Light metals 2013 : proceedings of the symposia sponsored by the TMS Aluminum Committee at the TMS 2013 Annual Meeting & Exhibition, San Antonio, Texas, USA March 3-7, 2013 978-3-319-65136-1, 3319651366

Table of contents :
Front Matter ....Pages i-xxviii
Front Matter ....Pages 1-1
Raw Material Impurities and the Challenge Ahead (Stephen J. Lindsay)....Pages 5-8
Impacts of Impurities Introduced into the Aluminium Reduction Cell (J. B. Metson, D. S. Wong, J. H. Hung, M. P. Taylor)....Pages 9-13
Changes in Global Refining and Its Impact on Anode Quality Petroleum Coke (Karl D. Bartholomew)....Pages 15-20
Impact of Higher Vanadium Levels on Smelter Operations (Chuck Coney, Lew Crabtree, John Gavin, Wes Marcrum, Andrea Weber, Les Edwards)....Pages 21-25
Impact on Smelter Operations of Operating High Purity Reduction Cells (Stewart Hamilton, Ray Cook)....Pages 27-32
Management of Impurities in Cast House with Particular Reference to Ni and V (M. A. Rhamdhani, J. F. Grandfield, A. Khaliq, G. Brooks)....Pages 33-38
An Initial Assessment of the Effects of Increased Ni and V Content in A356 and AA6063 Alloys (John Grandfield, Lisa Sweet, Cameron Davidson, Jason Mitchell, Aiden Beer, Suming Zhu et al.)....Pages 39-45
Front Matter ....Pages 47-47
Implementation of Logic Control by DCS to Measure the Caustic Concentration in Spent Liquor (Ayana Oliveira Santos, Aécio Carvalho, Bruno Urakawa, Milton Maciel, E. Antonio Santos)....Pages 51-53
Study of Influences on the Alumina/Caustic (A/C) Ratio and Discharge Digestion (DBO) Caustic of Through Design of Experiments (DOE) Statistic Tool (Américo Borges, Arthur Monteiro, Ayana Oliveira, Bruno Urakawa, Joel Miranda, Dauton Silva)....Pages 55-57
Particle size distribution model for kinetics of digesting alumina (Li Bao, Ting-an Zhang, Weimin Long, Anh V. Nguyen, Guozhi Lv, Jia Ma et al.)....Pages 59-64
Fractal kinetic model for digesting alumina (Li Bao, Ting-an Zhang, Anh V. Nguyen, Weimin Long, Jia Ma, Zhihe Dou et al.)....Pages 65-70
MAX HT® Bayer Sodalite Scale Inhibiter: A Green Solution to Energy Consumption (Morris Lewellyn, Alan Rothenberg, Calvin Franz, Frank Ballentine, Frank Kula, Luis Soliz et al.)....Pages 71-74
Sodalite Solids Formation at the Surface of Iron Oxide and Its Impact on Flocculation (Alexander Senaputra, Phillip Fawell, Franca Jones, Peter Smith)....Pages 77-82
Improvement on the Operation Management System of Vertical Pressure Filters (Tatiani Santos, Lucélia Moraes, Aline Sampaio, Milton Maciel, Humberto Lima, Joel Miranda et al.)....Pages 83-85
Using a Multivariate Statistical in the Identification of Alumina Loss in Red Mud (Alípio Júnior, Américo Borges, Ayana Oliveira, Humberto Lima, Joaquim Ribeiro, Joel Miranda et al.)....Pages 87-89
Bevill and the Aluminum Industry (TMS, Anthony E. Schoedel)....Pages 91-95
New Development Model for Bauxite Deposits - Dedicated Compact Refinery (Peter-Hans ter Weer)....Pages 97-102
Automatic Control of Drum Filters Operation (Aline Sampaio, Lucélia Moraes, Tatiani Santos, Humberto Lima, Américo Borges, Juarez Borges)....Pages 105-107
A New Technology for Dry Disposal of Alunorte’s Bauxite Residue (Marcelo Miranda de Castro, Carlos Roberto Abrunheiro Trindade, Ronaldo Guimarães Pantoja, Eduardo Raimundo Queiroz Alves Junior, Armênio Rabelo Martins)....Pages 109-112
Pilot Test of Bauxite Residue Carbonation with Flue Gas (Luis C. A. Venancio, José Antonio Silva Souza, Emanuel Negrão Macedo, Fernando Aracati Botelho, Glaucia Costa César)....Pages 113-118
Management of Industrial Waste: The Case of Effective Utilization of Red Mud and Fly Ash at Vedanta Aluminium Limited - Lanjigarh (Mukesh Kumar, Bimalananda Senapati, C. Sateesh Kumar)....Pages 119-123
Iron Recovery from Red Mud by Reduction Roasting-Magnetic Separation (Mingjun Rao, Jinqiang Zhuang, Guanghui Li, Jinghua Zeng, Tao Jiang)....Pages 125-130
Removal of Methylene Blue from Aqueous Solutions Using a Novel Granular Red Mud Mixed with Cement (Lu Shuaidan, Thiquynhxuan Le, Shaohua Ju, Peng Jin-hui, Zhang Li-bo)....Pages 131-136
Environmentally Safe Operation of Barometric Condensers (Matthew Jacobs)....Pages 139-141
Hatch - ETI Aluminyum Precipitation Modeling (E. Stamatiou, D. R. Chinloy, B. Çelikel, M. Kayaci, E. Savkilioglu)....Pages 143-146
Improve the Classification System in Hydro Alunorte Lines 4/5 (Emerson Moraes, Hans Haraldsen, Cleto Junior, Joaquim Ribeiro, Cesar Magro, Jose Chartouni et al.)....Pages 147-149
Increase in the Stability of Gravimetric Classification System of Precipitation at Hydro Alunorte (Victor Cruz, Emerson Moraes, Cleto Azevedo Junior, Denise Rodrigues, Adjane Souza, Alex Furtado et al.)....Pages 151-154
“Experience with Commissioning New Generation Gas Suspension Calciner” (Susanne Wind, Benny E. Raahauge)....Pages 155-162
Bayer Process Efficiency Improvement (Gu Songqing)....Pages 163-167
HyClass™ Technology for Improvement of Trihydrate Classification in the Bayer Process (Jing Wang, Jaqueline Herrera, Shawn Kostelak, Kody Frederic)....Pages 169-174
Metallic Impurities from the Mine to Metal Products (Stephen J. Lindsay)....Pages 177-181
The Control of Fluoride Concentration in ETİ Alüminyum Bayer Refinery Liquor (Esra Savkilioglu, Carl Carton, Serkan Ertugral, Meral Baygul, Kemal Dinc, Seyit Avcu)....Pages 183-186
Beneficiation of High Silica Bauxite Ores of India an Innovative Approach (Mukesh Kumar, Bimalananda Senapati, C. Sateesh Kumar)....Pages 187-190
Morphological Investigation of Sodium Oxalate Crystals Grown in Aqueous Sodium Hydroxide Solution (Weng Fu, James Vaughan)....Pages 191-194
Impurities in Raw Gas and Secondary Alumina (Svetlana Kalyavina, Arne Petter Ratvik, Thor Anders Aarhaug)....Pages 195-200
Innovative Technology for Alumina Production from Low-Grade Raw Materials (Alexander Senyuta, Andrey Panov, Alexander Suss, Yuri Layner)....Pages 203-208
Improving Characterization of Low Grade Diasporic Bauxite to be Utilize in Jajarm Alumina Plant (Mohammadtaghi Shadloo, Mohammad Zarbayani, Esmaeil Jorjani, Mojtaba Aram)....Pages 209-215
The Processing of High Quartz Bauxite (Edgar Gasafi, Alessio Scarsella, Vladimir Hartman, Hans. W. Schmidt)....Pages 217-221
Appropriate Reduction and Fe-Al Separation of High Iron Gibbsite (Liu Zheng-gen, Chu Man-sheng, Tang Jue, Han Yuan-ting, Wu Xiang-long)....Pages 223-227
Influence of MgO and C/A and Cooling System on Alumina Leaching Properties of Calcium Aluminate Slag (Zhifang Tong, Yingjie Li, Tao Chen)....Pages 229-232
Calcification - Carbonation Method for Alumina Production by Using Low-Grade Bauxite (Zhang Ting’an, Zhu Xiaofeng, Lv Guozhi, Pan Lu, Liu Yan, Zhao Qiuyue et al.)....Pages 233-238
Basic Research on Calcification Transformation Process of Low Grade Bauxite (Zhu Xiaofeng, Zhang Ting’an, Lv Guozhi, Liu Yan, Zhao Qiuyue, Dou Zhihe et al.)....Pages 239-244
Research on the phase transformation and separation performance in calcification - carbonation method for alumina production (Lv Guozhi, Zhang Ting’an, Zhu Xiaofeng, Pan Lu, Qin Mingxiao, Liu Yan et al.)....Pages 245-250
Front Matter ....Pages 251-251
Mechanical Properties of Al-Zn-Mg-Cu Alloys Processed with High-Pressure Torsion (Shigeru Kuramoto, Ichiro Aoi, Tadahiko Furuta)....Pages 255-258
High-Performance Be-Al Casting Alloys (Gary Schuster, Charles Pokross)....Pages 259-264
Structure Optimization of Al-Si-Type Alloys for Thermal and Mechanical High Loaded Components (Andreas Kleine, Marcel Rosefort, Ansgar Pithan, Christiane Matthies, Hubert Koch)....Pages 265-268
Development of High Strength Aluminium Alloys at BALCO (Mausumi Kar, Sachin Prasad, A. K. Paul, P. K. N. Raghavan)....Pages 269-275
Strength and Failure of Ultrafine Grain and Bimodal Al-Mg Alloy at High Temperatures (Andrew Magee, Leila Ladani)....Pages 279-282
Process development of AA3103 aluminum alloy for automotive thins (Marcelo Paes, Augusto D. Coelho, Roberto S. Netto, Fernando C. Aguiar)....Pages 283-287
Atom Probe Analysis of Sr Distribution in AlSi Foundry Alloys (Jenifer Barrirero, Michael Engstler, Frank Mücklich)....Pages 291-296
The Role of Sr on Microstructure Formation and Mechanical Properties of Al-Si-Cu-Mg Cast Alloy (Mohammadreza Zamani, Salem Seifeddine, Mona Aziziderouei)....Pages 297-302
Modification of the Eutectic Mg2si-Phase of AlMgSi-Cast Alloys (Thomas Pabel, Tose Petkov, Christian Kneissl, Peter Schumacher)....Pages 303-304
The influence of casting speed in the as cast strip mechanical properties of 8079 and 8006 alloys (Dionysios Spathis, John Tsiros)....Pages 305-309
Effect of Cooling Rate on Iron-rich Intermetallic Phases in 206 Cast Alloys (K. Liu, X. Cao, X.-G. Chen)....Pages 311-316
Effect of Iron in Al-Mg-Si-Mn Ductile Diecast Alloy (S. Ji, W. Yang, F. Gao, D. Watson, Z. Fan)....Pages 317-322
Oxidation Behavior of Al2Ca Added Al-5Mg Alloy in the Liquid State (Young-Ok Yoon, Seong-Ho Ha, Gil-Yong Yeom, Hyun Kyu Lim, Shae K. Kim)....Pages 323-326
Effect of the Thermal Modulus and Mould Type on the Grain Size of AlSi7Mg Alloy (I. Lizarralde, A. Niklas, A. I. Fernández-Calvo, J. Lacaze)....Pages 327-331
Alloy ALSi30 Cast in the Process of Rapid Solidification and Consolidated in the Process of Plastic Forming (Wojciech Szymański, Marcin Szymanek, Janusz Żelechowski, Mariusz Bigaj, Maciej Gawlik, Bartłomiej Płonka)....Pages 333-337
Effect of Homogenization Treatment Conditions on the Recrystallization Behavior of Al-1.2Mn Aluminum Alloy Sheets (Pizhi Zhao, Xinglin Chen, Wei Chen, Yonghao Zhang)....Pages 341-346
Toward a Recrystallized Microstructure in Extruded AA6005A Alloy (A. Bahrami, A. J. den Bakker, A. Miroux, J. Sietsma)....Pages 347-350
Grain Subdivision and Its Effect on Texture Evolution in an Aluminum Alloy Under Plane Strain Compression (Q. Ma, W. Mao, B. Li, P. T. Wang, M. F. Horstemeyer)....Pages 351-356
Fatigue Analysis of Ultrafine Grained Al 1050 Alloy Produced by Cyclic Forward Backward Extrusion (Hamid Alihosseini, Mohsen Asle Zaeem)....Pages 357-359
Effect of Zn Content and Process Parameters on Corrosion Behaviour of Twin-Roll Cast Aluminum Brazing Alloys (Murat Dündar, Mert Günyüz, Cemil Işiksaçan, Anıl Pastirmaci)....Pages 361-364
Growth ledges on silver-segregated θ′ (Al2Cu) precipitates (Julian Rosalie, Laure Bourgeois)....Pages 367-371
On the Aging Behavior of AA2618 DC Cast Alloy (P. Shen, E. M. Elgallad, X.-G. Chen)....Pages 373-377
The Effect of Cold Work on the Precipitation and Recrystallization Kinetics in Al-Sc-Zr Alloys (C. T. McNamara, S. L. Kampe, P. G. Sanders, D. J. Swenson)....Pages 379-382
A Novel Solution Heat Treatment of 7075-Type Alloy (M. F. Ibrahim, A. M. Samuel, S. A. Alkahtani, F. H. Samuel)....Pages 383-390
Experimental study of the Al-rich corner of the Al-Si-Ti system at 500 °C (Yang Li, Qun Luo, Jie-Yu Zhang, Qian Li)....Pages 391-393
Transient microstructural thermomechanical fatigue and deformation characteristics under superimposed mechanical and thermal loading, in AlSi based automotive diesel pistons (Roman Morgenstern, Scott Kenningley)....Pages 397-403
Mechanical Behaviour of Cold Formed Metal-Polymer Laminate and the Interaction of Its Layers (Feidhlim Ó Dubhlaing, David J. Browne, Robin Rennicks, Connor Rennicks)....Pages 405-410
Mechanical and Tribological Properties of AA2124-Graphene Self Lubricating Nanocomposite (A. Ghazaly, B. Seif, H. G. Salem)....Pages 411-415
Joining Vacuum High Pressure Die Cast A356 Under T4 Treatment to Wrought Alloy 6061 (Meng Wang, Yanda Zou, Henry Hu, Gary Meng, Patrick Cheng, Yeou-Li Chu)....Pages 417-421
Applications of the Horizontal Squeeze Casting Process for Automotive Parts Manufacturing (P. Dulyapraphant, E. Kittikhewtraweeserd, P. Kritboonyarit, N. Denmud)....Pages 425-429
Characterization of the Developed Precipitates in Al-2 at.%Zn - x at.%Mg, (x=1.8, 2, 2.4, 3, 4.2) (N. Afify, A. Gaber, Gh. Abbady)....Pages 431-436
Design and Development of a Permanent Mould for the Production of Motor-Cycle Piston in SEDI-Enugu. (C. E. Ilochonwu, E. I. Nwonye)....Pages 437-441
Development and Research of New Aluminium Alloys with Transition and Rare-Earth Metals and Equipment for Production of Wire for Electrotechnical Applications by Methods of Combined Processing (I. Matveeva, N. Dovzhenko, S. Sidelnikov, L. Trifonenkov, V. Baranov, E. Lopatina)....Pages 443-447
Influence of Machining Parameters on Al-4.5Cu-TiC In-Situ Metal Matrix Composites (Pradeep Kumar Jha, Anand Kumar, M M Mahapatra)....Pages 449-452
Effect of Mg Contents on Fluidity of Al-xMg Alloys (Nam-Seok Kim, Seong-Ho Ha, Young-Ok Yoon, Gil-Yong Yeom, Hyun Kyu Lim, Shae K. Kim)....Pages 453-456
Effect of Process Parameters on Centrifugally Cast Al-Si FGM (Aithal S. Kiran, Vijay Desai, S. Narendranath, P G Mukunda)....Pages 457-462
Effects of Minor Sc Addition on the Microstructures and Mechanical Properties of Al-Zn-Mg-Cu Casting Aluminum Alloy (Yang Guangyu, Liu Shaojun, Jie Wanqi)....Pages 463-467
Microhardness, Corrosion Behaviour and Microstructures of Directionally Solidified Al-Cu Alloys (Alicia Ares, Carlos M. Rodriguez, Claudia M. Mèndez, Carlos E. Schvezov, Mario R. Rosenberger)....Pages 469-473
Production of Single Cylinder Engine Components Through High Pressure Die Casting in Sedi Enugu (E. I. Nwonye, C. E. Ilochonwu, C. O. Nwajagu)....Pages 475-479
The Effect of Thermomechanical Ageing of Aluminium-Copper Alloy (MATLAB Approach) (Adegbola Adekunle Amos, Ghazali Akeem, Fashina Olugbenga Emmanuel, Omotoyinbo Joseph Ajibade, Olaniran Oladayo)....Pages 481-486
Front Matter ....Pages 487-487
Surface Crack Characterization of Twin Roll Caster Shells and Its Influence on As-Cast Strip Surface Quality (Murat Dündar, Barış Beyhan, Onur Birbaşar, Hatice M. Altuner, Cemil Işıksaçan)....Pages 491-495
The Effect of Magnesium Content on Microstructure Evolution During Hot Deformation of Aluminum Alloys (Trevor J. Watt, Shinya Yasuda, Koji Ichitani, Ken Takata, Alex Carpenter, Jakub Jodlowski et al.)....Pages 499-503
High Strength Nanostructured Al-Zn-Mg-Cu-Zr Alloy Manufactured by High-Pressure Torsion (Chao An, Huimin Lu, Shilai Yuan)....Pages 505-508
Corrosion Behavior of 2024 Aluminum Alloy Anodized in Sulfuric Acid Containing Inorganic Inhibitor (Maysam Mohammadi, Ali Yazdani, Farzad Mohammadi, Akram Alfantazi)....Pages 509-513
Laboratory simulation of wear during hot extrusion of aluminium (G. Kugler, M. Terčelj)....Pages 515-520
The Production of Wrought AlSi30Cu1.5Mg1.2Ni1.5Fe0.8 Alloy with Ultrafine Structure (Marcin Szymanek, Bogusław Augustyn, Wojciech Szymański, Dawid Kapinos)....Pages 521-525
The Structure and Properties of Wrought Aluminium Alloys Series 6xxx with Vanadium for Automotive Industry” (Marzena Lech-Grega, W. Szymański, B. Płonka, S. Boczkal, M. Gawlik, M. Bigaj et al.)....Pages 527-532
Front Matter ....Pages 533-533
In Depth Analysis of Energy-Saving and Current Efficiency Improvement of Aluminum Reduction Cells (Yan Feiya, Marc Dupuis, Zhou Jianfei, Ruan Shaoyong)....Pages 537-542
Rio Tinto Alcan AP4X Low Energy Cell Development (P. Thibeault, S. Bécasse, A. Blais, P. Côté, L. Fiot, F. Laflamme)....Pages 543-547
Energy Reduction Technology for Aluminum Electrolysis: Choice of the Cell Voltage (Feng Naixiang, Peng Jianping, Wang Yaowu, Di Yuezhong, Liao Xian’an)....Pages 549-552
Advancements of Dubal High Amperage Reduction Cell Technologies (Michel Reverdy, Abdalla Zarouni, Jean-Luc Faudou, Qassim Galadari, Ali Al Zarouni, Sergey Akhmetov et al.)....Pages 553-556
Development of Low-Voltage Energy-Saving Aluminum Reduction Technology (Li Jie, Lv Xiao-jun, Zhang Hong-liang, Liu Ye-xiang)....Pages 557-559
D18+: Potline Modernisation at DUBAL (Sergey Akhmetov, Daniel Whitfield, Maryam Mohammad Al-Jallaf, Ali Al-Zarouni, Alexander Arkhipov, Amer Al-Redhwan et al.)....Pages 561-565
Industry Test of Perforation Anode in Aluminium Electrolysis Technology (Yingfu Tian, Hesong Li, Longhe Wei, Xi Cao, Jianguo Yin)....Pages 567-571
The First Results of the Indus Trial Application of the Ecosøderberg Technology at the Krasnoyarsk Aluminum Smelter (Victor Buzunov, Victor Mann, Evgeniy Chichuk, Vladimir Frizorger, Andrey Pinaev, Evgeniy Nikitin)....Pages 573-576
Unsteady MHD modeling applied to cell stability (Renaudier Steeve, Bardet Benoit, Steiner Gilles, Pedcenko Alex, Rappaz Jacques, Molokov Sergeï et al.)....Pages 579-584
Impact of magnetohydrodynamic and bubbles driving forces on the alumina concentration in the bath of an Hall-Héroult cell (René von Kaenel, Jacques Antille, Michel V. Romerio, Olivier Besson)....Pages 585-590
Investigation of Electrolytic Bubble Behaviour in Aluminum Smelting Cell (Morshed Alam, Yos Morsi, William Yang, Krishna Mohanarangam, Geoff Brooks, John Chen)....Pages 591-596
Mathematical Model Validation of Aluminium Electrolysis Cells at DUBAL (Abdalla Zarouni, Lalit Mishra, Marwan Bastaki, Amal Al Jasmi, Alexander Arkhipov, Vinko Potocnik)....Pages 597-602
Production Application Study on Magneto-Hydro-Dynamic Stability of a Large Prebaked Anode Aluminum Reduction Cell (Ruan Shaoyong, Yan Feiya, Marc Dupuis, Valdis Bojarevics, Zhou Jianfei)....Pages 603-607
MHD of Aluminium Cells with the Effect of Channels and Cathode Perturbation Elements (Valdis Bojarevics)....Pages 609-614
Magnetohydrodynamic Model Coupling Multiphase Flow in Aluminum Reduction Cell with Innovative Cathode Protrusion (Wang Qiang, Li Baokuan, Wang Fang, Feng Naixiang)....Pages 615-619
Optimization of the Cathode Collector Bar with a Copper Insert Using Finite Element Method (Mathieu Gagnon, Patrice Goulet, Richard Beeler, Donald Ziegler, Mario Fafard)....Pages 621-626
Energy Savings in Aluminum Electrolysis Cells: Effect of the Cathode Design (Mathieu Blais, Martin Désilets, Marcel Lacroix)....Pages 627-631
Low Power Operation at Aluminium Dunkerque Smelter (Jean-Michel Peyneau, Laurent Fiot, Stéphane Mermet-Guyenet, Olivier Rebouillat)....Pages 635-639
Maximizing Creeping Value through Rigorous Methodology (Bénédicte Champel, Nicolas Monnet)....Pages 641-645
The Quick Shut Down and Restarting of 291 kA Pre-Baked Potline at JSC “RUSAL Sayanogorsk» from May to August 2011 (Victor Buzunov, Andrey Soldatov, Victor Mann, Aleksandr Pavin, Vasily Borisov, Sergey Zatepyakin et al.)....Pages 647-652
Production Growth and Future Challenges in Aluminium Bahrain (Alba) (Isa Al-Ansari, Abdulla Habib, A. C. Mittal, Nabeel Al-Jallabi)....Pages 653-658
High Frequency Power Modulation - TRIMET smelters provide primary control power for stabilizing the frequency in the electricity grid (Andreas Lützerath)....Pages 659-662
Autonomous Vehicle and Smelter Technologies (Ashley Tews, Paulo Borges)....Pages 663-668
Preventive Maintenance of Transport Vehicles is it improving production stability of a smelter? (Maarten Meijer)....Pages 669-671
Composition and Thermal Analysis of Crust Formed from Industrial Anode Cover (Qinsong Zhang, Mark P. Taylor, John J. J. Chen, David Cotton, Tania Groutzo, Xiaodong Yang)....Pages 675-680
Liquidus Temperatures of Na3AlF6-AlF3-CaF2-KF-LiF-Al2O3 Melts (Di Yuezhong, Peng Jianping, Bai Yunbin, Feng Naixiang)....Pages 681-684
The Effect of Calcium Fluoride on Alumina Solubility in Low Temperature Cryolite Melts (P. Tingaev, Yu. Zaikov, A. Apisarov, A. Dedyukhin, A. Redkin)....Pages 685-688
Conductivity of KF-NaF-AlF3 System Low-temperature Electrolyte (Jianhong Yang, Wangxing Li, Hengwei Yan, Dan Liu)....Pages 689-693
Numerical analysis of ionic mass transfer in the electrolytic bath of an aluminium reduction cell (Mohsen Ariana, Martin Désilets, Pierre Proulx)....Pages 695-699
Liquidus Temperature of Electrolytes for Aluminum Reduction Cells (Dong Shi, Bing-liang Gao, Zhao-wen Wang, Zhong-ning Shi, Xian-wei Hu)....Pages 701-704
Effect of LiAlO2 and KF on Physicochemical Properties for Industrial Aluminum Electrolyte (Lv Xiaojun, Chen Shiyue, Lai Yanqing, Tian Zhongliang, Li Jie, Zhang Hongliang)....Pages 705-709
Improvement of Alumina Dissolution Rate through Alumina Feeder Pipe Modification (Jayson Tessier, Gary P. Tarcy, Eliezer Batista, Xiangwen Wang, Patrice Doiron)....Pages 713-717
Reduction Cell Restart Method Influence on Cell Life Evolution (Mikhail Lukin, Richard Jeltsch)....Pages 719-724
Start of an Aluminum Reduction Cell without Liquid Bath (Kayron F. Lalonde, Brian D. Audie, Willy Kristensen, Timothy M. Snyder)....Pages 725-728
A MIMO Modeling Strategy for Bath Chemistry (Fabio M. Soares, Roberto C. L. Oliveira)....Pages 729-734
Cumulative Distributions of Metallic Impurities (Stephen J. Lindsay)....Pages 735-739
Sodium Content in Aluminum and Current Efficiency — Correlation through Multivariate Analysis (Lukas Dion, László Kiss, Patrice Chartrand, Gilles Dufour, François Laflamme)....Pages 741-746
Gas-Solid Flow Applications for Powder Handling in Aluminum Smelters Processes (Paulo Douglas S. de Vasconcelos, André L. Amarante Mesquita)....Pages 747-752
Operational Experience of Advanced Alumina Handling Technology in a Russian Smelter (Jan Paepcke, Arne Hilck, Sergey V. Marshalko)....Pages 753-759
Reduction in HF emission through improvement in operational practices (H. R. Devadiga, Ali Jassim Banjab, Maryam Mohamed Al Jallaf, Ali H. A. M. Al Zarouni, Kamel Al Aswad, A. Kumar et al.)....Pages 763-767
Trace Element Concentration in Particulates from Pot Exhaust and Depositions in Fume Treatment Facilities (Heiko Gaertner, Arne Petter Ratvik, Thor Anders Aarhaug)....Pages 769-774
The Study and Applications of Modern Potline Fume Treatment Plant (FTP) (Deng Xiang, Lv Weining, Liu Xun, Deng Qiyi, Yi Xiaobing)....Pages 775-780
F>C: Combined Treatment of Pot Gases and Anode Baking Furnace Fumes (B. Hureiki, C. Lim, A. Periers, E. Bouhabila, G. Girault, M. Leduc et al.)....Pages 781-785
Compact Filter Design for Gas Treatment Centers (Peter Verbraak, Peter Klut, Travis Turco, Erik Dupon, Edo Engel)....Pages 787-792
An Innovative Compact Heat Exchanger Solution for Aluminum Off-Gas Cooling and Heat Recovery (El Hani Bouhabila, Erling Næss, Victoria Kielland Einejord, Kolbeinn Kristjansson)....Pages 793-797
Latest Filter Developments Increasing Existing Aluminium Smelter Gas Treatment Centre Capacity and Reducing Emissions (Michael Neate, Brad Currell)....Pages 799-804
Reduced Ventilation of Upper Part of Aluminum Smelting Pot: Potential Benefits, Drawbacks, and Design Modifications (Ruijie Zhao, Louis Gosselin, Mario Fafard, Donald P. Ziegler)....Pages 805-810
Latest developments in potroom building ventilation CFD modelling (Nathalie Menet, Guillaume Girault, Nicolas Monnet, Catherine Turpin, Lionel Soulhac)....Pages 811-816
Solutions to Address Arc Welding Problems in an Operating Potline (Bill Paul, Yann El Ghaoui, Philippe Jadaud, John Anderson, Stephen A. L. Foulds)....Pages 819-822
Replacement of Damaged Electrical Insulators on Live Cross-Over Busbars inside a Tunnel: A Methodology Based on Risk Assessment and Numerical Simulation (Daniel Richard, André Yelle, Olivier Charette, Andre Felipe Schneider, Jean-François Nadeau, Mickael Glière et al.)....Pages 823-828
A Thermal-Mechanical Approach for the Design of Busbars Details (Andre Felipe Schneider, Olivier Charette, Daniel Richard, Charles Turcotte)....Pages 829-835
Study of Technology and Equipment on Magnetic Induction Intensity Weaken for Aluminum Reduction Cells Welding in the Condition of Pot Line Current (Wang Ziqian, Cao Bin, Yang Tao, Huang Jun, Li Meng)....Pages 837-841
Potline Shutdown and Restart Secured Solutions (Anne-Gaëlle Hequet)....Pages 843-844
Effect of Watering and Non-Watering Cooling Rates on the Mechanical Properties of an Aluminum Smelter’s Potshell (Ayoola T. Brimmo, Mohamed I. Hassan, M. O. Ibrahiem, Youssef Shatilla)....Pages 845-850
Mathematical Model of Cooling of a Stopped Pot and Its Validation (Mohamed I. Hassan, Ayoola T. Brimmo, M. O. Ibrahiem, Youssef Shatilla)....Pages 851-855
A Study of Low Voltage PFC Emissions at Dubal (Abdalla Zarouni, Michel Reverdy, Ali Al Zarouni, K. G. Venkatasubramaniam)....Pages 859-863
Continuous PFC Emissions Measured on Individual 400kA Cells (David S. Wong, Jerry Marks)....Pages 865-870
PFC and Carbon Dioxide Emissions from an Australian Aluminium Smelter Using Time-Integrated Stack Sampling and GC-MS, GC-FID Analysis (Paul Fraser, Paul Steele, Mark Cooksey)....Pages 871-876
Investigation on Formation Mechanism of Non-Anode Effect Related PFC Emissions from Aluminum Reduction Cells (Chen Xiping, Li Wangxing, Zhang Yanfang, Qiu Shilin, Chris Bayliss)....Pages 877-881
On the Mechanism Behind Low Voltage PFC Emissions (Jomar Thonstad, Sverre Rolseth, Rudolf Keller)....Pages 883-885
Frequency response analysis of anode current signals as a diagnostic aid for detecting approaching anode effects in aluminum smelting cells (C. Cheung, C. Menictas, J. Bao, M. Skyllas-Kazacos, B. J. Welch)....Pages 887-892
Reduction Strategies for PFC Emissions from Chinese Smelters (Li Wangxing, Chen Xiping, Qiu Shilin, Zhang Baowei, Chris Bayliss)....Pages 893-898
Off-Gas Analysis of Laboratory-Scale Electrolysis Experiments with Anodes of Various Compositions (Ole S. Kjos, Thor Anders Aarhaug, Egil Skybakmoen, Asbjørn Solheim, Henrik Gudbrandsen)....Pages 899-903
Hydrolysis of Carbonyl Sulfide (COS) on Smelting Grade Alumina (Aleksandr V. Mikhonin, Neal R. Dando, Michael Gershenzon)....Pages 905-908
A Thermodynamic Approach to the Corrosion of the Cathode Refractory Lining in Aluminium Electrolysis Cell: Modelling of the Al2O3-Na2O-SiO2-AlF3-NaF-SiF4 System (Guillaume Lambotte, Patrice Chartrand)....Pages 911-916
Effect of Current Density and Phosphorus Impurities on the Current Efficiency for Aluminum Deposition in Cryolite-Alumina Melts in a Laboratory Cell (Rauan Meirbekova, Gudrun Saevarsdottir, Geir Martin Haarberg, Joseph Prince Armoo)....Pages 917-920
Front Matter ....Pages 921-921
Very High Purity Ingot — An Endangered Species? (Stephen J. Lindsay)....Pages 925-928
The Challenge of Effectively Utilizing Trace Elements/Impurities in a Varying Raw Materials Market (Gyan Jha, Shridas Ningileri, Xiaoxuan Li, Randall Bowers)....Pages 929-934
Energy control in primary aluminium casthouse furnaces (Inge Johansen, Svenn Ivar Strømhaug)....Pages 935-939
Metal Contamination Associated with Dross Processing (Ray D. Peterson)....Pages 941-946
Ultrasonic Degassing and Processing of Aluminum (Victor Rundquist, Kiran Manchiraju)....Pages 949-955
Kinetics of Ultrasonic Degassing of Aluminum Alloys (Noé Alba-Baena, Dmitry Eskin)....Pages 957-962
Removal of Inclusions in Molten Aluminum by Flux Injection Under Counter-Gravity (Jianmin Zeng, Hong Gu)....Pages 963-965
Advanced Compact Filtration (ACF): An Efficient and Flexible Filtration Process (Francis Breton, Peter Waite, Patrice Robichaud)....Pages 967-972
Electromagnetic Priming of Ceramic Foam Filters (CFF) for Liquid Aluminum Filtration (Robert Fritzsch, Mark William Kennedy, Jon A. Bakken, Ragnhild E. Aune)....Pages 973-979
Plant Scale Investigation of Liquid Aluminium Filtration by Al2O3 and SiC Ceramic Foam Filters (Sarina Bao, Martin Syvertsen, Arne Nordmark, Anne Kvithyld, Thorvald Engh, Merete Tangstad)....Pages 981-986
Casting Practices Influencing Inclusion Distributions in Billets (Ghadir Razaz, Torbjörn Carlberg)....Pages 987-991
Oxidation of Commercial Purity Aluminum Melts: An Experimental Study (Stephen J. Bonner, John A. Taylor, Ji-Yong Yao, M. Akbar Rhamdhani)....Pages 993-997
Optimisation of Grain Refinement (Rein Vainik, John Courtenay, Bader Saglam)....Pages 1001-1008
Grain Refiner for Al-Si Alloys (Hari Babu Nadendla, Magdalena Nowak, Leandro Bolzoni)....Pages 1009-1012
Production of Al-Ti-B Grain Refining Master Alloys from B2O3 and K2TiF6 by Microwave Irradiation (Zhou Shu-cai)....Pages 1013-1016
Effects of Yb Additions on Refinement of Eutectic Si in Al-5Si Alloys (J. H. Li, P. Schumacher)....Pages 1017-1022
Influence of Vanadium on the Microstructure of A356 Foundry Alloy (Thomas H. Ludwig, Paul L. Schaffer, Lars Arnberg)....Pages 1023-1028
Influence of die and casting temperatures and titanium and strontium contents on the technological properties of die-cast A356 in the as-cast and T6 condition (Sebastian F. Fischer, Veronika F. Groten, Johannes Brachmann, Carolin Fix, Thomas Vossel, Andreas Bührig-Polaczek)....Pages 1031-1036
Horizontal Single Belt Strip Casting (HSBC) of Al-Mg-Sc-Zr Alloys (Mert Celikin, Donghui Li, Luis Calzado, Mihaiela Isac, Roderick I. L. Guthrie)....Pages 1037-1040
Front Matter ....Pages 1041-1041
Review of Different Techniques to Study the Interactions Between Coke and Pitch in Anode Manufacturing (Duygu Kocaefe, Arunima Sarkar, Shipan Das, Salah Amrani, Dipankar Bhattacharyay, Dilip Sarkar et al.)....Pages 1045-1050
Observations on the Coke Air Reactivity Test (Keith Neyrey, Les Edwards, James Marino)....Pages 1051-1056
Impurity Removal from Petroleum Coke (Alexandre Gagnon, Nigel Backhouse, Hans Darmstadt, Esmé Ryan, Laurence Dyer, David G. Dixon)....Pages 1057-1062
Calcined Coke Round Robin 19 and the Precision of Bulk Density Tests (Marvin Lubin, Les Edwards, Lorentz Petter Lossius)....Pages 1063-1068
A Method for the Rapid Characterization of Petroleum Coke Microstructure Using Polarized Light Microscopy (Andris Innus, Alain Jomphe, Hans Darmstadt)....Pages 1069-1073
Improvements of Vibrated Bulk Density Analysis at VM-CBA and Petrocoque S.A (Jean Carlos Pardo, Edinaldo Pereira da Silva, Paulo da Silva Pontes, André Nantes)....Pages 1075-1078
Influence of GPC Properties on the CPC Quality (Zhao Jingli, Zhao Qingcai, Zhao Qingbo, Yu Lei, Yu Pusheng)....Pages 1079-1083
Quality of Calcined Petroleum Coke and Its Influence on Aluminium Smelting (José Subero)....Pages 1085-1088
A Green Anode Plant Performance Analysis Tool Fully Embedded in the Plant Control System (Xavier Genin, Pasquale Calò)....Pages 1091-1096
Measures to Prevent Baked Anode Density Drop When Using High Porosity Cokes (Vinicius Piffer, Chin Woo, Fabiana Nicéas, Rafael Bacelar, Jeronimo Araujo, Leonardo Paulino)....Pages 1097-1100
New Green Anode Plant at EMAL Start-Up & Operation in the First Two Years (Raja Javed Akhtar, Rudolf Gemein, Manfred Beilstein)....Pages 1101-1104
Improving Anode Baked Density and Air Permeability through Process Optimization and Coke Blending (Bienvenu Ndjom, Muhammad Shafik Malik, Amer Al Marzouqi, Tapan Kumar Sahu, Saleh Ahmed Rabba)....Pages 1105-1110
Development of an Analytical Dynamic Model of a Vibro-Compactor Used in Carbon Anode Production (Fatma Rebaïne, Mohamed Bouazara, Daniel Marceau, Duygu Kocaefe, Brigitte Morais)....Pages 1111-1115
Driving Cost Reduction and Carbon Plant Productivity Improvement through Theory of Constraints and Planned Maintenance Capability (Keith A. Sinclair, Barry A. Sadler)....Pages 1117-1122
Optimum Vibration Time for Green Anode Production (Shoulei Gao, Chongai Bao, Shoujun Zhang, Huanxue Wang, Joe Woo, Euel Cutshall)....Pages 1123-1126
Comparison of Mixing Process Methods in Prebaked Anode Production (Sun Yi, Guan Huai, Zhou Shanhong, Liu Chaodong, Xu Haifei)....Pages 1127-1130
Hydro Aluminium’s Historical Evolution of Closed Type Anode Baking Furnace Technology (Michal Tkac, Anders Ruud, Inge Holden, Hogne Linga)....Pages 1133-1138
Use of Mathematical Modelling to Study the Behavior of a Horizontal Anode Baking Furnace (Yasar Kocaefe, Noura Oumarou, Mounir Baiteche, Duygu Kocaefe, Brigitte Morais, Marc Gagnon)....Pages 1139-1144
Study on Anode Baking Parameres in Open-Top and Closed-Type Ring Furnaces (Borzu Baharvand, Mohesn Ameri Siahouei, Mohammad Nabi Batoei, Saeb Sadeghi)....Pages 1145-1150
Energy Efficiency Improvement in Anode Baking Furnaces (Cassio Linhares, Fabiana Niceas, Rafael Bacelar, Helcio Campelo, Marcos A. Silva, Jeronimo Araujo et al.)....Pages 1151-1154
Anode Baking Process Improvement at ALRO (Pierre Mahieu, Nicolas Fiot, Arnaud Trillat, Ovidiu Balu, Cristian Stanescu)....Pages 1155-1161
Operational and Environmental Benefits of the New Baking Furnace at Boyne Smelters by Use of an Advanced Firing Technology (Andreas Himmelreich, Detlef Maiwald, Domenico Di Lisa, Glenn Cordon, Sathya Moodley)....Pages 1163-1168
Laser Mapping of Carbon Bake Furnaces (Ashley Tews, Mike Bosse, Robert Zlot, Paul Flick, Meaghan Noonan)....Pages 1169-1174
Pilot Scale Anodes for Raw Material Evaluation and Process Improvement (Lorentz Petter Lossius, Juraj Chmelar, Inge Holden, Hogne Linga, Michal Tkac)....Pages 1177-1182
Relationships between Coke Properties and Anode Properties — Round Robin 19 (Lorentz Petter Lossius, Marvin Lubin, Les Edwards, Julien Wyss)....Pages 1183-1188
Application of the Artificial Neural Network (ANN) in Predicting Anode Properties (Dipankar Bhattacharyay, Duygu Kocaefe, Yasar Kocaefe, Brigitte Morais, Marc Gagnon)....Pages 1189-1194
A Model for Predicting the Electrical Resistivity of Baked Anodes (Dipankar Bhattacharyay, Duygu Kocaefe, Yasar Kocaefe, Brigitte Morais, Marc Gagnon)....Pages 1195-1200
The Role of Electrode Quality in Metal Purity (Stephen J. Lindsay)....Pages 1201-1205
Electrochemical Characterization of Carbon Anode Performance (Rebecca Jayne Thorne, Camilla Sommerseth, Espen Sandnes, Ole Kjos, Thor Anders Aarhaug, Lorentz Petter Lossius et al.)....Pages 1207-1211
High Capacity Thermobalance Anode Reactivity Testing (Nick Janssen, James Baker, Frank Cannova, Barry Sadler)....Pages 1213-1218
Diagnosing Changes in Baked Anode Properties using a Multivariate Data-driven Approach (Julien Lauzon-Gauthier, Carl Duchesne, Jayson Tessier)....Pages 1219-1223
Evolution of the Thermo-Mechanical Properties of Ramming Paste from Ambient to Operating Temperature in a Hall-Heroult Cell (Stephane Tremblay, Lyne St-Georges, Laszlo Kiss, Lyès Hacini, Bénédicte Allard, Daniel Marceau)....Pages 1227-1231
New Compaction Method for the Production of Large Ramming Paste Samples for 3D Mechanical Characterization (Pierre-Olivier St-Arnaud, Donald Picard, Maxime Noël, Houshang Alamdari, Donald Ziegler, Mario Fafard)....Pages 1233-1238
Technology for Manufacturing Cathodes for Aluminium Reduction in China (Yang Hongjie, Liu Fengqin, Cai Suwei, Yang Xiaopei)....Pages 1239-1243
The Effect of Cryolite on the Formation of Aluminum Carbide at the Carbon Aluminum Interface (B. Novak, K. Tschöpe, A. P. Ratvik, T. Grande)....Pages 1245-1250
Critical Reflections on Laboratory Wear Tests for Ranking Commercial Cathode Materials in Aluminium Cells (Kati Tschöpe, Anne Støre, Egil Skybakmoen, Asbjørn Solheim, Tor Grande, Arne Petter Ratvik)....Pages 1251-1256
Model for Excessive Cathode Wear By a “Carbon Pump” at the Cell Bottom (Asbjørn Solheim, Kati Tschöpe)....Pages 1257-1262
Characterization of Porous Structure and its Correlation to Sodium Expansion of Graphite Cathode Materials Using Image Analysis (Xiang Li, Jilai Xue, Tong Chen)....Pages 1263-1267
Studies on the Resistance to Alkali Metal Penetration of Binders for TiB2-C Composite Cathode Materials (Fang Zhao, Zhang Kai, Lai Yan-qing, Li Lin-bo, Zhu Jun)....Pages 1269-1274
Mechanically alloyed Cu-Ni-Fe-Y material as inert anode for Al production (V. Ouvarov-Bancalero, D. Guay, L. Roué)....Pages 1277-1281
Cold Spray deposition of mechanically alloyed Cu-Ni-Fe material for application as inert anodes for aluminum production (G. Goupil, S. Helle, E. Irissou, D. Poirier, J. G. Legoux, D. Guay et al.)....Pages 1283-1287
Initial 1000A Aluminum Electrolysis Testing in Potassium Cryolite-Based Electrolyte (John Hryn, Olga Tkacheva, Jeff Spangenberger)....Pages 1289-1294
Electrochemical Behavior of Cermet Anodes in Na3AlF6-K3AlF6-Based Low-Melting Electrolytes for Aluminium Electrolysis (Guihua Wang, Xiaofei Sun)....Pages 1295-1298
Production of Aluminum Sulfide through Carbosulfidation Utilising H2S (Nazmul Huda, M. A. Rhamdhani, G. A. Brooks, B. J. Monaghan, L. Prentice)....Pages 1299-1304
Microstructural Evolution of Cast Iron Used for Cathode Rodding in Aluminium Electrolysis Cell (Alireza Hekmat-Ardakan, Gervais Soucy, Loig Rivoaland)....Pages 1305-1309
Preparing Al-Sc-Zr Alloys in Aluminum Electrolysis Process (Yi Qian, Jilai Xue, Qiaochu Liu, Jun Zhu)....Pages 1311-1314
Fume Treatment Systems Based on RTO Technology for Carbon Baking Furnaces (Matthias Hagen, Bernd Schricker)....Pages 1317-1322
AHEX- A New, Combined Waste Heat Recovery and Emission Control System for Anode Bake Furnaces (A. Sørhuus, S. Ose, G. Wedde)....Pages 1323-1328
Successful Start-Up of the Fume Treatment Centre at Boyne Smelter Carbon Bake Furnace #4 (Jonathan Higley, Glenn Cordon, Peter Klut, Rick Oliana, Erik Dupon, Travis Turco et al.)....Pages 1329-1334
Thermo-electro-mechanical characterization of anode interfaces at operating conditions (Hugues Fortin, Marie-Hélène Martin, Nédeltcho Kandev, Guillaume Gauvin, Donald Ziegler, Mario Fafard)....Pages 1335-1340
A Fully Coupled Thermal-Electrical-Mechanical Transient FEA Model for a 3D Anode Assembly (D. R. Gunasegaram, D. Molenaar)....Pages 1341-1346
Experimental and Numerical Investigation of Voltage Drop in Anode Assemblies (Ebrahim Jeddi, Daniel Marceau, Laszlo I. Kiss, Lyne St-Georges, Denis Laroche, Lyès Hacini)....Pages 1347-1352
Optimization of the Anode-Stub Contact: Effect of Casting Temperature, Contact Stress and Temperature and Surface Roughness (Bjarte Oye, Anne Store, Elin Haugland, Jorund Hop)....Pages 1353-1357
Experimental Investigation of Factors Affecting the Electrical Performance of the Stub to Carbon Connection (D. Molenaar, T. Kilpatrick, A. Montalto)....Pages 1359-1364
Back Matter ....Pages 1365-1377

Citation preview

13 Edited by BARRY A. SADLER

Light Metals 2013

142nd Annual Meeting & Exhibition

Light Metals 2013

Proceedings of the symposia sponsored by the TMS Aluminum Committee at the TMS 2013 Annual Meeting & Exhibition, San Antonio, Texas, USA March 3-7,2013

Edited by Barry A. Sadler

Editor Barry A. Saddler

Additional matr rialto this book ran be dOll"nloadrd from hllp:/Iextras.springer.com ISBN 978-3-3 19-65 135-4 DOr 10.1007/978-3-3 19-65136- 1

ISBN 978-3-3 19-65136- 1 (cBook)

Chemistry and Materials ScielU;c: Professional Copyright " 20 16 by The Minerals, Metals & Materials Society rublished by Springer hnemational Publishers, Switzer land, 2016 Reprint of the original edition published by John Wiley & Sons, Inc. , 20 13, 978· 1·11860-572·1 This work is subject to copyright. All rights arc reserved by the Publisher, whether the whole or part of the materiul is concerned, specificall y the rights of truns lation. repr inting. reuse of illustrations, recitation, broadcasting. reproduction on microfilms or in any other physical way, and transmission or information storage and retrie llal. electronic adaptat ion. computer software. or by simi lar or dissi milar methodology noll' known or hereafte r devel oped. The usc of general descriptive names, registered names, trademarks, seT\liee marks, etc. in this puhlicati on does not imply, even in the absence o f a specirlc statement, that such names are exem pt from the relevant protective laws and regu lutions und therefore frc;:e for general usc. The publisher, the authors and the editors are safe to assume that the advice and information in this book are he lieveJ to he true and aCCUnlte at the date of puhlieation. Nei ther the publisher nor the authors or the editors give a lI"urranty. express or implied. with respect to the mater ial containcd hcrein or for any crrors or omissions th at llIay havc bccllmadc. Printed on acid·rree paper This Sprin ger imprint is publisheJ by Springer Natu re Thc rcgistcred company is Springer Intcrnational Publishing AG The registcreJ company address is: Gewcrbcstrassc II. 6330 Cham, SwilZCrlanJ

TABLE OF CONTENTS Light Metals 2013 Preface ........................................................................................................................................................................ xxi About the Editor ....................................................................................................................................................... xxiii Program Organizers ................................................................................................................................................... xxv Aluminum Committee ............................................................................................................................................. xxvii

2013 Aluminum Keynote: Impurities in the Aluminum Supply Chain Keynote Session Raw Material Impurities and the Challenge Ahead ....................................................................................................... 5 S. Lindsay Impacts ofImpurities Introduced into the Aluminium Reduction Cell ......................................................................... 9 J Metson, D. Wong, J Hung, and M Taylor Changes in Global Refining and Its Impact on Anode Quality Petroleum Coke ......................................................... 15 K. Bartholomew Impact of Higher Vanadium Levels on Smelter Operations ........................................................................................ 21 C. Coney, L. Crabtree, J Gavin, W Marcrum, A. Weber, and L. Edwards Impact on Smelter Operations of Operating High Purity Reduction Cells .................................................................. 27 S. Hamilton, and R. Cook Management ofImpurities in Cast House with Particular Reference to Ni and V ...................................................... 33 M Rhamdhani, J Grandfield, A. Khaliq, and G. Brooks An Initial Assessment ofthe Effects of Increased Ni and V Content in A356 and AA6063 Alloys ........................... 39 J Grandfield, L. Sweet, C. Davidson, J Mitchell, A. Beer, S. Zhu, X Chen, and M. Easton

Alumina and Bauxite Digestion Implementation of Logic Control by DCS to Measure the Caustic Concentration in Spent Liquor. ........................... 51 A. Oliveira-Santos, A. Carvalho, B. Urakawa, M. Maciel, and A. Santos Study ofTnfluences on the Alumina/Caustic (A/C) Ratio and Discharge Digestion (DBO) Caustic of Through Design of Experiments (DOE) Statistic Tool .............................................................................................................. 55 A. Borges, A. Monteiro, A. Oliveira, B. Urakawa, J Miranda, and D. Silva Particle Size Distribution Model for Kinetics of Digesting Alumina .......................................................................... 59 L. Bao, T Zhang, W Long, A. Nguyen, G. Lv, J Ma, and Y Liu Fractal Kinetic Model for Digesting Alumina ............................................................................................................. 65 L. Bao, T Zhang, A. Nguyen, W Long, J Ma, Z. Dou, and G. Lv

v

MAX HT® Bayer Sodalite Scale Inhibiter: A Green Solution to Energy Consumption .............................................. 71 M. Lewellyn, A. Rothenberg, C. Franz, F. Ballentine, F. Kula, L. Soliz, Q. Dai, and S. Moffatt

Clarification Sodalite Solids Formation at the Surface of Iron Oxide and Its Impact on Flocculation ............................................. 77 A. Senaputra, P. Fawell, F. Jones, and P. Smith Improvement on the Operation Management System of Vertical Pressure Filters ...................................................... 83 T Santos, L. Moraes, A. Sampaio, M Maciel, H. Lima, J Miranda, A. Junior, and J Borges U sing a Multi variate Statistical in the Identification of Alumina Loss in Red Mud .................................................... 87 A. Junior, A. Borges, A. Oliveira, H. Lima, J Ribeiro, J Miranda, and R. Podversek Bevill and the Aluminum Industry .............................................................................................................................. 91 A. Schoedel New Development Model for Bauxite Deposits - Dedicated Compact Refinery ........................................................ 97 P. ter Weer

Red Mud Automatic Control of Drum Filters Operation ........................................................................................................... 105 A. Sampaio, L. Moraes, T Santos, H. Lima, A. Borges, and J Borges A New Technology for Dry Disposal of Alunorte's Bauxite Residue ....................................................................... 109 M. de Castro, R. Trindade, R. Pantoja, E. Alves, and A. Martins Pilot Test of Bauxite Residue Carbonation with Flue Gas ........................................................................................ 113 L. Venancio, J Souza, E. Macedo, F. Botelho, and G. Cesar Management of Industrial Waste: The Case of Effective Utilization of Red Mud and Fly Ash at Vedanta Aluminium Limited - Lanjigarh ................................................................................................................................ 119 M. Kumar, B. Senapati, and C. Kumar Iron Recovery from Red Mud by Reduction Roasting-Magnetic Separation ............................................................ 125 M Rao, J Zhuang, G. Li, J Zeng, and T Jiang Removal of Methylene Blue from Aqueous Solutions Using a Novel Granular Red Mud Mixed with Cement ...... 131 L. Shu aidan, L. Thiquynhxuan, S. Ju, P. Jin-hui, and Z. Li-bo

Precipitation and Calcination Environmentally Safe Operation of Barometric Condensers ..................................................................................... 139 M. Jacobs Hatch - ETI Aluminyum Precipitation Modeling ...................................................................................................... 143 E. Stamatiou, D. Chinloy, B. C;elikel, M. Kayaci, and E. Savkilioglu Improve the Classification System in Hydro Alunorte Lines 4/5 .............................................................................. 147 E. Moraes, H. Haraldsen, C. Junior, J Ribeiro, C. Magro, J Chartouni, E. Santos, and D. Gomes Increase in the Stability of Gravimetric Classification System of Precipitation at Hydro Alunorte .......................... 151 V Cruz, E. Moraes, C. Junior, D. Rodrigues, A. Sousa, A. Furtado, and D. Silva

VI

Experience with Commissioning New Generation Gas Suspension Calciner ........................................................... 155 S. Wind, and B. Raahauge Bayer Process Efficiency Improvement .................................................................................................................... 163 S. Gu HyClass™ Technology for Improvement of Trihydrate Classification in the Bayer Process ................................... 169 J Wang, J Herrera, S. Kostelak, and K. Frederic

Impurities Metallic Impurities from the Mine to Metal Products ............................................................................................... 177 S. Lindsay The Control of Fluoride Concentration in ETI AlUminyum Bayer Refinery Liquor ................................................. 183 E. Savkilioglu, C. Carton, S. Ertugral, M. Baygiil, K. Dine;, and S. Avcu Beneficiation of High Silica Bauxite Ores of India - An Innovative Approach ........................................................ 187 M. Kumar, B. Senapati, and C. Kumar Morphological Investigation of Sodium Oxalate Crystals Grown in Aqueous Sodium Hydroxide Solution ............ 191 W Fu, and J Vaughan Impurities in Raw Gas and Secondary Alumina ........................................................................................................ 195 S. Kalyavina, A. Ratvik, and T Aarhaug

Low Grade Alumina Sources Innovative Technology for Alumina Production from Low-grade Raw Materials .................................................... 203 A. Senyuta, A. Panov, A. Suss, and Y. Layner Improving Characterization of Low Grade Diasporic Bauxite to Be Utilize in Jajarm Alumina Plant ..................... 209 M. Shadloo, M. Zarbayani, E. Jorjani, and M. Aram The Processing of High Quartz Bauxite .................................................................................................................... 217 E. Gasaji, A. Scarsella, V Hartmann, and H. Schmidt Appropriate Reduction and Fe-AI Separation of High Iron Gibbsite ........................................................................ 223 Z. Liu, M Chu, J Tang, Y Han, andX Wu Influence ofMgO and CIA and Cooling System on Alumina Leaching Properties of Calcium Aluminate Slag ...... 229 Z. Tong, Y. Li, and T Chen Calcification-Carbonation Method for Alumina Production by Using Low-Grade Bauxite ..................................... 233 T Zhang, X Zhu, G. Lv, L. Pan, Y Liu, Q. Zhao, Y Li, X Jiang, and J He Basic Research on Calcification Transformation Process of Low Grade Bauxite ..................................................... 239 X Zhu, T Zhang, G. Lv, Y Liu, Q. Zhao, Y Li, and Z. Dou Research on the Phase Transformation and Separation Performance in Calcification-Carbonation Method for Alumina Production ................................................................................................................................................... 245 G. Lv, T Zhang, X Zhu, L. Pan, M. Qin, Y. Liu, Q. Zhao, X Jiang, and Y. Li

VII

Aluminum Alloys: Fabrication, Characterization and Applications Development and Application Mechanical Properties of AI-Zn-Mg-Cu Alloys Processed with High-pressure Torsion .......................................... 255 S. Kuramoto. I. Aoi. and T Furuta

High-Performance Be-AI Casting Alloys .................................................................................................................. 259 G. Schuster, and C. Pokross

Structure Optimization of AI-Si-Type Alloys for Thermal and Mechanical High Loaded Components .................. 265 A. Kleine, M Rose{ort, A. Pithan, C. Matthies, and H. Koch

Development of High Strength Aluminium Alloys at BALCO ................................................................................. 269 M Kar, S. Prasad, A. Paul, and P. Raghavan

Corrosion Resistance Performance Strength and Failure of Ultra fine Grain and Bimodal AI-Mg Alloy at High Temperatures ...................................... 279 A. Magee, and L. Ladani

Process Development of AA3103 Aluminum Alloy for Automotive Thins .............................................................. 283 M. Paes, A. Coelho, R. Netto, and F. Aguiar

Casting and Solidification Atom Probe Analysis of Sr Distribution in AISi Foundry Alloys ............................................................................. 291 1. Barrirero, M. Engstler, and F. Miicklich The Role of Sr on Microstructure Formation and Mechanical Properties of AI-Si-Cu-Mg Cast Alloy .................... 297 M Zamani, S. Sei{eddine, and M Aziziderouei

Modification of the Eutectic Mg 2 Si-Phase of AIMgSi-Cast Alloys .......................................................................... 303 T Pabel, T Petkov, C. Kneissl, and P. Schumacher

The Influence of Casting Speed in the as Cast Strip Mechanical Properties of 8079 and 8006 Alloys ..................... 305 D. Spathis, and 1. Tsiros Effect of Cooling Rate on Iron-Rich Intermetallic Phases in 206 Cast Alloys .......................................................... 311 K. Liu, X Cao, and X Chen

Effect of Iron in AI-Mg-Si-Mn Ductile Diecast Alloy .............................................................................................. 317 S. Ji, W Yang, F. Gao, D. Watson, and Z. Fan Oxidation Behavior of AbCa Added AI-5Mg Alloy in the Liquid State ................................................................... 323 Y. Yoon, S. Ha, G. Yeom, H. Lim, and S. Kim Effect ofthe Thermal Modulus and Mould Type on the Grain Size of AISi7Mg Alloy ........................................... 327 I. Lizarralde, A. Niklas, A. Fernandez-Calvo, and 1. Lacaze Alloy ALSI30 Cast in the Process of Rapid Solidification and Consolidated in the Process of Plastic Forming ..... 333 W Szymanski, M Szymanek, 1. Zelechowski, M Biga}, M Gawlik, and B. Plonka

VIII

Thermal Mechanical Processing Effects of Homogenization Treatment Conditions on the Recrystallization Behavior of AI-I.2Mn Aluminum Alloy Sheets ............................................................................................................................................................... 341 P. Zhao. X Chen. W. Chen. and Y. Zhang Toward a Recrystallized Microstructure in Extruded AA6005A Alloy .................................................................... 347 A. Bahrami, A. Bakker, A. Miroux, and J Sietsma

Grain Subdivision and Its Effect on Texture Evolution in an Aluminum Alloy Under Plane Strain Compression .. 351 Q. Ma, W. Mao, B. Li, P. Wang, and M Horstemeyer

Fatigue Analysis of Ultrafine Grained AI 1050 Alloy Produced by Cyclic Forward Backward Extrusion ............... 357 H. Alihosseini, and M. Zaeem Effect of Zn Content and Process Parameters on Corrosion Behaviour of Twin-Roll Cast Aluminum Brazing Alloys ........................................................................................................................................................... 361 M Dundar, M Gunyuz, C. Isiksac;an, and A. Pastirmaci

Solutioning and Aging Growth Ledges on Silver-Segregated 8' (AbCu) Precipitates ................................................................................... 367 J Rosalie, and L. Bourgeois

On the Aging Behavior of AA2618 DC Cast Alloy .................................................................................................. 373 P. Shen, E. Elgallad, and X Chen

The Effect of Cold Work on the Precipitation and Recrystallization Kinetics in AI-Sc-Zr Alloys ........................... 379 C. McNamara, S. Kampe, P. Sanders, and D. Swenson A Novel Solution Heat Treatment of 7075-Type Alloy ............................................................................................ 383 M. Ibrahim, A. Samuel, S. Alkahtani, and F. Samuel Experimental Study of the AI-rich Corner of the AI-Si-Ti System at 500°C ............................................................ 391 Y Li, Q. Luo, J Zhang, and Q. Li

Emerging Technology Transient Microstructural Thermomechanical Fatigue and Deformation Characteristics under Superimposed Mechanical and Thermal Loading in AISi Based Automotive Diesel Pistons ........................................................... 397 R. Morgenstern, and S. Kenningley Mechanical Behaviour of Cold Formed Metal-Polymer Laminate and the Interaction of Its Layers ....................... .405 F. 6 Dubhlaing, D. Browne, R. Rennicks, and C. Rennicks Mechanical and Tribological Properties of AA2124-Graphene Self Lubricating Nanocomposite ........................... .411 A. Ghazaly, B. Self, and H. Salem Joining Vacuum High Pressure Die Cast A356 under T4 Treatment to Wrought Alloy 6061 ................................. .417 M Wang, Y Zou, H Hu, G. Meng, P. Cheng, and Y Chu

IX

General Poster Session Applications ofthe Horizontal Squeeze Casting Process for Automotive Parts Manufacturing .............................. .425 P. Dulyapraphant, E. Kittikhewtraweeserd, P. Kritboonyarit, and N. Denmud Characterization of the Developed Precipitates in AI-2 at.%Zn-x at.%Mg, (x=1.8, 2, 2.4, 3, 4.2) ........................... 431 N AfifY, A. Gaber, and G. Abbady Design and Development of a Permanent Mould for the Production of Motor-Cycle Piston in Sedi-Enugu ............ 437 C. llochonwu, and E. Nwonye Development and Research of New Aluminium Alloys with Transition and Rare-Earth Metals and Equipment for Production of Wire for Electrotechnical Applications by Methods of Combined Processing ............................ .443 I. Matveeva, N. Dovzhenko, S. Sidelnikov, L. Trifonenkov, V. Baranov, and E. Lopatina Influence of Machining Parameters on AI-4.5Cu-TiC In-Situ Metal Matrix Composites ........................................ .449 P. lha, A. Kumar, and M. Mahapatra Effect ofMg Contents on Fluidity of AI-xMg Alloys ............................................................................................... 453 N Kim, S. Ha, Y Yoon, G. Yeom, H. Lim, and S. Kim Effect of Process Parameters on Centrifugally Cast AI-Si FGM ............................................................................... 457 K. Aithal, V. Desai, N. S, and P. Mukunda Effects of Minor Sc Addtion on the Microstructures and Mechanical Properties of AI-Zn-Mg-Cu Casting Aluminum Alloy ........................................................................................................................................................ 463 G. Yang, S. Liu, and W lie Microhardness, Corrosion Behaviour and Microstructures of Directionally Solidified AI-Cu Alloys ...................... 469 A. Ares, C. Rodriguez, C. Mendez, C. Schvezov, and M Rosenberger Production of Single Cylinder Engine Components through High Pressure Die Casting in Sedi Enugu .................. 475 E. Nwonye, C. llochonwu, and C. Nwajagu The Effect of Thermomechanical Aging of Aluminium-Copper Alloy (MATLAB Approach) ................................ 481 A. Adegbola, A. Ghaza/i, 0. Fashina, A. Omotoyinbo, and 0. Olaniran

Aluminum Processing Aluminum Processing I Surface Crack Characterization of Twin Roll Caster Shells and Its Influence on As-Cast Strip Surface Quality .... .491 M. Diindar, B. Beyhan, 0. Birbasar, H. Altuner, and C. /siksar;an

Aluminum Processing II The Effect of Magnesium Content on Microstructure Evolution during Hot Deformation of Aluminum Alloys .... .499 T Watt, S. Yasuda, K. /chitani, K. Takata, A. Carpenter, 1. lodlowski, and E. TalefJ High Strength Nanostructured AI-Zn-Mg-Cu-Zr Alloy Manufactured by High-Pressure Torsion ........................... 505 C. An, H. Lu, and S. Yuan Corrosion Behavior of 2024 Aluminum Alloy Anodized in Sulfuric Acid Containing Inorganic Inhibitor ............. 509 M Mohammadi, A. Yazdani, F. Mohammadi, andA. Al{antazi x

Laboratory Simulation of Wear during Hot Extrusion of Aluminium ....................................................................... 515 G. Kugler, and M. Tercelj The Production of Wrought AISi30Cul.5Mgl.2Nil.5FeO.8 Alloy with Ultrafine Structure .................................... 521 M Szymanek, B. Augustyn, W Szymanski, and D. Kapinos The Structure and Properties of Wrought Aluminium Alloys Series 6xxx with Vanadium for Automotive Industry .................................................................................................................................................. 527 M Lech-Grega, W Szymanski, B. Plonka, S. Boczkal, M Gawlik, M Biga}, and P. Korczak

Aluminum Reduction Technology Cell Design and Performance In Depth Analysis of Energy-Saving and Current Efficiency Improvement of Aluminum Reduction Cells ............ 537 F. Yan, M Dupuis, J Zhou, and S. Ruan Rio Tinto Alcan AP4X Low Energy Cell Development.. .......................................................................................... 543 P. Thibeault, S. Becasse, A. Blais, P. Cote, L. Fiot, and F. Laflamme Energy Reduction Technology for Aluminum Electrolysis: Choice ofthe Cell Voltage .......................................... 549 F. Naixiang, P. Jianping, W Yaowu, D. Yuezhong, and L. Xian 'an Advancements of Dubal High Amperage Reduction Cell Technologies ................................................................... 553 M Reverdy, A. Zarouni, J Faudou, Q. Galadari, A. Al Zarouni, S. Akhmetov, K. Al Aswad, M AI-Jallaf, W Al Sayed, and V Potocnik Development of Low-Voltage Energy-Saving Aluminum Reduction Technology ................................................... 557 J Li, X Lv, H. Zhang, and Y. Liu DI8+: Potline Modernisation at DUBAL .................................................................................................................. 561 S. Akhmetov, D. Whitfield, M. AI-Jalla/, A. Al Zarouni, A. Arkhipov, A. AI-Redhwan, and W Abou Sidou Industry Test of Perforation Anode in Aluminium Electrolysis Technology ............................................................ 567 Y Tian, H. Li, L. Wei, X Cao, and J Yin The First Results of the Industrial Application of the EcoSoderberg Technology at the Krasnoyarsk Aluminium Smelter ....................................................................................................................................................................... 573 V Buzunov, V Mann, E. Chichuk, V Frizorger, A. Pinaev, and E. Nikitin

Fundamentals: Modelling Unsteady MHD Modelling Applied to Cell Stability ................................................................................................ 579 S. Renaudier, B. Bardet, G. Steiner, A. Pedcenko, J Rappaz, S. Molokov, and A. Masserey Impact of Magnetohydrodynamic and Bubbles Driving Forces on the Alumina Concentration in the Bath of an Hall-Heroult Cell ....................................................................................................................................................... 585 R. von Kaenel, J Antille, M. Romerio, and 0. Besson Investigation of Electrolytic Bubble Behaviour in Aluminium Smelting Cell .......................................................... 591 MAlam, Y Morsi, W Yang, K. Mohanarangam, G. Brooks, and J Chen Mathematical Model Validation of Aluminium Electrolysis Cells at DUBAL ......................................................... 597 A. Zarouni, L. Mishra, M Bastaki, A. Al Jasmi, A. Arkhipov, and V Potocnik XI

Production Application Study on Magneto-Hydro-Dynamic Stability of a Large Pre baked Anode Aluminum Reduction Cell ........................................................................................................................................................... 603 S. Ruan. F. ran, M. Dupuis, V. Bojarevics, and 1. Zhou MHD of Aluminium Cells with the Effect of Channels and Cathode Perturbation Elements ................................... 609 V. Bojarevics Magnetohydrodynamic Model Coupling Multiphase Flow in Aluminum Reduction Cell with Innovative Cathode Protrusion .................................................................................................................................................... 615 Q. Wang, B. Li, F. Wang, and N. Feng Optimization ofthe Cathode Collector Bar with a Copper Insert Using Finite Element Method ............................. 621 M. Gagnon, P. Goulet, R. Beeler, D. Ziegler, and M. Fafard Energy Savings in Aluminum Electrolysis Cells: Effect ofthe Cathode Design ...................................................... 627 M Blais, M Desilets, and M Lacroix

Potline Operation I: Smelter Operations Low Power Operation at Aluminium Dunkerque Smelter.. ....................................................................................... 635 1. Peyneau, L. Fiot, S. Mermet-Guyenet, and 0. Rebouillat Maximizing Creeping Value through Rigorous Methodology .................................................................................. 641 B. Champel, and N Monnet The Quick Shut Down and Restarting of291 kA Pre-Baked Potline at JSC "RUSAL Sayanogorsk" from May to August 2011 .......................................................................................................................................................... 647 V. Buzunov, A. Soldatov, V. Mann, A. Pavin, V. Borisov, S. Zatepyakin, E. Scherbakov, and A. Gouzenkov Production Growth and Future Challenges in Aluminium Bahrain (Alba) ............................................................... 653 1. AI-Ansari, A. Habib, A. Mittal, and N AI-Jallabi High Frequency Power Modulation - TRIMET Smelters Provide Primary Control Power for Stabilizing the Frequency in the Electricity Grid .............................................................................................................................. 659 A. Lutzerath Autonomous Vehicle and Smelter Technologies ....................................................................................................... 663 A. Tews, and P. Borges Preventive Maintenance of Transport Vehicles: Is It Improving Production Stability of a Smelter? ........................ 669 M Meijer

Fundamentals: Chemistry Composition and Thermal Analysis of Crust Formed from Industrial Anode Cover ................................................ 675 Q. Zhang, M Taylor,1. Chen, D. Cotton, T Groutzo, andX Yang Liquidus Temperatures ofNa3AIF6 -AIFrCaFrKF-LiF-Ab03 Melts ...................................................................... 681 D. Yuezhong, P. Jianping, B. Yunbin, and F. Naixiang The Effect of Calcium Fluoride on Alumina Solubility in Low Temperature Cryolite Melts ................................... 685 P. Tingaev, Y. Zaikov, A. Apisarov, A. Dedyukhin, and A. Redkin

XII

Conductivity ofKF-NaF-AIF 3 System Low-temperature Electrolyte ....................................................................... 689 1. Yang, W Li, H. Yan, and D. Liu Numerical Analysis oflonic Mass Transfer in the Electrolytic Bath of an Aluminium Reduction Cell ................... 695 MAriana, M Desilets, and P. Proulx Liquidus Temperature of Electrolytes for Aluminum Reduction Cells ..................................................................... 701 D. Shi, B. Gao, Z. Wang, Z. Shi, andX Hu Effect of LiAI02 and KF on Physicochemical Properties for Industrial Aluminum Electrolyte .............................. 705 X Lv, S. Chen, Y. Lai, Z. Tian,1. Li, and H. Zhang

Cell Operations and Process Control Improvement of Alumina Dissolution Rate through Alumina Feeder Pipe Modification ......................................... 713 1. Tessier, G. Tarcy, E. Batista, X Wang, and P. Doiron Reduction Cell Restart Method Influence on Cell Life Evolution ............................................................................. 719 M. Lukin, and R. leltsch Start of an Aluminum Reduction Cell without Liquid Bath ...................................................................................... 725 K. Lalonde, B. Audie, W Kristensen, and T Snyder A MIMO Modeling Strategy for Bath Chemistry ...................................................................................................... 729 F. Soares, and R. Oliveira Cumulative Distributions of Metallic Impurities ....................................................................................................... 735 S. Lindsay Sodium Content in Aluminum and Current Efficiency - Correlation through Multivariate Analysis ....................... 741 L. Dion, L. Kiss, P. Chartrand, G. Dufour, and F. Laflamme Gas-Solid Flow Applications for Powder Handling in Aluminum Smelters Processes ............................................. 747 P. Vasconcelos, and A. Mesquita Operational Experience of Advanced Alumina Handling Technology in a Russian Smelter .................................... 753 1. Paepcke, A. Hilck, and S. Marshalko

Environment I Reduction in HF Emission through Improvement in Operational Practices .............................................................. 763 H. Devadiga, A. Banjab, M. AUallaf A. Al Zarouni, K. Al Aswad, A. Kumar, G. Meintjes, S. Gowda, and M. Khan Trace Element Concentration in Particulates from Pot Exhaust and Depositions in Fume Treatment Facilities ...... 769 H. Gaertner, A. Ratvik, and T Aarhaug The Study and Applications of Modern Potline Fume Treatment Plant (FTP) ......................................................... 775 D. Xiang, L. Weining, L. Xun, D. Qiyi, and Y Xiaobing F>C: Combined Treatment of Pot Gases and Anode Baking Furnace Fumes ..................................................... 781 B. Hureiki, C. Lim, A. Periers, E. Bouhabila, G. Girault, M. Leduc, and S. Delenclos Compact Filter Design for Gas Treatment Centers .................................................................................................... 787 P. Verbraak, P. Klut, T Turco, E. Dupon, and E. Engel

XIII

An Innovative Compact Heat Exchanger Solution for Aluminium Off-Gas Cooling and Heat Recovery ................ 793 E. Bouhabila, E. Naess, V. Einenjord, and K. Kristjansson Latest Filter Developments Increasing Existing Aluminium Smelter Gas Treatment Centre Capacity and Reducing Emissions .................................................................................................................................................. 799 M Neate, and B. Currell Reduced Ventilation of Upper Part of Aluminum Smelting Pot: Potential Benefits, Drawbacks, and Design Modifications ............................................................................................................................................................. 805 R. Zhao, L. Gosselin, M. Fafard, and D. Ziegler Latest Developments in Potroom Building Ventilation CFD Modelling ................................................................... 811 N. Menet, G. Girault, N. Monnet, C. Turpin, and L. Soulhac

Potline Operation II: Equipment Solutions to Address Arc Welding Problems in an Operating Potline ...................................................................... 819 B. Paul, Y. EIGhaoui, P. Jadaud, 1. Anderson, and S. Foulds Replacement of Damaged Electrical Insulators on Live Cross-Over Busbars inside a Tunnel: A Methodology Based on Risk Assessment and Numerical Simulation ............................................................................................. 823 D. Richard, A. Yelle, 0. Charette, A. Schneider, 1. Nadeau, M Gliere, Y Drouet, and P. Breme A Thermal-Mechanical Approach for the Design of Busbars Details ....................................................................... 829 A. Schneider, 0. Charette, D. Richard, and C. Turcotte Study of Technology and Equipment on Magnetic Induction Intensity Weaken for Aluminum Reduction Cells Welding in the Condition of Pot Line Current. .......................................................................................................... 837 Z. Wang, B. Cao, T Yang, 1. Huang, and M. Li Potline Shutdown and Restart Secured Solutions ...................................................................................................... 843 A. Hequet Effect of Watering and N on-Watering Cooling Rates on the Mechanical Properties of an Aluminum Smelter's Potshell ...................................................................................................................................................................... 845 A. Brimmo, M Hassan, M Ibrahiem, and Y Shatilla Mathematical Model of Cooling ofa Stopped Pot and Its Validation ....................................................................... 851 M. Hassan, A. Brimmo, M. Ibrahiem, and Y. Shatilla

Environment II: PFCs A Study of Low Voltage PFC Emissions at Dubal .................................................................................................... 859 A. Zarouni, M. Reverdy, A. Al Zarouni, and K. Venkatasubramaniam Continuous PFC Emissions Measured on Individual400kA Cells ........................................................................... 865 D. Wong, and 1. Marks PFC and Carbon Dioxide Emissions from an Australian Aluminium Smelter Using Time-Integrated Stack Sampling and GC-MS, GC-FID Analysis ................................................................................................................. 871 P. Fraser, P. Steele, and M Cooksey Investigation on Formation Mechanism of Non-Anode Effect Related PFC Emissions from Aluminum Reduction Cells ......................................................................................................................................................... 877 X Chen, W. Li, Y. Zhang, S. Qiu, and C. Bayliss

XIV

On the Mechanism Behind Low Voltage PFC Emissions ......................................................................................... 883 1. Thonstad, S. Rolseth, and R. Keller Frequency Response Analysis of Anode Current Signals as a Diagnostic Aid for Detecting Approaching Anode Effects in Aluminum Smelting Cells ......................................................................................................................... 887 C. Cheung, C. Menictas, 1. Bao, M Skyllas-Kazacos, and B. Welch Reduction Strategies for PFC Emissions from Chinese Smelters .............................................................................. 893 W Li, X Chen, S. Qiu, B. Zhang, and C. Bayliss Off-gas Analysis of Laboratory-Scale Electrolysis Experiments with Anodes of Various Compositions ................ 899 0. Kjos, T Aarhaug, E. Skybakmoen, A. Solheim, and H. Gudbrandsen Hydrolysis of Carbonyl Sulfide (COS) on Smelting Grade Alumina ........................................................................ 905 A. Mikhonin, N. Dando, and M. Gershenzon

Cell Fundamentals, Phenomena and Alternatives I (2012) A Thermodynamic Approach to the Corrosion ofthe Cathode Refractory Lining in Aluminium Electrolysis Cell: Modelling of the AbOrNa20-SiOrAIFrNaF-SiF4 System ............................................................................. 911 G. Lambotte, and P. Chartrand Effect of Current Density and Phosphorus Impurities on the Current Efficiency for Aluminum Deposition in Cryolite-Alumina Melts in a Laboratory Cell ............................................................................................................ 917 R. Meirbekova, G. Saevarsdottir, G. Haarberg, and 1. Armoo

Cast Shop for Aluminum Production Aluminum Cast Shop I Very High Purity Ingot - An Endangered Species? ................................................................................................... 925 S. Lindsay The Challenge of Effectively Utilizing Trace Elements/Impurities in a Varying Raw Materials Market.. ............... 929 G. Jha, S. Ningileri, X Li, and R. Bowers Energy Control in Primary Aluminium Casthouse Furnaces ..................................................................................... 935 I. Johansen, and S. Strf3mhaug Metal Contamination Associated with Dross Processing .......................................................................................... 941 R. Peterson

Aluminum Cast Shop II Ultrasonic Degassing and Processing of Aluminum ................................................................................................. 949 V Rundquist, and K. Manchiraju Kinetics of Ultrasonic Degassing of Aluminum Alloys ............................................................................................ 957 N. Alba-Baena, and D. Eskin Removal of Inclusions in Molten Aluminum by Flux Injection under Counter-Gravity ........................................... 963 1. Zeng, and H. Gu

xv

Advanced Compact Filtration (ACF): An Efficient and Flexible Filtration Process ................................................. 967 F. Breton, P. Waite, and P. Robichaud Electromagnetic Priming of Ceramic Foam Filters (CFF) for Liquid Aluminium Filtration .................................... 973 R. Fritzsch, M Kennedy, J Bakken, and R. Aune Plant Scale Investigation of Liquid Aluminum Filtration by Ab03 and SiC Ceramic Foam Filters ......................... 981 S. Bao, M Syvertsen, A. Nordmark, A. Kvithyld, T Engh, and M Tangstad Casting Practices Influencing Inclusion Distributions in Billets ............................................................................... 987 G. Razaz, and T. Carlberg Oxidation of Commercial Purity Aluminium Melts: An Experimental Study ........................................................... 993 S. Bonner, J Taylor, J Yao, and M. Rhamdhani

Aluminum Cast Shop III Optimisation of Grain Refinement .......................................................................................................................... 1001 R. Vainik, J Courtenay, and B. Saglam Grain Refiner for AI-Si Alloys ................................................................................................................................ 1009 H. Nadendla, M Nowak, and L. Bolzoni Production of AI-Ti-B Grain Refining Master Alloys from B 20 3 and K2 TiF 6 by Microwave Irradiation .............. 10 13 S. Zhou Effects ofYb Additions on Refinement of Eutectic Si in AI-5Si Alloys ................................................................. 1017 J Li, and P. Schumacher Influence of Vanadium on the Microstructure of A356 Foundry Alloy .................................................................. 1023 T Ludwig, P. Schaffer, and L. Arnberg

Aluminum Cast Shop IV Influence of Die and Casting Temperatures and Titanium and Strontium Contents on the Technological Properties of Die-Cast A356 in the As-Cast and T6 Condition ............................................................................... 1031 S. Fischer, V Groten, J Brachmann, C. Fix, T Vossel, and A. Buhrig-Polaczek Horizontal Single Belt Strip Casting (HSBC) of AI-Mg-Sc-Zr Alloys ................................................................... 1037 M Celikin, D. Li, L. Calzado, M /sac, and R. Guthrie

Electrode Technology for Aluminium Production Anode Raw Materials Review of Different Techniques to Study the Interaction Between Coke and Pitch in Anode Manufacturing ....... 1045 D. Kocaefe, A. Sarkar, S. Das, S. Amrani, D. Bhattacharyay, D. Sarkar, Y Kocaefe, B. Morais, and M. Gagnon Observations on the Coke Air Reactivity Test ........................................................................................................ 1051 K. Neyrey, L. Edwards, and J Marino Impurity Removal from Petroleum Coke ................................................................................................................ 1057 A. Gagnon, N Backhouse, H. Darmstadt, E. Ryan, L. Dyer, and D. Dixon XVI

Calcined Coke Round Robin 19 and the Precision of Bulk Density Tests .............................................................. 1063 M. Lubin, L. Edwards, and L. Lossius A Method for the Rapid Characterisation of Petroleum Coke Microstructure Using Polarised Light Microscopy .............................................................................................................................................................. 1069 A. Innus, A. Jomphe, and H Darmstadt Improvements of Vibrated Bulk Density Analysis at VM-CBA and Petrocoque S.A ............................................. 1075 J Pardo, E. daSilva, P. da Silva, and A. Nantes Influence of GPC Properties on the CPC Quality .................................................................................................... 1079 J Zhao, Q. Zhao, Q. Zhao, L. Yu, and P. Yu Quality of Calcined Petroleum Coke and Its Influence on Aluminium Smelting .................................................... 1085 J Subero

Paste Plant Operations A Green Anode Plant Performance Analysis Tool Fully Embedded in the Plant Control System .......................... 1091 X Genin, and P. Calo Measures to Prevent Baked Anode Density Drop When Using High Porosity Cokes ............................................ 1097 V Pif{er, C. Woo, F. Niceas, R. Bacelar, J Araujo, and L. Paulino New Green Anode Plant at EMAL - Start-Up and Operation in the First 2 Years .................................................. 1101 R. Akhtar, R. Gemein, and M. Bei/stein Improving Baked Anode Density and Air Permeability through Process Optimization and Coke Blending .......... 1105 B. Ndjom, M. Malik, A. AIMarzouqi, T Sahu, and S. Rabba Development of an Analytical Dynamic Model of a Vibro-Compactor Used in Carbon Anode Production .......... 1111 F. Rebarne, M Bouazara, D. Marceau, D. Kocae{e, and B. Morais Driving Cost Reduction and Carbon Plant Productivity Improvement through Theory of Constraints and Planned Maintenance Capability ............................................................................................................................. 1117 K. Sinclair, and B. Sadler Optimum Vibration Time for Green Anode Production .......................................................................................... 1123 S. Gao, C. Bao, S. Zhang, H. Wang, J Woo, and E. Cutshall Comparison of Mixing Process Methods in Pre baked Anode Production ............................................................... 1127 S. Yi, G. Huai, Z. Shanhong, L. Chaodong, and X Hai{ei

Bake Furnace Design and Operation Hydro Aluminium's Historical Evolution of Closed Type Anode Baking Furnace Technology ............................ 1133 M Tkac, A. Ruud, 1. Holden, and H Linga Use of Mathematical Modelling to Study the Behavior of a Horizontal Anode Baking Furnace ............................ 1139 Y Kocae{e, N Oumarou, M Baiteche, D. Kocaefe, B. Morais, and M Gagnon Study on Anode Baking Parameters in Open-Top and Closed-Type Ring Furnaces .............................................. 1145 B. Baharvand, M. Siahouei, M. Batoei, and S. Sadeghi Energy Efficiency Improvement in Anode Baking Furnaces .................................................................................. 1151 C. Linhares, F. Niceas, R. Bacelar, H. Campelo, M. Silva, J Araujo, and R. Reis XVII

Anode Baking Process Improvement at ALRO ....................................................................................................... 1155 P. Mahieu, N. Fiot, A. Trillat, 0. Balu, and C. Stanescu Operational and Environmental Benefits on the New Baking Furnace at Boyne Smelter by Use of an Advanced Firing Technology ................................................................................................................................................... 1163 A. Himmelreich, D. Maiwald, D. DiLisa, G. Cordon, and S. Moodley Laser Mapping of Carbon Bake Furnaces ............................................................................................................... 1169 A. Tews, M Bosse, R. Zlot, P. Flick, and M Noonan

Anode Quality and Performance Pilot Scale Anodes for Raw Material Evaluation and Process Improvement .......................................................... 1177 L. Lossius, J. Chmelar, I. Holden, H. Linga, and M. Tkac Relationships between Coke Properties and Anode Properties - Round Robin 19 .................................................. 1183 L. Lossius, M. Lubin, L. Edwards, and J. Wyss Application ofthe Artificial Neural Network (ANN) in Predicting Anode Properties ............................................ 1189 D. Bhattacharyay, D. Kocae{e, Y Kocaefe, B. Morais, and M Gagnon A Model for Predicting the Electrical Resistivity of Baked Anodes ....................................................................... 1195 D. Bhattacharyay, D. Kocae{e, Y Kocaefe, B. Morais, and M Gagnon The Role of Electrode Quality in Metal Purity ........................................................................................................ 1201 S. Lindsay Electrochemical Characterization of Carbon Anode Performance .......................................................................... 1207 R. Thorne, C. Sommerseth, E. Sandnes, 0. Kjos, T Aarhaug, L. Lossius, H. Linga, and A. Ratvik High Capacity Thermobalance Anode Reactivity Testing ...................................................................................... 1213 N Janssen, J. Baker, F. Cannova, and B. Sadler Diagnosing Changes in Baked Anode Properties using a Multivariate Data-driven Approach ............................... 1219 J. Lauzon-Gauthier, C. Duchesne, and J. Tessier

Cathode Materials and Wear Evolution of the Thermo-Mechanical Properties of Ramming Paste from Ambient to Operating Temperature in Hall-Heroult Cell ..................................................................................................................................................... 1227 S. Tremblay, L. St-Georges, L. Kiss, L. Hacini, B. Allard, and D. Marceau New Compaction Method for the Production of Large Ramming Paste Samples for 3D Mechanical Characterization ................................................................................................................................... 1233 P. St-Arnaud, D. Picard, M Noel, H. Alamdari, D. Ziegler, and M Fa{ard Technology for Manufacturing Cathodes Used in Aluminum Reduction in China ................................................. 1239 H. Yang, F. Liu, S. Cai, andX Yang The Effect Of Cryolite on the Formation of Aluminum Carbide at the Carbon Aluminum Interface ..................... 1245 B. Novak, K. TschOpe, A. Ratvik, and T Grande Critical Reflections on Laboratory Wear Tests for Ranking Commercial Cathode Materials in Aluminium Cells ...................................................................................................................................................... 1251 K. Tschope, A. Stare, E. Skybakmoen, A. Solheim, T Grande, and A. Ratvik

XV1l1

Model for Excessive Cathode Wear by a "Carbon Pump" at the Cell Bottom ........................................................ 1257 A. Solheim, and K. Tschope

Characterization of Porous Structure and Its Correlation to Sodium Expansion of Graphite Cathode Materials Using Image Analysis .............................................................................................................................................. 1263 XLi, J Xue, and T Chen

Studies on the Resistance to Alkali Metal Penetration of Binders for TiBrC Composite Cathode Materials ........ 1269 Z. Fang, K. Zhang, Y Lai, L. Li, and J Zhu

Inert Anodes, Cell Materials and Alternative Processes Mechanically Alloyed Cu-Ni-Fe-Y Material as Inert Anode for Al Production ..................................................... 1277 V. Guvarov-Bancalero, D. Guay, and L. Roue

Cold Spray Deposition of Mechanically Alloyed Cu-Ni-Fe Material for Application as Inert Anodes for Aluminum Production ............................................................................................................................................. 1283 G. Goupil, S. Helle, E. Irissou, D. Poirier, J Legoux, D. Guay, and L. Roue Initial 1000A Aluminum Electrolysis Testing in Potassium Cryolite-Based Electrolyte ........................................ 1289 J Hryn, 0. Tkacheva, and J Spangenberger

Electrochemical Behavior of Cermet Anodes in Na3AIF6-K3AIF6-Based Low-Melting Electrolytes for Aluminum Electrolysis .............................................................................................................................................................. 1295 G. Wang, and X Sun

Production of Aluminum Sulfide through Carbosulfidation Utilising H2 S ............................................................. 1299 N. Huda, M. Rhamdhani, G. Brooks, B. Monaghan, and L. Prentice Microstructural Evolution of Cast Iron Used for Cathode Rodding in Aluminum Electrolysis Cell ...................... 1305 A. Hekmat-Ardakan, G. Soucy, and L. Rivoaland

Preparing AI-Sc-Zr Alloys in Aluminum Electrolysis Process ................................................................................ 1311 Y Qian, J Xue, Q. Liu, and J Zhu

CBF Environmental & Anode Electrical Connections Fume Treatment Systems Based on RTO Technology for Carbon Baking Furnaces .............................................. 1317 M. Hagen, and B. Schricker

AHEX-A New, Combined Waste Heat Recovery and Emission Control System for Anode Bake Furnaces ......... 1323 A. Sorhuus, S. Gse, and G. Wedde Successful Start-Up of the Fume Treatment Centre at Boyne Smelter Carbon Bake Furnace #4 ........................... 1329 J Higley, G. Cordon, P. Klut, R. Ghana, E. Dupon, T Turco, and E. Engel

Thermo-Electro-Mechanical Characterization of Anode Interfaces at Operating Conditions ................................. 1335 H. Fortin, M. Martin, N. Kandev, G. Gauvin, D. Ziegler, and M. Fafard

A Fully Coupled Thermal-Electrical-Mechanical Transient FEA Model for a 3D Anode Assembly ..................... 1341 D. Gunasegaram, and D. Molenaar

Experimental and Numerical Investigation of Voltage Drop in Anode Assemblies ............................................... 1347 E. Jeddi, D. Marceau, L. Kiss, L. St-Georges, D. Laroche, and L. Hacini

XIX

Optimization ofthe Anode-Stub Contact: Effect of Casting Temperature, Contact Stress, Temperature and Surface Roughness .................................................................................................................................................. 1353 B. 0e, A. Store, E. Haugland, and 1. Hop Experimental Investigation of Factors Affecting the Electrical Performance of the Stub to Carbon Connection ... 1359 D. Molenaar, T Kilpatrick, and A. Montalto Author Index ............................................................................................................................................................ 1365 Subject Index ........................................................................................................................................................... 1371

xx

PREFACE It is my honour to welcome you to the 142nd TMS Annual Meeting and Exhibition at San Antonio, Texas and to present the Light Metals 2013 proceedings. As always, Light MetaL'i' 2013 is the collective output of the huge intellectual efforts by Authors, Session Chairs, Subject and Symposium Organisers, and TMS Support Staff. We owe these colleagues our sincere thanks.

We are meeting with the aluminium smelting industry continuing to face difficult times, with ongoing oversupply and metal prices generally declining through 2012. Enduring post GFC instabilities in a number of economies around the world mean that right now the light at the end of the tunnel is not easy to see. During tough times, the typical response of companies is to "tighten the belt" to conserve cash, which unfortunately leads to lower attendance at meetings such as this. I firmly believe that participation at the TMS Annual Meeting, with its inherent technical interchange opportunities, should not be seen as a cost to be cut, but rather as an investment to be made. As a Graduate at the Comalco Research Centre, a request to attend my first TMS meeting in 1987 was approved on the basis that I find plant improvements that deliver net savings more than ten times the attendance cost. The considerable cost of travel to the United States meant this was not a trivial task. I can say, however that this target was easily exceeded for the first, and all subsequent TMS meetings I attended as a Comalco employee. The target was also exceeded for all the TMS meetings attended by others that I approved travel for, as they had the same task. The ideas that will give you this return on investment are to be found in the papers in this volume, Author presentations, informal discussions over coffee, visiting the Exhibition, and very importantly, from the contacts made that can last a career. In addition to economic issues, and the sustainability and energy concerns that are inherent parts of our business, we now face a further challenge. The changing quality of raw materials, most notably petroleum coke for anodes, is increasing Casthouse metal impurities. A special Keynote Presentation session is devoted to this topic during the meeting. Les Edwards has done an excellent job of organising this session to cover the issues from raw materials and process impacts through to metal quality considerations. I believe impurities are a growing concern that will require serious attention right across our industry and urge you to attend the Keynote session and engage with the issues. The contribution of Steve Lindsay to this topic deserves special recognition as he is making a Keynote Presentation as well as presenting papers related to impurities in four of the Aluminum Committee sponsored symposia. On behalf of the organisers of the Light Metals 2013 conference and proceedings, I would like to thank the TMS staff for their support and ability to deliver on our requests. I would also like to thank Steve Lindsay and Carlos Suarez (Past and current Chair of the Aluminum Committee) for their support. My deepest gratitude go to the Authors for their efforts, and the Subject Organizers who have done most of the work: Pat Clement, Les Edwards, Mark Cooksey, Gyan Jha, Kai Karhausen, Zhengdong Long, Subodh Das, William Golumbfskie, and Tongguang Zhai. To end on a positive note, I am sure that you will find that Light 2013 maintains the position of this series as the preeminent source of developments in aluminum process knowledge and as such can help us through the difficult current times.

BaiTY A. Sadler

XXI

EDITOR1S BIOGRAPHY

BARRY A. SADLER LIGHT METALS 2013 EDITOR Barry Sadler has been involved in the aluminium industry for more than 30 years in a range of positions, but always with a focus on anode carbon technology. He has a Ph.D. in Metallurgy, and commenced his career in 1982 at the Comalco Research Centre (CRC) in Melbourne, Australia. Barry established the Carbon Technology Group at CRC before moving to Comalco's New Zealand Aluminium Smelter (NZAS) as Carbon Plant Manager in 1989. After more than seven years at NZAS, and following a stint as the General Manager of Organisational Effectiveness for Hamersley Iron, Barry was appointed as a Corporate Technical General Manager at Comalco Aluminium's headquarters in Brisbane, Australia. Barry resigned from Rio TintolComalco in 2002 to successfully set up "Net Carbon Consulting Pty Ltd". As a consultant, he provides advice, training, and support to clients on improving carbon plant performance and process technology, always maintaining a strong focus on the practical application of statistical thinking and methods to process management. Barry has authored or co-authored over 25 technical papers, is a regular lecturer on anode technology at post-graduate courses run by the University of Auckland, and has been an invited speaker at a range of conferences and meetings. He has been an active member of TMS for 25 years as a presenter, session chair, subject organism', and member of the Aluminum Committee.

XX 111

PROGRAM ORGANIZERS ALUMINUM KEYNOTE: IMPURITIES IN THE ALUMINUM SUPPLY CHAIN AND ELECTRODE TECHNOLOGY FOR ALUMINUM PRODUCTION Les Edwards is Vice President of Technical Services at Rain CII Carbon based in Houston, Texas. He has been with Rain ClI since 1998 and is responsible for customer technical support, R&D activities, laboratory services, and quality management. Les has been a regular contributor to TMS meetings over the last 20 years as an author, presenter, and session chair. Les was Program Organizer of the Carbon Teclmology sessions at the 2001 TMS meeting. Prior to joining Rain cn Carbon, he spent 11 years working in the aluminum industry in Australia, based at the Comalco Research Center in Melbourne. Les holds a B.S. from the University of Western Australia and an M.B.A. from Tulane University in New Orleans.

ALUMINA AND BAUXITE Pat Clement graduated from the Colorado School of Mines with B.S. and M.S. in Metallurgical Engineering. He originally joined ASARCO Inc. as a Research Engineer working in various non-ferrous metals. After leaving ASARCO he worked for Brush Wellman in the powder fabrication of Beryllium components. Pat began working in the alumina industry in 1994 when he joined Ormet at their alumina facility in Burnside, Louisiana, eventually becoming Technical Manager for the facility. He is currently a Technical Specialist for Alcoa at their Point Comfort Operations Alumina Refinery.

ALUMINUM ALLOYS: FABRICATION, CHARACTERIZATION AND APPLICATIONS Zhengdong Long is an Alloy Development Engineer at Kaiser Aluminum in Spokane, Washington, United States. He has been active for over a decade in the area of physical and mechanical metallurgy of aluminum alloy and superalloys. Dr. Long's diverse experience includes casting, thermo-mechanical processing, mechanical properties, corrosion behaviors as well as casting and rolling process modeling. lIe specializes in microstructure characterization and mechanical behavior of metallic materials. His principle research interests arc in the development of industrial processes for manufacturing. He has two patents to his credit and more than 30 published papers. He received his doctorate in Materials Science and Engineering from Central Iron and Steel Research and Institute in Beijing, China, in 2000.

xxv

ALUMINIUM PROCESSING Kai F. Karhausen is depmiment manager for process teclmology at the central Rolled Products R&D of Hydro Aluminium in Bonn, Germany. Dr. Karhausen earned his doctorate at the RWTH Aachen and worked in the industrial aluminum research for 15 years both in Norway and Germany. His principal work is focused on the modeling and optimization of materials behavior in industrial production processes. Dr. Karhausen has issued 75 scientific presentations and publications. In 2003 he was awarded the Georg-Sachs-Preis of the German Materials Society (DGM) for important achievements in the field of integrated modeling of metal forming and materials behavior.

ALUMINUM REDUCTION TECHNOLOGY Mark Cooksey is Program Leader - Process Integration at CSIRO Process Science and Engineering, where he leads a group of approximately 90 engineers, physicists, chemists, and technical statIto develop step-change technologies for the resource processing industries. He is also responsible for the aluminium smelting project portfolio at CSIRO. Mark's career began in 1997 as a Research Engineer at Comalco Research Centre, initially working on the development of improved aluminium alloys and aluminium casting processes. After becoming a Senior Research Engineer in 2000, he gradually moved toward aluminium smelting research. He also worked for GE Plastics as a Production Technologist in 2003. Mark joined CSIRO in 2004 as a project leader for multiple research projects in aluminium and magnesium production. In 2008 he was appointed Group Leader - Process Engineering, managing approximately 30 engineers and technical staff. While at CSIRO, Mark has completed a Ph.D. in Chemical and Materials Engineering, developing a technique to directly measure ohmic resistance in aluminium reduction cells.

CAST SHOP FOR ALUMINUM PRODUCTION Gyan Jha has spent 31 years in the aluminum sheet and packaging business. Currently CTO for Tri-Arrows Aluminum, Gyan has extensive experience in all aspects rigid container sheet processing and the manufacturing of aluminum for packaging. Gyan's work experience includes 25 years at Tri-Arrows Aluminum (formerly ARCO Aluminum).

XXVI

ALUMINUM COMMITTEE 2012-2013 Chairperson Carlos Enrique Suarez, Sr. Alcoa Inc. Al Khobar, Saudi Arabia

Secretary Alan David Tomsett Pacific Aluminium Queensland, Australia

Vice Chairperson Barry A. Sadler Net Carbon Consulting Pty. Ltd. Victoria, Australia

Member-at-Large Robeti F. Baxter Bechtel Corporation London, Great Britain

Past Chairperson Stephen J. Lindsay Alcoa Inc. Temlessee, USA

Member-at-Large Naixiang Feng Northeastern University Shen Yang, China

Light Metals Division Chairperson John N. Hryn Argonne National Laboratory Illinois, USA

Member-at-Large Ender Suvaci Anadolu University Eskisehir, Turkey

JOMAdvisor Alton T. Tabereaux Alabama, USA

MEMBERS THROUGH 2013 Pierre LeBrun Constellium CRV Voreppe, France

Gilles Dufour Alcoa Canada Quebec, Canada

Everett C. Phillips N alco Company Illinois, USA

XXVII

MEMBERS THROUGH 2014 KetHA. Rye Alcoa Mosj0en Mosj0en, Norway

Charles Mark Read Bechtel Corporation Quebec, Canada

John F. Grandfield Grandfield Technology Pty. Ltd. Victoria, Australia

MEMBERS THROUGH 2015 Geoffrey Paul Bearne Rio Tinto Alcan Voreppe, France

James B. Metson University of Auckland Auckland, New Zealand

Geoffrey A. Brooks Swinburne University ofTeclmolgy Hawthorn, Australia

Abdulla Habib Ahmed AluminumAG Bahrain

MEMBERS THROUGH 2016 John Johnson Johnson's Consulting Group Krasnoyarsk, Russia

Benny E. Raahauge Fl Smidth Minerals Copenhagen, Denmark

Olivier Martin Rio Tinto Alcan Saint Jean, France

Trond Furu Hydro Oslo, Norway

Morten Sorlie Alcoa Norway Kristiansand, Norway

XXV11I

I

ORGANIZER

Edwards Rain cn Carbon Kingwood, TX USA

I

Keynote Session SESSION CHAIR

Edwards Rain cn Carbon Kingwood, TX USA

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

RAW MATERIAL IMPURITIES AND THE CHALLENGE AHEAD Stephen J. Lindsay Alcoa, Inc.; Primary Metals; 300 N. Hall Rd. MS S-29, Alcoa, Tennessee, 37701-2516, USA Keywords: impurities, raw materials, vanadium, nickel, sulfur include challenges in parameters like reactive silica or trace levels of elements such as manganese, chromium, and beryllium. Even so, traditional Bayer processing does not pose an impending or direct threat to higher impurities in metal. The challenges for traditional alumina processing may come from a different direction. Some smelting customers may look to alumina sourcing to offset higher impurities of Si0 2 and Fe203 in other raw materials. Pressure on alumina refineries may also become focused on parameters such as CaO and Na20 due to smelting cost pressures. These include costly bath dilution that is driven in part by higher CaO in specific sources of coke. Then there is the cost of excess bath generation. This is driven mostly by Na20 content that is higher than smelter consumption rates. [2]

Abstract

The impurities contained in the raw materials used by the aluminum industry pose challenges that must be managed from various perspectives. These include; product quality, costs, and impact upon the work environment and areas that surround smelters. As the industry continues to grow, impurities, and changes in impurities, will take on greater meaning for process control, equipment design and selection, metal products, and environmental, health, and safety. The author provides his insights into these emerging issues. Introduction

When it comes to impurities in aluminum, market expectations tend to become tighter as the years pass. This has been the trend since aluminum first began being traded as a commodity on the London Metals Exchange. Some regions of the primary metals world such as Australasia have migrated to even higher expectations for standard ingot purity. Refer to figure # 1.

Growth in the global alumina supply is also likely to continue to include some producers with bauxite/raw material that must pass through an acid leaching step prior to Bayer processing. Dealing with higher levels of traditional impurities such as Si0 2, Ga203, and P20 S will pose some challenges. Other impurities such as Li 20, MgO, and even residual amounts of chlorine from acid leaching may create new challenges for bath chemistry, metallic impurities, and even environment, health & safety, EHS, factors.

Minimum Ingot Purity Requirements

Conventional anode technology will require greater consumption of; petroleum coke, coal tar & petroleum based pitch, and perhaps carbon from other sources. Many primary metal producers have taken steps to adjust raw material specification limits to accommodate a changing landscape of what is available in the market. Impurities such as vanadium and nickel are already of growing concern in our industry. Refer to figure #2.

9!1.iIO%

LME 1918

360, 340 -3Z()

lMt1989

Figure 1 - Standard ingot purity requirements for the LME and the Australasian region

1300 :l)zso

Primary aluminum production is anticipated to continue to grow. CRU has estimated that global primary capacity of 47.2 million metric tons in 2012 will grow to 69.4 million metric tons of capacity world wide by 2020 [1].

i.$ 200

ISP :100

Regardless of the accuracy of this prediction the trend is clear. More aluminum production will place greater demand on raw material supplies. Costs of development for new raw material sources will no doubt face making some trade-offs between production capability and impurity levels that are acceptable to markets and to customers.

2004

2005

2006

2001

2008

2008

lOlO

2011

2012

¥(:!-ar

Figure 2 - Examples of trends of vanadium in baked anodes Other impurities such as Ca and Na in coke will pose threats to anode quality and reactivity. Indirectly these have the potential to affect levels of process Fe and Si that enter reduction cells. More reactive anodes are likely to cause greater variation in butt thickness and air burning resistance. Greater variation in anode butt thickness will affect iron contamination of the metal. Greater amounts of carbon dust from more reactive anodes can cause the

Discussion - Raw Materials

Greater levels of aluminum production will require new reserves of bauxite to be extracted for processing. Some is likely to

5

percentage of hot pots to increase. This ultimately concludes with greater input of Fe and Si to the metal.

Depending upon its concentration in bath and the region of the globe there may already be strict and costly regulatory requirements in place that are aimed at worker protection.

Higher metal production rates will also eventually stress reserves of acid grade fluorspar that is desired by the aluminum industry for production of AlF 3. Higher levels of Si0 2 and P2 0 S would be of greatest concern. Special levels of metal purity may then all but require AlF3 that is produced from the fluosilicic acid process.

Another impurity of concern in certain regions of the globe is nickel. There is clear evidence that this trace impurity primarily reports to the metal. But, there is also evidence from a few aluminum producers that small amounts of nickel leave as fugitive emissions that are likely to be in the form of NiF 2. However, this speciation as NiF2 is speculative, not proven.

As the industry grows and as customer demands for purity increase there will be no lack of challenges for primary aluminum producers to face with regard to impurities in raw materials.

At some smelters the mass input rate of nickel from all raw material sources is fully accounted for in aluminum production. At others, up to 30% of Ni does not report to the metal and appears to leave the pots as a fugitive emission. For example, it has been found in wet scrubber sludge. Refer to Figures #3 & #4.

Discussion - Environment, Health and Safety

Impurities can also pose threats that are beyond processes and products. Some impurities of concern, such as sulfur, obviously face challenging environmental regulations. Some regions of the world have placed strong focus on employee exposure to beryllium. Others are beginning to place more focus on elements such as nickel in their emissions inventory reporting requirements.

The variable loss mechanism is not fully understood. It appears to be related in part to cell technology. But, increases of Ni in coke could lead to greater regulatory concern. At this time nickel is only required to be included in smelter emissions inventories, but in a growing number of countries.

With or without changes of %S in anodes a growing fraction of the industry has been moving towards S02 scrubbing. Even projects that are not being built with tail gas scrubbers are often constructing their gas treatments centers, GTCs, such that S02 scrubbing may be added-on later. Some consider this to only be a matter of time before it becomes a standard in the primary metals industry. Increasing levels of sulfur as an impurity in coke can only help to hasten such a change in the status quo.

%Ni in Meta,l- Example #1 100%NlwMetal

Many smelters continue to focus on control of percent sulfur in anode coke as a way to manage S02 emissions. For many of these, especially older facilities, requirements to add tail gas scrubbers would likely cause them to become non-competitive. Therefore, this impurity has the potential to change the standard footprint of primary aluminum production. It also may change the designs of gas treatment systems. GTCs are currently focused on their capability to capture and control total fluoride and particulate emissions with very high efficiency, often at >99.8% capture. If tail-gas scrubbing for S02 becomes an industry norm some tradeofIs are likely that could reduce GTC costs, alumina recirculation rates, and scrubbing efficiency since small amounts of residual HF would easily be captured by wet S02 scrubbers.

1Q%

0% Q-il()Q{l
Rho => Gamma => Gamma' => Delta => Theta => Alpha Conversion

The purpose of the venturi and the first two cyclone stages is to utilize the heat for drying, pre-heating and pre-calcination of the hydrate. The foremost processesl unit operations are suspension of the solids in the gas phase for proper heat- and mass-transfer and gas-solids separation.

The Gas Suspension FurnacelReactor is at the heart of the GSC process:

Calcination in Theory and Practice The overall basic and simplified thermo chemical reactions taken place, when converting alumina hydrate to smelter grade alumina by calcination is:

(/) Pre-calcination (POI-P02): 250 - 3S0°C A1 20 3, 3H20 + Heat => A1 20 3, xH20 + (3 - x)H20, L'lHR (l < x::; 3) = 96.S Kllmole Al 20 3 = 4S.4 K.T/mole H 2 0

(2) Calcination (P04-P03): 3S0 - 1075°C A1 20 3, xH 20+Heat => y-A1203 + (x-y)AI203H20 + yH20 L'lHR (0 a-A1 20 3 + Heat L'lHR = - 23.9 Kllmole Al 20 3

The hot combustion gases entrain the partially calcined alumina from the furnace/reactor (P04) into the Holding vessel (HV03), where sufficient retention time is provided to reach the tinal degree of calcination (LOT, SSA and alpha phase), and thus smelter grade alumina quality. The Furnace/Reactor has the multiple functions of:

? ? ? ?

Figure 4. Phase changes during Calcination [19].

156

Providing sufficient gas retention time for combustion of the fuel at the prevailing temperature; Provide dispersion of the pre-calcined alumina particles entering at about 320°C; Provide fluid dynamic condition for etfective heat transfer between the burning fuel gas and the surface of the pre-calcined alumina particles. Provide sufficient solids retention time for calcination of the pre-calcined alumina particles entering with an LOI (0 - 1000°C) of 7 - 25%, subject to particle size (see

>

Figure 8 below), without generating excessive particle breakdown. Provide fluid dynamic condition for effective mass transfer of water vapor from the surface of the alumina particles and into the bulk gas phase leaving the Furnace/Reactor at an elevated temperature.

HEATR£COVERY

CALCINING

In the above context, calcination of the pre-calcined alumina particles means the four (4) process steps comprising: (1) Conduction of heat from the particle surface through the porous transition alumina phase to the reaction front, where reaction (2) needs heat to take place. (2) Release of solid state "crystal water" as water vapor at the reaction front, when close hydroxyls react through a proton capture mechanism [4]. (3) Diffusion of water vapor from the reaction front in the particle to the surface of the particles. (4) Phase transformation and partial formation of Alpha phase according to reaction (3) in that part of the particle where crystal water has been removed.

Figure 7. FLS Vessel.

esc flow sheet with Fluidized Holding

This design allows the calcination temperature in the Gas Suspension FurnacelReactor to be reduced from about 1050°C to less than 950°C, without sacrificing insufficient gas retention time to achieve complete combustion of the fuel used.

As a result of the above endothermic calcination processes (l) and (2), the calcined part of each solid particle becomes very porous with a large surface area as seen in Figure 6 below.

However, the approximately 100°C lower temperature ofthe gas out of the Furnace/Reactor and into the holding vessel (furnace cyclone), HV03, contain less energy available for the pre-calcination. This means that some of the precalcination according to reaction (1) is moved from P02 cyclone riser into the Furnace/Reactor. This is confirmed by analysing the underflow from the second preheating cyclone, P02, which reveals higher amounts of gibbsitelLOI in calcinations systems with holding vessels compared to calcinations systems without. This has no effect on the final alumina quality as the material is fully calcined in the Furnace/Reactor and Holding Vessel, HV03, anyway as seen from Figure 8 below.

Yarwun K3 Dec. 2011 TM·100tU!789

40

OS:21

Figure 6: Alumina from

esc plant.

30

The internal porosity of the calcined particle makes it relatively easy for water vapour to escape from the reaction front, in the interior of the particle, and diffuse to the surface of the particle. Therefore no excessive water vapour pressure is formed that causes particle breakage contrary to what has been reported elsewhere [4]. On the other hand, Alpha - phase formation tends to start at the outer surface of the particles owing to the relatively low rate of heat conduction through the porous calcined layer of intermediate alumina phases.

II LOI 300-1000 II LOI 300

O(

O(

20 10

o

New esc Process Flow Sheet

Figure 8. Size Fraction LOI (O-lOOO°C) from P02 versus esc Produced Alumina.

The latest design of GSC plant incorporates a fluidized holding vessel, HV03, with several minutes of solids retention time (Figure 7).

The introduction of a proven Fluidized Holding Vessel with overflow discharge, into the new GSC process flow sheet in a Semi - Vertical arrangement (see Figure 7 above), do not in

157

any way reduce the easy, responsive and stable operating characteristics of the GSC process. The advantages and disadvantages of installing a holding vessel is: ./ ./ ./ ./ ./ ./

Reduced calcination temperature => Smaller vessel size => Lower Capex; Lower Alumina temperature to Fluid Bed => Smaller quantity of cooling water => Lower Opex; Less refractory thickness installed to reach the same shell temperature => Less Capex; Lower thermal stress on the refractory lining => Longer service life => Lower Opex; Installation of High Temperature Fluid Bed => Higher Capex; Blowing and Heating fluidization air to stack temperature => Higher Opex; (insigniticant)

Overall, the installation of the fluidized Holding Vessel resulted in lower Capex and Opex. Several other new design features have been introduced into the new generation GSC units in order to reduce the dust circulation around the Fabric Filter, so that all dust is re-cycled completely without the need of any additional process equipment such as a dust lift calciner [ 3], or the necessity of bleeding dust to the Fluid Bed cooler.

Figure 9. GSC Produced Alumina with no Gibbsite. Alumina for Removal of HF in Dry Scrubbers The Specific Surface Area (SSA) is the primary physical property of the alumina specified, for capturing HF gas emitted from the smelting pots, in the Dry Scrubbing, Gas Treatment Centres. However, during calcination, the pore size distribution in alumina has recently been of interest [8].The development of pore size during Gas Suspension Calcination is shown in Table 1 below.

The Hydrate Drying and Pre-HeatinglPre-Calcination (POI & P02) comprise two stages that are very similar in all calcination flow sheets. This is dictated by the drying requirement of hydrate with typically 6 - 9% surface moisture and the thermo chemistry of the calcination process. However, the dust collected from cyclone PO I contains both gibbsite and a fraction of a-alumina phase and constitutes a small Hydrate By-pass.

Physical Pore Pore BET (m2/g) Parameter / size volume (cm 3/g) GSC Design (nm) GSC with HV 4.2 109.2 0.07 P02 material GSC withHV 6.6 0.14 80.6 alumina GSC without HV 8.8 0.3 75.5 Alumina I Table 1. Pore SIze DIstributIOn of GSC Alumma.

In the FLS and Alcoa calciner tlow sheets, dust from cyclone POI enters an ESP (4) or Bag House [4], from where it can be either fully or partially recycled back (dust management) to the calcination process. No gibbsite is detected in the alumina from the FBC (Figure 9) due to all the dust being re-entered into the main alumina stream. This is because the dust enters a cooling cyclone with high enough temperature for complete pre-calcination to take place in accordance with reaction (I). The below TGA DT A curve for alumina shows no presence of Gibbsite.

The pore size distribution of alumina from GSC units, is mono dispersed with an average pore size of 6 - 9 nm. The range 6-8 nm is believed to be optimal [7], though still not proven to be of importance [7, 8] for the HF adsorption capacity of alumina. Here it is worthwhile to remember that the size ofHF and H 20 molecules respectively, is of the order of 0.092 and 0.096 - 0.152 nm, only. According to earlier work [9], there is no apparent correlation between alumina phase composition, pore size, pore volume, surface acidity, MOl or LOI (300-1000°C), and adsorption of HF on alumina plant samples. HF adsorption on plant samples is correlated with SSA, though not totally linear like for laboratory calcined samples from the same hydrate source [9].

Smelter Grade Alumina Quality The primary quality requirements from Smelters are to receive a consistent Smelter Grade Alumina (SGA) quality with each shipment. This is especially so with respect to bulk density and Particle Size Distribution (PSD) for modern cells with point feeders. Ideally this should be possible to a large extent given well known and accepted international specifications [6].This is however not always easy, when considering the many chemical and physical parameters to fulfil [7] owing to the dual application of the smelter grade alumina for (1) removal of HF from the smelter gasses in Dry Scrubbers, before it is used as (2) feed stock for production of primary aluminium in the electrolytic pot/cells. So what can the real impact from Calcination be?

Alumina Degree of Calcination The Degree of Calcination parameters comprises Gibbsite, LOT (300-1000 C), SSA and Alpha phase content, in Alumina from GSC units equipped with and without Fluidized Holding Vessel, and having different calcining capacities as well.

158

The below Table 2 shows typical values of Degree of Calcination in Alumina from GSC Units:

u;

PS., % >45 Mfcron

...............................................................................................................................................................................................

: w t · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .•.........................

'i =

Degree of Calcination

-ltl5Al

..

•.

•

Figure 11: Particle Breakdown 45 Micron (Y-axis) with increasing fraction of hydrate> 45 Micron (X-axis).

Table 2. Degree of Calcination of GSC Alumina. Alumina Strength and Particle Size

It is the coarse particles that break the most [11], when their

threshold value is exceeded, as can be seen from the increase in particle breakdown with increasing size traction exceeding 45 micron in the hydrate to the calcinations plant.

Dry Scrubbers and alumina handling in general requires a strong alumina particle [10], as measured by the Attrition Index (Al), in addition to sufficient Specific Surface Area (SSA) to meet the guaranteed scrubbing efficiency and PSD of the alumina as pot feed. The strength, or rather friability of alumina particles (AI) from any stationary calciner is mainly determined by the hydrate particle morphology (be it mosaic or radial) established in the precipitation circuit of the Alumina Refinery [I I]. However, increasing the heating rates decreases the Al [12], as seen when comparing Al values of alumina from a rotary kiln with Al values of alumina from a GSC [11].

This is not new or surprising information, as it has been shown previously [13], that over coarsening the hydrate particles to meet the requirement of maximum 10% < 45 micron in the alumina shipped [6]. results in the precipitation of less strong hydrate particles. which upon calcinations results in increased particle breakdown. and not a coarser alumina.

Alumina as Primary Aluminium feed stock The dissolution rate of alumina is of major importance in smelters using point feeder technology [14], while crust formation, properties and its subsequent breakage and slow dissolution rate [15] are of major importance in cells with side feeding.

12 10 8

6

However, the dissolution process is improved by good dispersion and wetting of the alumina. by the molten electrolyte bath. when fed to the pot where the following reaction is taken place [14]:

4

2

o

o

5000

10000

15000

20000

25000

Figure 10: Particle Breakdown 45 Micron (V-axis) for Industrial Calciners 12100 - 4500 tpdl versus the parameter AI*U 2 (X-axis),I11I.

In the above reaction (5), wetting is promoted by the volatiles content on the alumina as expressed by the LOI of the alumina. The LOT (300-1 OOOC), on the other hand, shall be less than 1.0% in order to limit the generation ofHF to less than can be efficiently adsorbed in the Dry Scrubber system [4].

The Particle Breakdown seems to increase signiticantly when the parameter AI*U 2 exceeds 7.500 - 8.000, regardless of the plant considered has a Holding Vessel or not. Since the parameter Al*U 2 is a measure of the friability (Al) and kinetic energy (U2) of the particles, when passing the cyclones in the calcinations plant, it seems the particles will break when they exceed a certain threshold value of impact energy or strength.

159

Also GSC units without Holding Vessel can produce alumina with Alpha phase < 5%, if so desired, by adjustment of the fuel distribution into the Furnace/ Reactor (Figure 14). It is however, questionable if it is at all possible to prove a significant difference in smelter operation or performance from a change in alpha phase from, say 4 to say 7 % [4, 7].

SSA and LOI y =0.0122x - 0.056 R2 =0.3134

1.4 1.2 1

OJ·8 ....I

0.6

y =0.0119x - 0.161 R2 =0.3278

0.4 0.2 0 0

100

50 SSA

+GSCl w HV .sGSC3

w HV

150

rn GSC2 w HV • GSC wo HV

Figure 12. LOI (300-1000°C) versus SSA in GSC Alumina. Figure 14. Fuel Distribution by Burner Nozzle Allocation.

Optimization of the SSA versus LOT (300-1000°C) to the 0.8 - 1.0% range benefits the dissolution process. Gibbsite shall be minimized in the alumina in order to avoid "VulcanoEffects" causing excessive HF and dust emissions.

Chemical Purity and Contamination There will be no impact on the chemical purity of alumina from a new GSC process, when using natural gas as fuel, provided a good and durable refractory lining is installed. Except for a very short period of time right after the start up of a new GSC unit, it has not been possible to detect an increase in contamination from Iron and Silica Oxides. Contamination of the alumina with Na20 and CaO, that changes the bath chemistry [16], must be addressed in the Bayer process itself

The above LOI versus SSA (Figure 12) shows how the GSC operator can optimize the alumina in response to smelter demands for GSC units with and without Holding Vessel by adj usting the calcinations temperature. The smelter requirements have led to the general specification of max 10% Alpha phase in SGA [6]. However, to further improve dissolution in modern smelters with point feeders, the Gamma-to- Alpha ratio shall be optimized, and the Alpha phase is specified to be less than or equal to 5%. This requirement is easily achievable in the new generation GSC with Holding Vessel as seen from Figure 13 below. 2DtJIlOO1-iJZ

'*

Energy Efficiency and Specific Heat Consumption The cooling section (COl, C02, C03 and C04) is used for recuperating heat from the hot calcined alumina. This is efficiently done by using four cyclone cooler stages. The stage wise counter-current flow of air and alumina obtained with four (4) cyclone stages of inherent co-current flow, is providing a high thermal efficiency with respect to cooling the alumina and simultaneously pre-heating the combustion air as discussed elsewhere [17].

%

M%

The last section of the GSC system comprises a water cooled fluid bed cooler, which reduces the alumina temperature from 180-200 °C to a temperature low enough for the alumina to be transported with a belt conveyor. The reporting and comparison of the specific heat consumption for different calcination plants is difficult because the number reported depends on several variables beyond control of the calcination process, such as: Moisture content of the production hydrate, typically varying from 5 - 9%, subject to the content ofwt -% of particles < 20 and 45 micron in the production hydrate fed to calcinations? Use or not of dewatering surfactants? Use or not of installed Hydrate By-Pass?

Figure 13. TGA DTA ofGSC Alumina with XRD of alumina showing 6% Gamma and 4% Alpha Phase.

160

> > >

Battery limits of the calcinations process including or excluding the water cooled Fluid Bed Cooler? Credits received for extracting heat in the Fluid Bed Cooler from the hot alumina, to be used elsewhere, i.e. for heating the condensate used to wash the production hydrate before it is fed to calcinations? Types offuel used, and whether reporting and comparison of the specific heat consumption of the calcinations process is based on the High Heating Value (HHV) or Lower Heating Value (LHV) of the fuel?

With the advent of 10m long bags Fabric Filters will most likely be the economic choice of the future as well [18]. Bag life experienced by FLS exceeds 2 year with dust emissions significantly below 50 mg/Nm 3 (dry).

Operational Reliability and Availability The lD-Fan is the prime gas mover in the GSC plant, creating a pressure below ambient throughout the flow sheet, when no Forced Draft Fan is installed. This makes it very easy and reliable to control the pressure profile, and thus the gas flow and calcining capacity of the calcination plant throughout its capacity range, varying from its design capacity to less than 50% of its design capacity.

Unfortunately, the HHV is used as basis for paying the fuel consumption bill. But, only the LHV of the fuel can be used in the calcination process itself. The use of the HHV of the fuel would require that the water vapour from the combustion process is condensed, and that is not taken place within any oftoday's proven stationary calcination technologies. In the case of Natural Gas, the LHV is only approximately 90% of the HHV, while the LHV is approximately 95% of the HHV for Heavy Fuel Oil.

Availability Availability ofGSC units depends to a large extent upon the quality ofthe refractory installed and subsequent quality of the refractory maintenance work performed. Replacement quantity of refractory may range from less than 1% to 20% of original installed refractory per major overhaul for different GSC units without Holding Vessel. In some GSC units, with 36 month between each major refractory repair, availability has exceeded 98.5% in the interim period between major refractory repairs. Still other GSC units have only reached an average of 94%, when including the GSC commissioning period.

The battery limits for reporting or guaranteeing the specific heat consumption of GSC units by FLS do not include the Fluid Bed Cooler, and any credits for utilizing the hot aluminalwater from the Fluid Bed Cooler for other heating purposes. The specific energy consumption for a standard GSC unit with good performing hydrate filters is typically about 2795 kJ/kg based on LHV of the fuel. When decreasing the moisture content in the hydrate the specific heat consumption will decrease about 25 kJ/kgl % decrease in hydrate moisture. When introducing the holding vessel one would intuitively think that the specific heat consumption would decrease due to the lower calcining temperature, and this is also the FLS experience. Understanding specific numbers for comparison is not possible without a clear definition of hydrate filtration pre-conditions, measurement and reporting conditions, and in addition hereto, reporting is commercially sensible.

Environmental Compliance All the latest GSC units now under commissioning are installed with Fabric Filters. Earlier on [3] the selection of a Fabric Filters with 6 m long bags, instead of Electrostatic Precipitators, was often dictated by Fabric Filters absolute filter characteristics minimizing the sensitivity of emission towards process upsets and power failure, rather than from an economic point of view.

Figure 16.

esc units with availability> 98.0%.

It is expected that GSC units with Holding Vessel has the potential to match and even surpass the best experience reported above. Provided, however, good quality refractory work is performed in the first place and recommended operating practice is followed, including the use of the Hot Stand-By facility ofthe GSC unit control system.

Conclusion The recent commissioning experiences with GSC units including several new design features improving specific heat consumption and flexible dust management, shows that the GSC producing SGA at specifications, is as flexible and easy to operate as ever, and at the same time living up to the expectation from the design stage.

References 1. T.A.Venugopalan,"Experience with Gas Suspension Calciner for Alumina", Proceedeings 1st International Alumina Quality Workshop, pp 53-66,(1988).

Figure 15. Fabric Filter with 6 m long bags.

161

2. 3. 4.

5. 6. 7.

8. 9. 10. II.

12 . 13. 14.

15. 16. 17. 18. 19.

Li Zhaoxia et all, "The Key Technologies for Energy Efficient Al(OH)3 Dilute Phase Fluidized Bed Roasting Furnace",TMS Light Metals,(2012). .T.Fenger et all, "Experience with 3 x 4500 tpd Gas Suspension Calciners (GSC) for Alumina",TMS Light Metals,(2005). L.M.Perander et all," Impact ofCalciner Technologies on Smelter Grade Alumina Microstructure and Properties" Proceedeings 8th International Alumina Quality Workshop, pp 103-107 (2008). K. Yamada, "Dehydration Products of Gibbsite by Rotary Kiln and Stationary Calciner" TMS Light Metals, pp 157-171, (1984). S.J .Lindsay,"SGA Requirements in Coming Years" ,TMS Light Metals, pp 117-122 (2005). A.J .Meyer et all, "Examination of Drop in Bath Acidity due to Change-Over of Alumina Qualities in the Sunndal Aluminium Smelter, Norway", " Proceedeings 9th International Alumina Quality Workshop, pp 316321, (2012). L.M.Perander, J.B.Metson and C.Klett,"Two Perspectives on the Evolution and Future of Alumina",TMS Light Metals, ppI51-155, (2011). J.F.Coyne et all,"The Influence of Physical and Chemical Properties of Alumina on Hydrogen Flouride Adsorption",TMS Light Metals, pp 35-39,(1987). Chandraskar, S. et all,"Alumina Fines' Journey from Cradle to Grave". Proceedings ofthe 7th International Alumina Quality Workshop, Perth, Australia, 2005. S.Wind,C. Jensen-Holm and Benny E. Raahauge,"Development of Particle Breakdown and Alumina Strength During Calcination", TMS Light Metals, (2010). .T.D. Zwicker, 'The Generation of Fines due to Heating of Aluminium Trihydrate", TMS Light Metals, pp 373395 (1985). .T.V. Sang:"Factors Affecting the Attrition Strength of Alumina Products", TMS Light Metals, pp 121127,(1987). B.J.Welch, A. Alzaroni and S.Lindsay,"Modern Smelting Technology and its Impact on Alumina Requirements",ICSOBA Proccedings, Zhengzhou, PRC, (201 0). .T.Gerlach and G.Winkhaus,"Interactions of Alumina with Cryolite-based Melts", TMS Light Metals, pp301313,(1985). S.J.Lindsay,"Customer Impacts ofNa20 and CaO in Smelting Alumina". TMS Light Metals, (2012). S.Wind and B.E.Raahauge: "Energy Efficiency in Gas Suspension Calciners (GSC)",TMS Light Metals, (2009). C.V.Rasmussen and H.V.Pedersen," Fabric Filter Operating Results with 10 m Long Bags and Low Purging Pressures",TMS Light Metals, (2012). Wefers,K. and C.Misra, eds. Oxides and Hydroxides of Aluminum. Alcoa Technical paper No. 19. 1987, Aluminum Company of America: Pitssburgh,P A.

162

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

Bayer Process Efficiency Improvement Gu Songqing China Aluminum Corporation Limited, No.62 Xizhimenbei Street, Beijing 100082 Keywords: alumina production, Bayer process, process efficiency, productivity, energy consumption

Abstract

It is known from the systematic analysis and study on Bayer

The process efficiency has a great influence on the production yield, energy consumption, investment and operation cost in Bayer process.

process that the efficiency in Bayer process is closely related to three important parameters: caustic concentration of spent liquor, alumina extracted from bauxite during digestion and alumina precipitated from pregnant liquor during seeded precipitation. In this paper the factors are analyzed in detail and the possibility to develop technology for process efficiency improvement is discussed.

The factors affecting the Bayer efficiency are particularly discussed in this paper. Reducing the molar ratio of pregnant liquor and raising the molar ratio of spent liquor and the caustic concentration of spent liquor will enhance whole output efficiency in Bayer process. Higher concentration of spent liquor will bring about higher productivity in digestion process and a higher pregnant liquor concentration will be required for reducing energy consumption at evaporation stage as well.

Analysis of the factors impacting Bayer efficiency Bayer process is mainly composed of four stages: bauxite digestion, red mud separation, seed precipitation and liquor evaporation as shown in Fig. 1.

The key technology development to improve the Bayer efficiency in Chinese alumina production is revealed.

Introduction The energy consumption is of great significance to the operation cost in all the alumina refineries, about 40% of which is related to energy consumption in China including vapor for Bayer digestion and evaporation, gas for calcination and power for pumps, agitators and fans etc. Nevertheless the energy consumption depends on the Bayer process efficiency to a great extent. The process efficiency has a great influence on the production yield, energy consumption, investment and operation cost in Bayer process. Figure 1. Typical Bayer Cycle

The efficiency improvement will lead to more productivity for the same flow rate and energy consumption reduction for unit output in the Bayer cycle. So the energy saving could be achieved not only by application of energy saving processes and equipment, such as fluidized calcinations instead of rotary kiln, indirect preheating in digestion process, efficient red mud settling and sweetening process etc. but also by increasing the Bayer process efficiency without much change for the process and equipment.

In Bayer process red mud separation and liquor evaporation are basically physical processes without major chemical reactions to take place. Bauxite digestion and seeded precipitation are the main chemical reactions. And the processes occurred in the stages can be considered as the reverse reactions: Al(OH)3 (or A100H) + NaOH

Bayer efficiency improvement becomes more important and necessary since the technology and equipment in almost all the stages in Bayer process have been up-to- dated in the past years. Energy shortage and its price rise have driven alumina producers to reduce energy consumption and operation cost. The best way is to improve Bayer process efficiency.

.... NaA10 2 + H 20

It can be considered that the Bayer process is a stable cycling process and the spent liquor for digestion remains in a same composition after every cycle, that means its caustic and alumina concentrations will keep the same. And it is supposed that all the alumina extracted from bauxite in digestion will decomposed to alumina hydrate in precipitation process without any loss to residue.

A further study on the factors to influence process efficiency and how to improve efficiency should be carried out for the relevant technology development.

Based on the above the alumina output (A) for every cubic meter of spent liquor in a Bayer cycle could be considered as follows:

163

E=Ap -As

It can be seen from Figures2 and 3 that the higher Bayer

efficiency can be achieved by increasing caustic concentration and reducing MRp, i.e. MR of pregnant liquor.

where: Ap is the alumina concentration in pregnant liquor in

180.0

giL

I'f')

...6

As is the alumina concentration in spent liquor in giL

170.0 160.0 150.0 140.0

E is the Bayer efficiency in kg/m3

130.0 120.0

Since the molar ratio of Na20 to AI 20 3 in liquor will not be changed in flash tanks and during slurry dilution a technical term (molar ratio MR) is used for efficiency study:

110.0 100.0 90.0

2.6

2.7

2.8

2.9

3.0

Spent Liquor MRs

MR =1.645*Nk/A Figure 4. Bayer efficiency E vs. MRs

in which A is alumina concentration in giL; Nk is caustic concentration in giL

Figure4. shows that the higher MRs of spent liquor is the higher Bayer efficiency can be achieved.

Therefore,

In order to increase Bayer cycling efficiency in a refinery three kinds of process improvement can be carried out:

E=1.645*Nk*(IIMRs-IIMRp) where. MRs and MRp is the molar ratio of spent liquor and pregnant liquor respectively; and the caustic concentration Nk is considered unchanged from the beginning to the end during digestion.

(2) Reducing MRp;

So the Bayer process efficiency E is directly proportional to caustic concentration of spent liquor and related to MR change before and after digestion.

MRs usually is in the range from 2.5 to 3.2 and MRp is between 1.2 and 1.6 for a general Bayer process.

'6

(I) Increasing caustic concentration Nk;

(3) Increasing MRs.

It can be found by comparison of Figures 3. and.4. that slopes of

180.0

the curves for E vs. MRp are much higher than the slopes ofE vs. MRs. The detai led calculations show that the increase of E by 0.1 reduction of MRp will be about four times higher than that by 0.1 increase of MRs. Therefore reducing MRp will lead to a greater Bayer efficiency improvement than increasing MRs.

170.0 160.0 .:01. 150.0 C 140.0 E 130.0 t 20.0 ;::s 110.0 100.0 90.0 80.0 70.0

OJJ

;g

160

180

200

220

240

The most important chemical reactions happen in the digestion and precipitation stages in the Bayer process. So the Bayer efficiency improvement is closely related to the reaction efficiency and productivity in both stages.

260

Caustic concentration Nk,g/L ---MRJF1.3 -MRJFJ.45

-MRJFJ.35 -o-MRJFI.5

The digestion temperature, caustic concentration and MRs are the major technical parameters for bauxite digestion. And the digestion efficiency and MRp are the major results to impact Bayer efficiency. The higher caustic concentration Nk and MRs will increase digestion efficiency and reduce MRp, which bring about higher Bayer efficiency.

Figure 2. Bayer efficiency E vs. Nk

t

180.0

.:01 150.0 140,0

Precipitation is the longest stage in Bayer process and its productivity determines the output efficiency for whole Bayer process to a great extent. The higher precipitation efficiency (lower MRs) can be achieved by seed addition, temperature control. residence time and suitable caustic concentration. The higher caustic concentration in precipitation will probably lead to reducing the precipitation efficiency but higher precipitation output. So a trade otf analysis should be considered for the best results.

&

130,0 '-l

120.0 110.0 100,0

90.0 80.0 70.0

1.35

1.3

1.4

1.45

1.5

MRp of Pregnant Liquor Nk=l60 -Nk=180 -Nk=200 -Nk=240 ---Nk=260

Figure 3.

Reduction of the caustic concentration difference between digestion and precipitation is very important for the energy saving

Bayer efficiency E vs. MRp

164

at evaporation process. So a higher Bayer efficiency can be achieved by increasing Nk of spent liquor and a systematic energy saving can be obtained by higher Nk of pregnant liquor as well.

Reducing the molar ratio of the pregnant liquor Reducing the molar ratio of pregnant liquor as much as possible is the most important technological solution for improving Bayer efficiency.

The optimization of concentration system in whole Bayer cycle is very important for higher Bayer efficiency and lower energy consumption in the retineries.

Relatively lower MRp can be obtained by intensifYing digestion process including suitably raJsmg digestion temperature, increasing digestion time and caustic concentration. Digestion temperature is the key condition to intensify the process and depends on the equipment limitation, heat resource, energy and water balance etc.

Technological solutions to improve Bayer efficiency It is well known that the operation cost of alumina production in the refineries mainly depends on energy consumption, operation efficiency and investment cost etc.

The intensifying digestion technology has been developed and applied in China since 20 years ago. Since all the Chinese retineries have to treat diasporic bauxite the higher caustic concentration Nk and higher digestion temperatures are used for a better digestion efficiency. Figure 5, shows a technical scheme of intensitied digestion process.

All the factors are closely related to the operation efficiency in whole Bayer process and at the different stages. As discussed above it is concluded that the best way to achieve a higher Bayer efficiency and energy savings are to reduce molar ratio of pregnant liquor, relatively increase the molar ratio and caustic concentration of spent liquor and decrease liquor concentration differences as much as possible.

The problem for the intensifying digestion is the lower digestion operation rate due to fast scaling on the indirect preheating surface. The technical solutions developed in China are as follows:

Firstly all the stages in Bayer process should be intensified for a higher stage efficiency, i.e. higher digestion efficiency, precipitation efficiency and evaporation efficiency etc. The intensification of the process stages by improvement of process parameters and application of more efficient equipment will provide better process conditions and higher caustic concentrations for the chemical reactions to complete in a shorter period and to save energy and raw materials, such as bauxite, caustic and lime etc.

(I) Reducing scale by enlarging the preheating surface and increasing slurry flow rate; (2) Installing more spare equipment and developing fast switch technology for scale removal; (3)

Secondly the systematic optimization of the Bayer process is another important technical solution for improving Bayer efficiency, which includes optImIzations of the caustic concentration and molar ratio systems in whole Bayer process.

Setting up predesilication tank(s) before digestion and in the stage to reduce reactions of silica minerals on preheating surface.

The operation rate has been greatly increased by scaling inhibition mentioned above. On the other hand on order to improve Bayer efficiency higher digestion caustic concentration and MRs should be provided, which depend on the Nk and efficiency in precipitation stage and evaporation.

A series of key technology have to be developed to achieve the goals of both intensifYing all the stages and realizing systematic optimization of Bayer process.

Efficient Crushing & Grinding

Retention Tanks with Stirs

Lime Milk

Tanks in Seies

Dilution Tanks without Vapor Emission _ _

Spent Vapor

...

K,.......,L.----,[] ...

Efficient Multistage Flash Tanks

,.......,L...---'[]D

Figure 5 Schematic diagram forthe intensified digestion stage

165

Retention Tanks without Stirs

To keep accurate recipe during bauxite slurry grinding is necessary for obtaining both better MRp and higher bauxite digestion efficiency. An online bauxite and slurry test technology is developed for this purpose.

A key precipitation technology with higher efficiency for sandy alumina has been developed in China. The technical solutions are:

The key technology of intensifying digestion is widely applied in Chinese refineries for both higher digestion efficiency and lower MRp at the higher Nk.

(2) Optimizing the temperature reduction gradient during precipitation by using well designed intermediate heat exchangers;

(1) Increasing Nk and reducing MRp for higher precipitation productivity;

(3) Forecasting of hydrate size distribution change by regular testing the minor hydrate size and analysis and control system.

In order to make MR of pregnant liquor for precipitation as close to that of bauxite digestion slurry as possible the precipitation which might occur in the red mud settlers and washers should be alleviated, i.e. the MR difference between both liquors should be reduced. The best way for this is to keep the settling process at higher temperatures and Nk.

(4) Applying the additives for a stable hydrate size distribution. There has been a great progress in the field of reducing hydrate suspension from 2-10 grams per liter to less than 1 giL in the filtrate from hydrate filtration.

Sweetening process is an effective solution for lower MR of pregnant liquor and higher productivity at precipitation ..

Optimizing liquor concentration system

Increasing molar ratio of spent liquor

Optimization of the liquor concentration system is carried out based on the principles: greater Bayer efficiency, larger productivity at the main stages and smaller liquor concentration difference between the digestion and precipitation processes for reducing energy consumption at evaporation.

Increasing MRs of spent liquor is a key factor to improve Bayer efficiency and digestion productivity. Increase of MRs mainly depends on the precipitation efficiency and reduced hydrate suspension in spent liquor after hydrate filtration. The precipitation efficiency basically relies on the precipitation temperature system including the liquor primary and final temperatures and the temperature reduction gradient in the process. And it also depend on the seeding system including seed rate, seed size distribution and washing and addition ways, and liquor caustic concentration, precipitation time and additives as well. The lower precipitation temperature and MRp and higher seed content will make higher MRs and precipitation efficiency, which mainly relies on the process parameters control and optimization.

The relatively higher spent liquor concentration is more important for diasporic bauxite digestion and anelevated digestion productivity will be obtained. Nevertheless a higher precipitation concentration is needed for reducing evaporation energy consumption. Increasing pregnant liquor concentration might have a negative effect on precipitation efficiency and sandy alumina production. So the optimized caustic concentration system has to be specially designed according to the status of the process parameters and major equipment. Figure 6. shows the optimized high efficient Bayer process with series of key technologies including intensified and efficient digestion, low loss settling, efficient precipitation for sandy alumina and energy saving evaporation ..

Efficient Crushing & Grinding

High Caustic & Low MRp Digested Slurry

Lime Milk

High Efficient Precipitation for Sandy Alumina

Efficient Hydroxide Filtration

High Caustic & Low MRp Pregnant Liquor

Figure 6 Schematic diagram of optimized high efficient Bayer process

166

It is concluded that the design of a new alumina refinery should be

optimized based on both technological and economical considerations for a relatively higher operation efficiency and productivity, lower raw materials and energy consumption. And for an old refinery some process parameters are supposed to be improved for achieving the same goals.

Conclusion (I) The Bayer efficiency mainly depends on the caustic concentration of spent liquor, MR of pregnant liquor and spent liquor. The Bayer efficiency will be improved by increasing liquor caustic concentration, reducing MR of pregnant liquor and enhancing MR of spent liquor. (2) Reducing MRp will lead to the greater Bayer efficiency improvement than increasing MRs. (3) Bayer efficiency is improved by increasing liquor caustic concentration, which will bring about lower precipitation efficiency and higher evaporation energy consumption. So a optimized caustic concentration system should be designed for both higher Bayer efficiency and lower energy consumption. (4) The key technologies to intensify all the stages in Bayer process will provide the basic conditions for high Bayer efficiency. (5) An optimized caustic concentration and MR system are put forward to build a high efficient Bayer process for systematic energy saving and all the consumption reducing.

References [I] Qi Lijuan, Gu Songqing Wang Qingwei. "The Comprehensive Energy Saving in China Alumina Industry" Light Metals 2005 p.35-40 [2] Yang Zhongyu. "Alumina Production Processes" (in Chinese). The Metallurgical Industry Press. Beijing 1993 p.47-50 [3] K. Solymar et al. "Co-processing of Different Types of Bauxite with High Efficiency". Light Metals 2001 p.105III [4] 1. Anich et. al. "The Alumina Technology Roadmap". Light Metals 2002 p.193-198 [5] Gu Songqing, Wu Jianqiang "Review on the Energy Saving Technologies Applied in Bayer Process in China" Proceedings of the 9th International Alumina Quality Workshop 2012 Perth, Australia p.379-384

167

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

HYCLASS™ TECHNOLOGY FOR IMPROVEMENT OF TRIHYDRATE CLASSIFICATION IN THE BAYER PROCESS ling Wang l , laqueline Herrera l , Shawn Kostelak2, Kody Frederic2 lNalco Company, 1601 West Diehl Road, Naperville, IL, 60563, U.S.A. 2

Noranda Alumina LLC, 1111 Airline Highway, US 61, Gramercy, LA 70052, U.S.A. Keywords: trihydrate classification, alumina, flocculation, Bayer process demonstrated some capacity to reduce overflow solids and increase hydrate settling rate, there has been a clearly established need for further improvements in the classification and flocculation of trihydrate in Bayer process. The benefits of further reducing the overflow solids in the tertiary settlers include: (I) a direct net production increase, (2) a reduction in heat exchange scale and associated energy consumption [5], and (3) a reduction in the quantity of trihydrate that must later be redigested. In addition, the operational efficiency of the tertiary thickeners, seed dewatering and precipitation can be impacted significantly through (4) faster settling rates, (5) improved settling rate doseresponse at higher feed solids and (6) marked improvement in settled hydrate underflow density and rheology.

Abstract

The classification section in the Bayer plants is a critical part of the process. The final separation of the finest hydrate particles from the spent liquor takes place in the tertiary thickener vessels. Efficient capture of trihydrate particles in the classification circuit is essential to improve the productivity of the Bayer process. It is widely accepted that the efficiency of the classification circuit can be improved by the use of additives. Nalco is continually developing new polymer chemistries to improve the efficiency of trihydrate flocculation. A novel hydrate classification technology, HyClass™, was developed by Nalco and field evaluated in the tertiary thickener units of Noranda Alumina. HyClass technology significantly improves the flocculation performance against an industrial standard 85700 with respect to increased hydrate capture, improved settling dose-response, higher underflow densities and most importantly, further improved slurry rheology. The lab development and field evaluation work is summarized in this paper.

Therefore, it is highly desirable to develop a novel polymer chemistry with superior performance for hydrate flocculation, which is the objective ofthis work. In current work, a new hydrate classification technology, HyClass, was developed by Nalco and field evaluated in the tertiary thickener units of Norand a Alumina. This paper summarizes the lab development and field evaluation work for HyClass technology as trihydrate ±1occulant in Bayer process.

Introduction

In the Bayer process for the production of alumina, bauxite ore is pulverized, slurried in a caustic solution, and then digested at elevated temperatures and pressures. After removal of red mud impurities, the clarified green liquor is then cooled to a supersaturation condition and seeded with recirculated fine aluminum trihydrate crystals in precipitation tanks to induce the precipitation of alumina trihydrate from liquor. These trihydrate particles are then separated in the classification circuit according to Stokes law - coarse particles settle easily, and fine particles settle slowly. In particular, in the final classification step (tertiary thickener), the fine particles settle so slowly that it becomes very difficult to separate them from the spent liquor, potentially resulting in trihydrate production loss. In addition, the poor underflow rheology of densely compacted fine trihydrates can lead to premature tank failures due to excessive rake torques, scale formation in classification vessels and reduction in seed filtration efficiency. Therefore, within the settling steps of the classification system, trihydrate flocculants can be used to enhance particle capture, settling rate and improve underflow rheology. As a result, the efficiency of thickeners and the process can be improved.

Experimental

In the initial lab development, the ±1occulation performance of HyClass products was compared to 85700 benchmark using (1) cylinder settling test, (2) Imhoff cone method and (3) Focused Beam Reflectance Measurement (FBRM) [6]. Field evaluation was conducted in the tertiary thickener units at Noranda Alumina in Gramercy, LA. Cylinder Test In the cylinder test, I-L bottles of Secondary Overflow slurry (containing about 50 giL solids) were collected and stored in an oven at 75°C. For a given test, a bottle was removed from the oven, shaken to re-suspend the hydrate solids, and, then dosed with a specific amount of flocculant solution. This bottle was then mixed by hand to allow the flocculant to contact the solids for 1.0 minute, prior to transferring the slurry into a I L graduated cylinder. The settling hydrate interface was recorded at I minute intervals over 4 minutes to calculate the hydrate settling rate. The amount of solids in the overflow of each sample was determined after 4mins of settling by standard methods in the lab.

The conventional flocculant technologies, synthetic water soluble polyacrylate flocculants and/or dextran flocculants, are used to improve the settling characteristics of the alumina trihydrate particles in the classification process and reduce the amount of solids in the spent liquor [1-4]. While various flocculants are often used in classification circuit of Bayer plants and have

169

Imhoff Cone Method In the Imhoff cone test, I-L bottles of Secondary Overflow slurry (containing about 50-140 gIL solids) were collected and stored in an oven at 75°C. For a given test, a bottle was removed from the oven, shaken to re-suspend the hydrate solids, then dosed with a specific amount of flocculant solution. This bottle was then mixed by hand to allow the flocculant to contact the solids for 1.0 minute, prior to transferring the slurry into aIL Imhoff cone. The hydrate settling rate and overflow clarity were determined using a similar method as described in Cylinder settling test procedure The tlowability of the compacted hydrate solids was measured by determining the elution time needed for the hydrate to pass out of the Imhoff cone after 15 minutes of compaction.

$@ttlnd.@fY

Cb,ili.,

Focused Beam Reflectance Measurement (FBRM) Figure I. Diagram of HyClass flocculant application Focused Beam Retlectance Measurement (FBRM) allows for insitu measurement of particle (tloccule) size by insertion ofa probe into the continually agitated slurry. The probe scans a highly focused laser beam at a fixed speed across the hydrate particles in the slurry and measures the back scattered light or the duration of the reflection of the beam from the particles. The backscattered light is expressed as chord length, which is related to the particle size. The unflocculated particles have short mean chord lengths. As flocculation occurs, the mean chord lengths increase and the total number of discrete particles decreases. In each test, 200mL of slurry was prepared comprising of 50g/L aluminum trihydrate solids and spent liquor (equilibrated at 60°C in 250ml Nalgene bottles for 1 hour). A commercially available fine standard aluminum trihydrate seed was added to each bottle and mixed for 30 seconds. The slurry was poured into a 250ml glass reactor connected to a hot water bath (60°C). The FBRM probe and stirrer were then placed in the reactor. After Imin of continuous agitation, data acquisition with FBRM started. After a certain point of time, a specific amount of flocculant solution was added. The information on the change in the mean chord length (which is proportional to the real particle size) was obtained over time.

Results and Discussion Lab Development Impact of HyClass on Trihydrate Capture and Settling Rate

Plant use of trihydrate flocculants is traditionally justified on the value of solids captured, which is the reduction of overflow solids in tertiary thickener vessels. Therefore, to evaluate the tlocculation efficiency of HyClass flocculant against the 85700 benchmark, cylinder settling tests were conducted on seed secondary overflow. Figures 2 and 3 show the clarity and settling rate results of each flocculant versus dosage. It is apparent that the addition of HyClass flocculant significantly improves the overflow clarity and hydrate settling rate compared to 85700. In detail, with the treatment of 0.6ppm 85700, the overflow solid was 17g/1. However, with HyClass product at the same dosage, the overflow solid was only 5g/1. Furthermore, as shown in Figure 3, with the treatment of HyClass flocculant at 3ppm, a settling rate of over 14ft/hr was achieved. At the same dosage, with 85700, the hydrate settling rate was only about 7ft/hr. It should be noted that a plateau in the settling rate is observed with 85700, but with the HyClass product, settling rate increases linearly with dose.

Field Test The trial was conducted in two of the three operating tray thickeners at Noranda Alumina during a five-week test period. The product was applied neat to each tray feed line (Figure 1). The current hydrate flocculant feeding system was used to run the test without any additional requirement of equipment, pumps or pipelines. During the trial, the feed slurry, tray overflow and underflow samples were collected three times per day. The solids (gpl) in each of the samples was determined by standard procedures in the plant quality Lab.

170

As shown in Figure 4, compared to the blank and 85700, the settled hydrate flowability was increased significantly using the HyClass product. The flowability of settled trihydrate for the blank was very poor (only 0.1 Umin). With 85700, the flowability of flocculated material increased to 0.4 Umin, and reached a plateau. However, the addition of HyClass flocculant can increase the hydrate flowability dramatically to about 1.9 Umin at the same dose. As the product dose of HyClass is increased, underflow rheology was continually improved. This result clearly demonstrates the substantial impact of the new flocculant on the tlow properties of settled trihydrate.

__

- - HyCbss

5.00

-""" 1.00

tOO

....

Ud

--

Similar tests were subsequently conducted with various feed slurry solids from 50 to 140g/1. From Figure 5, it is apparent that the flowability of untreated settled hydrate diminishes as the feed solid loading is increased from 50 to 140g/1. Treatment with 85700, improved the hydrate tlowability to some extent. However, with addition of HyClass flocculant, a dramatic increase in hydrate flowability was achieved. In detail, for the Blank and 85700 treated hydrate, the hydrate tlowability is less than 1 and 2 Umin respectively at a solid loading of 50g/1. However, with HyClass product, even at a much higher solid loading of 140g/l, the same hydrate tlowability of 2Umin was maintained. Therefore, HyClass technology shows great promise in allowing plants to operate thickeners at a more desirable condition of higher underflow density without the risk of unacceptable rake torques or "rat-holing".

3.00

jlpm Pj'

LL

Jajarm Alumina plant has been designed to produce Alumina based on the Chamosite-Diasporic domestic Bauxite by Bayer method [1]. Due to some mistake in its digestion unit design, plant commissioned on March 2002 with Gibbsitic imported Bauxite from India, and after doing remedy actions, on May 2003, has been switched to domestic bauxite as the plant feed. Based on the quality, quantity and mineralogical characteristics of Elburz Bauxite, and also economical consideration, all parties agreed to feed plant with a Bauxite AJS ratio around 4.4-4.7 coming from nearby mine. Now, after 8-9 years exploring and exploiting Jajarm mine by open-pit method, not only a large amount of low-grade, Kaolinite and Shale Bauxite has been piled in the mine but also scare up Bauxite with an AJS ratio in the agreed range become very difficult. So very intensive investigations have been started from 2007 to find processing methods to increase domestic Diasporic Bauxite quality which feeds the plant and also capable to produce a reasonable feed from low-grade, Kaolinite and Shale Bauxite piles.

Mineral

@J

Chemical Formula

....0

@J

Diaspore

AIO.H 0

47.05

Hematite

Fe 0

14

11.67

11.6

Kaolinite

2SiO .AI 0 . 2H 0

8.95

13.8

Anatase

TiO

2

3

2

2

5.01

3.7

SiO .'/16 AI 0 .'123M gO.Y, I< O. 'l,sFeO. %H 0

3.49

4.5

Fe 0 .H 0

1.49

3.9

Quartz

SiO

1.24

6

Calcite

CaCO

1.19

2.3

cancrinite

2

2

2

Maghemite Pyrite Magnetite

2

2

2

3

2SiO .AI 0 . Na O.%CaCO

1.0

2.4

AIO.H 0

1.0

1.9

TiO

0.66

0.63

0.61

0.8

l.5AI 0 .P 0 .CaO. 7/zH 0

0.3

0.2

CaCO . MgCO

0.24

0.1

.

4000

:54000

'JJ.

§

i o. 61°",:;6 (011) 9.11

.Co3AI, (Si0 4) (OH) 8

2000

"

2000

20

28

40

n

60

21O"C

80

20

100

40

60

80

2 e (0) 240'C

100

Fig.3 XRD patterns of calcified slag at different temperature Fig.3 shows that silicon phase changed into hydrogarnet and

sodium-silicon residue.

and

3000,------------------------,

calcium

aluminate hydrate phase is existed when calcification

2500

temperature is lower than 200'C. When the calcification

the silicon begin to dissolved in calcium aluminate hydrate

Ca3AIFe(Si04)(Oll)g

•

Na(jLAlSi0 4 J(j'411 2 0

•

Fe20]

*--Si0 2 2000

temperature get above 200'C, the calcium aluminate hydrate phase disappeared with the dissolution effect of silicon, and

....

Din:'clly digeslioll I'ed lilud

""

1500

1000

phase and change into hydrogamet. At the calcification temperature of 240 ·C, the aluminate hydrate phase

500

disappeared completely, so (»240'C is considered as a 20

optimal calcification temperature. The calcification process of

40

60

80

2H( • )

FigA XRD patterns of calcification slag of gibbsite

247

100

is

Through calcification treatment, the calcified slag was

diasporic bauxite is same with the pure materials synthesis experiment.

treated with carbonation process, the reaction of this process is:

Calcification process of gibbsite

3CaO· AI20 3·xSi0 2·

XRD patterns of calcification slag of gibbsite are shown in

xCa2Si 04+(3-2x)CaC03 +2AI( OH)3+(3-2x)H20 Through carbonation treatment, hydrogarnet phase is

Fig.4. The calcification conditions of gibbsite is: molar ratio of CaO:Si0 2=3:0.64,

calcification

concentration of mother liquid of 140g/L,

Na20

changed into Al(OH)3' 2CaO· Si0 2 and CaC0 3, the AI(OH)3 is

of mother liquid of

digested by alkali when the digestion temperature is lower than

of

time

ak

30mins,

3.1. Fig.4 shows that through calcification treatment, high iron

100 "C, and the main phase of the new type red mud is

content in gibbsite lead the part of Fe203 instead A1 20 3, and the

2CaO·Si0 2 and CaC0 3 We have researched the carbonation

main phase of calcification slag is iron-hydrogarnet. The

conditions such as temperature, CO 2 pressure, time and

optimal calcification temperature is 180"C.

liquid-solid ratio on the carbonation effects. The result at the

Carbonation oerformance of diasooric bauxite and gibbsite

optimal conditions is shown in Table 2

Table 2 A/S and chemical content in new type red mud through calcification-carbonation treatment Materials

Carbonation times

Diasporic bauxite

Gibbsite

Experiment result AI2Oi%

SiOi%

Na201%

A/S in new type red mud

I

11.09

11.02

--

1.01

2

4.41

10.02

0.12

0.44

I

11.27

10.16

0.67

1.11

2

8.49

10.32

0.57

0.82

Table 2 shows that through calcification-carbonation

silicon phase of bauxite into hydrogarnet, so in the new method,

treatment, the A/S and Na20 content in new type red mud by

lime addition is higher than that of Bayer process (mass of lime

using diasporic bauxite lowed to 0.44 and 0.12%, meet the

addition amount for diasporic bauxite is m(CaO):m(Si02)= 1:

requirements of cement industry. But for gibbsite, although the

1, in Bayer process of gibbsite lime addition is 0, but in

AlS is lower than 1, but it is also higher than that of diasporic

calcification-carbonation process is n(CaO):n(Si02)=3:0.64,

bauxite, the main reason is the optimal carbonation conditions

m(CaO):m(Si0 2)

of iron-hydrogarnet is different from hydrogarnet, the effect of

calcification slurry is about 2-2.5 times of Bayer digested slurry.

carbonation

Tn this section, settling performance of calcification slurry is

condition

on

the

decomposability

of

= 4.375:

1). So the solid content in

iron-hydrogarnet is in researching step.

researched firstly, and the settling speed of calcification slurry

Separation performance of calcified slurry

and Bayer digested slurry by using diasporic bauxite and gibbsite is shown in Fig.5.

The propose of calcification process is to change the

-----Gayer slurry @-----Cal c; fi cd slurry

_______ Kaycr slurry

100

-@-----

Ca lei fi cd slurry 80

80

':; CD

(b)

60

(a)

40

40

i

I

....l

20

20

1 10

20

30

40

50

10

60

Sott ling t i mc/mi n

m

mm

20

i

30

III 40

Scttl ing time/min

(a)Diasporic bauxite

(b )Gibbsite

Fig.5 The height of settlement layer go with time under different lime addition

248

50

60

The settling performance experiment results in Fig.S show

so it is the main reason of settling speed of calcified slurry is the

that:

same with Bayer slurry. (I) For diasporic bauxite, the solid content of calcified

For finding the reasonable separation method for the

slurry effects on the settling performance is great, the settling

calcified slurry (especial for the calcified slurry of diasporic

speed of calcified slurry is much more lower than that of Bayer

bauxite), pressure filtration performance of calcified slurry were

slurry. The reason is: in Bayer or calcification process for

tested. In this paper, the filter constant K is calculated to

diasporic bauxite, the Na20 concentration in the mother liquid is

represent the filter performance of the calcified slurry. The

higher than that of gibbsite (about 240g/L), the solid content of

calculation method for K is:

Bayer slurry is about 120glL, and in the calcified slurry is

Constant pressure filtration equation:

higher than 200glL. High solid content lead the settling slurry

(2)

change into interference settlement status, and the settling speed

Where q is volume of filtrate of unit area;

decreased greatly. (2)

For

qe is virtual volume of filtrate of unit area;

gibbsite,

although

the

lime

addition

B is actual filtration time:

is

n(CaO):n(Si02)=3:0.64 as well, but the Na20 concentration in

Be is virtual filtration time;

the mother liquid is about 140g/L, and the total gibbsite and

K is filter constant.

lime addition amount is lower than that of diasporic bauxite,

From equation (2), we get:

which lead the solid content of calcified slurry is lower than

(3)

200g/L, the interference settlement phenomenon do not So when we get the liner relationship between LYJIL!:q and

happened in settling process. Meanwhile, in the Bayer process of gibbsite, goethite in settling process is easy to change into

q, the slope is 21K.

colloid Fe(OHh, which will decrease the settling speed of

The liner relationship and between LJBI LJq and q for

digested slurry. But in calcified slurry, Fe content changed into

diasporic bauxite and gibbsite at different pressure are shown in

iron-hydro garnet, which will not effect the settling performance,

Fig.6 and Fig.7.

0000,-------------------------, 6000

'" "

S

=

.

=

",.

2000

5

1500

//,/'

4(ID /,/'

"0"

:1 = 50 flm, 150 ppm is a minimum value to achieve a reasonable moditication • There are no indications that the Fe-rich intermetallics are affected by Sr additions. • Porosity volume fractions are mostly cooling rate dependent rather than moditication level related. • Highly modified samples with SDAS 10 flm offered the highest elongation to fracture values. • The strength of the modified alloys is a function of the overall coarseness of the microstructure. • AI-Si modification is not a guarantee for realizing higher ductility in EN AC 46000 alloys; although samples with SDAS of 25-50 flm were highly modified, the Fe-rich intermetallics govern the elongation to fracture.

14.

15.

16. 17.

18.

References 1.

Kaufman, J.G., Introduction to aluminum alloys and tempers2000: Asm IntI.

302

Caceres, C. and J. Taylor, Enhanced ductility in Al-SiCu-Mg foundry alloys with high Si content. Metallurgical and Materials Transactions B, 2006. 37(6): p. 897-903. Gowri, S. and F. Samuel, Effect of alloying elements on the solidification characteristics and microstructure of Al-Si-Cu-Mg-Fe 380 alloy. Metallurgical and Materials Transactions A, 1994.25(2): p. 437-448. Lu, S.Z. and A. Hellawell, The mechanism of silicon modification in aluminum-silicon alloys: impurity induced twinning. Metallurgical and Materials Transactions A, 1987. 18(10): p. 1721-1733. Ceschini, L., et aI., Effect of Fe content and microstructural features on the tensile and fatigue properties of the AISil OCu2 alloy. Materials & Design, 2011. Dinnis, C., et aI., The influence of strontium on porosity formation in Al-Si alloys. Metallurgical and Materials Transactions A, 2004. 35( 11): p. 3531-3541. Lu, L., et aI., Eutectic solidification and its role in casting porosity formation. JOM Journal of the Minerals, Metals and Materials Society, 2004. 56(11): p.52-58 . Djurdjevic, M., H. Jiang, and J. Sokolowski, On-line prediction of aluminum-silicon eutectic modification level using thermal analysis. Materials characterization, 2001. 46( 1): p. 31-38 . Handbook, A.S.M.M., Metals Park. Ohio: American Society for Metals, 1972. Thall, B.M. and B. Chalmers, Modification in Aluminium Silicon Alloys. Journal of the Institute of Metals, 1950. 77(1): p. 79-&. Davies, V.D.L., Direct Microscopic Observation of Solidification of Metals. Journal of the Institute of Metals, 1963. 92(4): p. 127-&. Zarif, M., B. Mckay, and P. Schumacher, Study of heterogeneous nucleation of eutectic Si in high-purity Al-Si alloys with Sr addition. Metallurgical and Materials Transactions A, 2011. 42(6): p. 1684-1691. Ceschini, L., et aI., Microstructure, tensile and fatigue properties of the Al-JO% Si-2% Cu alloy with different Fe and Mn content cast under controlled conditions. Journal of Materials Processing Technology, 2009. 209(15): p. 5669-5679. Seifeddine, S., S. Johansson, and 1.L. Svensson, The influence of cooling rate and manganese content on the fJ-AI< sub> 5 FeSi phase formation and mechanical properties of Al-Si-based alloys. Materials Science and Engineering: A, 2008. 490( 1): p. 385-390. Dinnis, C.M., et aI., The influence of strontium on porosity formation in Al-Si alloys. Metallurgical and Materials Transactions a-Physical Metallurgy and Materials Science, 2004. 35A(II): p. 3531-3541. Campbell, J., Castings2003: Butterworth-Heinemann. Shih, T.S., L.W. Huang, and Y.J. Chen, Relative porosity in aluminium and in aluminium alloys. International Journal of Cast Metals Research, 2005. 18(5): p. 301-308. Shabestari, S.G., et aI., Effect of Mn and Sr on intermetallics in Fe-rich eutectic Al-Si alloy. International Journal of Cast Metals Research, 2002. 15(1): p. 17-24.

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

MODIFICATION OF THE EUTECTIC MG2SI-PHASE OF ALMGSI-CAST ALLOYS Thomas Pabel l, Tose Petkov l, Christian l(neissl l, Peter Schumacher l,2 lAustrian Foundry Research Institute, Leoben, Austria 2 Chair of Casting Research, University of Leoben, Leoben, Austria Keywords: AlMgSi-alloys, moditication, eutectic Mg2Si-Phase

Abstract & Introduction

The present work shows a possible method for refining the microstructure of AlMgSi cast alloys in a similar fashion to the method which has been state of the art for many years in the case of hypoeutectic AlSi alloys, The effects of different melt purification processes, purification agents and the incorporation of different alloy and microalloying elements to modify the eutectic Mg 2 Si phase within near eutectic AlMgSi alloys are investigated, As a base material a near-eutectic alloy of type Al-Mg2 Si is investigated in the project as the base, Systematic studies lead to an optimum composition of the purging gases and the duration of the melt purification treatment Also the negative effects of other purification agents and trace elements on the formation of the eutectic phase were documented.

Figure 2. Deep etch of the non-modified Mg2 Si-Phase (left) and modified Mg2 Si-Phase (right) Mechanical Testing The results of the tensile test and the resistance to bending correlate well with the moditication found in the alloy and the results obtained by image analysis (see Figure 3).

Investigations and Results

Microstructure 1116.2

11.1

250

The effect of the modification upon the microstructure has been elucidated, in the first instance, by the determination of the most significant structural parameters, such as the form factor (FF), laminar separation (LA) and morphological factor of the microstructure (GF), using quantitative image analysis (see Figure 1) and deep etching methods (see Figure 2).

200 150 100 50

FF []

LA

GM []

P-Fl

Av [J]

Rm [MPa]

A [%]

Figure 3. Results of microstructure parameters and tensile tests (orange none moditied, blue moditied) Scanning electronic microscope In complementary research using scanning electronic microscope (SEM, see Figure 4) and transmission electronic microscope (TEM, see Figure 5), the effects of different melt treatment variants on the morphology of the Mg2 Si microstructure were analyzed in detail.

..

...

Figure 1. Microstructure of the non-modified Mg2Si-Phase (left) and modified Mg2Si-Phase (right)

303

Figure 4. Fracture surface (SEM) of the non-moditied Mg2SiPhase (left) and moditied Mg2Si-Phase (right)

Figure 5. TEM investigations of the non-moditied Mg2Si-Phase (left) and moditied Mg2Si-Phase (right) Summary All investigations have shown that the microstructure of Mg 2 Si changes its morphology with melting treatment, from a coarse laminar morphology to tine globular morphology. Owing to the spheroidization in the eutectic phase, ductility increases and offers new application fields and possible use for AIMgSi alloys in sand and permanent mould castings.

304

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013 THE INFLUENCE OF CASTING SPEED IN THE AS CAST STRIP MECHANICAL PROPERTIES OF 8079 AND 8006 ALLOYS

Dionysios Spathis and John Tsiros Metallurgy Laboratory - Technology Department Hellenic Aluminum Industry (EL VAL SA) Keyword: microhardness, mechanical properties, conductivity, casting speed Vicker's hardness measurements across the strip thickness of the alloy 8006. The error bars in Diagram 1 and Diagram 2 represents the standard deviation of the hardness measurements (St Div. = 1,5 Vicker hardness).

Abstract The influence of casting speed on mechanical properties from twin-rolled cast strip was investigated. Low and high speed coils of 8079 and 8006 alloys were produced from FATA HUNTER twin roll caster of ELV AL.

So

Microhardness and tensile tests of low casting speed coils for both alloys exhibit higher hardness and mechanical properties compared to the higher casting speed coils.

45

Electrical Conductivity measurements and metallographic examinations helped to explain the difference in mechanical properties between the high and low casting speed coils. Increased microstructural knowledge allowed EL VAL to optimize the cold rolling and annealing processes to meet customer specifications.

35

-tow spted coils 4>-high sp!Wd colis 30 3

'1

mm

Introduction

4

5

The twin roll casting process is a well-established process for 8079 and 8006 foil stock alloys. The casting speed is one of the major controlling parameters of the metallurgical quality of the caster product. Low and high casting speed coils of 8079 alloy and 8006 alloy were produced from FATA HUNTER twin roll caster of EL VAL SA. In order to fully characterize the metallurgical quality of the twin roll caster product, hardness and mechanical properties were determine on low and high speed as-cast strips from both alloys. 35

Further material characterization was done by strip grain structure study and Electrical conductivity measurements. A strong correlation of the as-cast strip mechanical properties and hardness with the caster speed was observed.

-!PI'! sp,ec!l (oils o high sp,ecd roils

4

7

mm

The higher manganese (Mn) and iron (Fe) containing 8006 alloy showed higher hardness relative to 8079. Reducing the casting speed for both alloys cause a displacement of the hardness curve to higher hardness values.

Vickers Microhardness Results Strip hardness was measured by Vicker's microhardness indentation method. Indentations were done across the strip thickness for low and high casting speed coils. The indentation load was constant at 100g and the indentation spacing was 0,5 mm, well above the recommended closest permitted spacing between adjacent indentations (3 times the Pyramid indentation diagonal distance) [1].

Tensile yield strength results of the as cast strip Tensi Ie test specimens parallel to the casting direction were prepared from the as cast strip. Diagram 3 and Diagram 4 show the yield strength of low and high casting speed coils for 8079 alloy and 8006 alloy. Both alloys, exhibits significant higher yield strengths when the casting speed is reduced.

Diagram 1 presents the Vickers hardness measurements across the strip thickness of the alloy 8079. Diagram 2 presents the

305

Diagram 5: EtC values of low and high speed coils • Alloy 8079 34.COO

It is obvious that the yield strength data confirms the micro hardness testing results. The yield strength difference between the two casting speeds for 8079 alloy is almost 13 N/mm2 and for 8006 alloy is almost 19 N/mm2.

33,800

.. coil 1 low speed 011 coil 2 low spped " Coil 3 lew speed .. coil 4 lew speed

Oiacram 3: : VS lIalue 01 low alld high casting spelld tolls I

co ()

/

/

Q)

0.6

(f)

•

1.2

8

(S)

Figure 6. a. Evolution of the slope of the regression lines in figure 5 versus the grain size in the thermal cup TGTh.Al . b: Evolution of the constant in the regression lines versus the solidification time in the thermal cup. Acknowledgments Acknowledgment are due to the Basque government for its tinancial support (Project ProFUTURE, Etorkek 20 I 0).

331

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

ALLOY ALSI30 CAST IN THE PROCESS OF RAPID SOLIDIFICATION AND CONSOLIDATED IN THE PROCESS OF PLASTIC FORMING Wojciech Szymanski l, Marcin Szymanek l, Janusz Zelechowski l, Mariusz Bigajl, Maciej Gawlikl, BarHomiej Plonka l linstitute of Non-Ferrous Metals Light Metals Division, Pilsudskiego 19, 32-050 Skawina, Poland Keywords: aluminum alloys, silicon addition, rapid solidification, consolidation, plastic working Abstract

AISi alloys with high Si content have good strength properties at elevated temperatures and resistance to thermal expansion. The main problem in the production of the classic methods of casting AISi alloys with Si content above 20% is the formation of large precipitates of primary silicon. To prevent the formation of large precipitates of silicon can be used casting methods of rapid solidification and subsequent consolidation in the process of plastic forming. This paper presents the results of structure alloy AISi30 cast in rapid solidification process on the wheel and consolidated in the process of extrusion and continuous rotary extrusion. The study was conducted on light microscopy and scanning electron microscope. It was found that the entire volume of the alloy is uniformly distributed fine Si particles.

The rapid solidification by melt spinning consists in pouring the molten metal onto a spinning copper wheel, providing very rapid heat transfer. The result is almost immediate solidification and metal leaves the wheel in the form of a thin ribbon (Fig. 1).

Introduction One of the methods to produce aluminium alloys of an uncommon composition and structure is by the combined process of casting with rapid solidification and the following plastic forming. [I] When modern advanced methods ofrapid cooling of the melt are used, the alloy structure solidifies as a powder in the atomiser or as ribbons when cast onto a rapidly rotating copper wheel. If optimum conditions for the process of casting and rapid consolidation are satisfied, it is possible to control some structure parameters like the size of the particles, the size of the precipitates, etc[2, 3]. Additionally, the production of aluminium alloys by rapid solidification allows introducing the alloying constituents that are incompatible with the state of equilibrium. The consolidation of material made by rapid solidification is achieved in one of the numerous variations of the plastic forming processes, among which the most commonly used are the direct extrusion and continuous rotary extrusion (CRE).[4-8] AISi alloys with high Si content are characterised by satisfactory mechanical properties at elevated temperatures and resistance to thermal expansion. The basic problem in the production of AISi alloys with Si content above 20% by the traditional casting route is the formation of large primary silicon precipitates. To prevent this adverse phenomenon, casting by rapid solidification methods followed by consolidation in a plastic forming process can be applied. This paper presents the results of examinations of the structure of an AISi30 alloy cast in the process of rapid solidification by melt spinning and consolidated in the processes of direct extrusion and continuous rotary extrusion.

Fig. 1 A ribbon of AISi30 alloy obtained by melt spinning This ribbon is next cut into small chips in a special mill (Fig. 2).

Fig.2 Chips made from a ribbon cut in a cutting mill

Methodology Tests were carried out on an AISi30 alloy of the composition given in Table 1 cast in the rapid solidification process by melt spinning.

Thus obtained material was next consolidated by direct extrusion and continuous rotary extrusion.

333

F or the direct extrusion process, the material was first subjected to cold pre-consolidation in a tube made of a 6xxx alloy, to be next hot extruded in a 60Tz vertical hydraulic press with instrumentation adapted to the extrusion process. The process was carried out at 460 0 C and the extrusion ratio was A. = 14, the diameter of the extruded rod was = Smm. The continuous rotary extrusion (CRE) was carried out in an MC-260 device. The die temperature was about 350°C. The produced rod had a diameter = 15mm. For microstructure examinations, an Olympus GX71 light microscope and a Philips XL30 scanning electron microscope with an attachment for the EDX chemical analysis in microregions were used. The size of the crystallites was measured by Sherrer method using Bruker DS Advance X-ray apparatus. Results Examinations have proved that, from the wheel side, the ribbon microstructure was characterised by the presence of a layer of the supersaturated AISi solution. Moving away from the wheel, the supersaturated solution underwent decomposition and silicon precipitates in the form of rosettes started appearing, and then coagulated into particles of I-311m size (Fig. 3).

* Fig.4 Microstructure of AISi30 alloy chips Examinations of the microstructure formed in rods obtained by direct extrusion have showed that the material consolidated by this method contained in prevailing part very tine precipitates of about I 11m diameter (Fig. 5). On the other hand, in the material consolidated by CRE, a significant amount of precipitates with a diameter of less than 0.5 11m was found (Fig.6). Tn both materials, a few areas appeared where the Si precipitates were of a size larger than 10 11m. SEM examinations showed that the material consolidated by direct extrusion contained precipitates of two types, i.e. silicon precipitates and precipitates containing Fe, Ni and Cu (Fig. 7). The alloy matrix was mainly composed of Al with minor additions of the alloying elements. On the other hand, in practice, the material consolidated by CRE contained only the precipitates of Si, while Fe, Ni and Cu were mainly found in the matrix, where they appeared as a fine network of the spots locally enriched with these elements (Fig. S).

Fig. 3 Microstructure of AISi30 alloy ribbon cast by melt spinning Similar structure was observed in the ribbons cut into chips (Fig. 4).

334

Fig. 5 Microstructure extrusion

alloy consolidated by direct

Fig.6 Microstructure of AlSi30 alloy consolidated by eRE

335

!IiI1

.... ru!

!:.f'$Wa"f··k¢'tI i.;$·l!i!:l(ls:;11

Q!

AI

'"

11ft

1'W

.....

1\'

Fig.S Microstructure of AlSi30 alloy after consolidation by eRE with chemical analysis of the precipitates.

Fig.7 Microstructure of AlSi30 alloy after consolidation by direct extrusion with chemical analysis of the precipitates

336

The study wasfinancedfrom a Strategic "ZAMAT" Project No. POJG. OJ. 03.0J-00-OJ5/09 entitled "Advanced materials and technologiesfor their production, " co:financedfrom the structural fund; the project implementation period is 20102013.

The results of the measurements of the size of the crystallites done by Sherrer method are shown in Table 2. Table 2 The results of the measurement of the size of crystallites Material Size of crystallites [flm1 Ribbons 1.233 Chips 1.346 Direct extrusion 6.273 CRE 0.808 Summary The method of melt spinning applied to produce an AISi30 alloy subjected next to the process of consolidation by direct extrusion or continuous rotary extrusion has yielded the material with fine Si precipitates of an average size of about I flm uniformly distributed in the matrix. The consolidation by CRE introduced only minor changes to the alloy microstructure compared with the consolidation by hot extrusion process. There was no formation of the precipitates of the phases containing Fe, Ni and Cu; only a network of areas enriched with these elements existed in the matrix. The crystallites did not grow in size, either. The reason was most probably the time of exposure to the effect of elevated temperatures, too short in the case of the material that did not have to be pre-heated for the CRE process, just opposite as it happened in the direct extrusion. References 1.

2.

3.

4.

5.

6.

7.

8.

L. Katgerman. F. Dom Rapidly solidified aluminium alloys by meltspinning Mat. Sc. Eng A375- 377 (2004) pp 12121216. Grant EPlE04060811 Development of Bulk Nanostructured Aluminium Alloys for High Strength Applications. (University of Oxford) 01. Juli 2007- 31 December 2010. B.Pucun "Microstructure and mechanical properties of a large billet of spray formed AIZnMgCu alloy with high Zn content Mat. Sc. Eng 2009. Frank Palm "Hypereutectic high Strength AIMgSc profile materials Melt- spun Scalmalloy- a new family of weldable and corrosion free Al alloys with 500- 850 MPa strength" Aeromat 2006 15-18 th of May 2006. Seattle. Washington. USA. A.J.Bosch. R.Senden. W.Etelmann. F.Palm. Nanostructured and High Strength Alloys: Scamalloy- a Unique High Strength and Corrosion Insensitive AIMgScZn Material Concept Aluminium Alloys Their Physical and Mechanical Properties Vol 1 Edited by Jiirgen Hirsch Wiley- VCH 2008. Feng Wang, Baiquing Xiong, Yongan Zhang, ... "Microstructure and mechanical properties of spraydeposited AI-Zn-Mg-Cu alloy" Materials dand Design 28 (2007) 1154-1158 Akihisa Inoue, Hisamichi Kimura ., Fabrication and mechanical properties of bulk amorphous, nanocrystaline, anaoquasicrystalline alloys in aluminium-based system" Journal of Light Metals 1 (2001) 31-41 P.M. Thomas Conform- the use of alternative feedstock materials APT Aluminium September 2004.

337

Thermal Mechanical Processing SESSION CHAIR

Xiyu Wen University of Kentucky Lexington, KY USA

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013 EFFECT OF HOMOGENIZATION TREATMENT CONDITIONS ON THE RECRYSTALLIZATION BEHAVIOR OF AL-1.2MN ALUMINUM ALLOY SHEETS Pizhi Zhao, Xinglin Chen, Wei Chen, Yonghao Zhang

Suzhou Research Institute for Nonferrous Metals, Suzhou, Jiangsu 215026, China

Keywords: AI-I.2Mn Aluminum alloy, recrystallization, homogenization annealing - foil rolling. It is necessary to control the grain

Abstmct

structures during intermediate annealing process because the structure greatly affects the surface topography offoil.

Aluminum foil is commonly used as a cathode current collector within lithium ion battery. To reduce thickness of the foil, one kind of high strength Al-1.2Mn alloy has been developed. In its

It is supposed that grain structure during intermediate annealing

fabrication

after

process varies from different homogenization conditions of

intermediate annealing greatly affects the surface topography of

AI-Mn series aluminum alloy due to different Mn content in

process,

the

recrystallization

structure

foil. Tn this study, the relationship between recrystallization

solid solution. It is known that precipitates which formed prior

structure during intermediate annealing after rolling and

to or during annealing have significant influence on nucleation

homogenization treatment conditions of Al-1.2Mn alloys has

and growth of the recrystallized grains, i.e. Zener pinning effect

been investigated. Homogenization was carried out with one

[1]. Several works have been carried out to investigate the

step treatment or two steps treatment. It was found that one step

relationship between precipitates and recrystallization behavior

homogenization

higher

in AI-Mn alloys [2-8]. It is well described that recrystallization

recrystallization temperature and fmer complete recrystallization

is prevented by the precipitation at low annealing temperature

grains

treated

while it can complete before precipitation occurs at high

cold-rolling sheets. This phenomenon can be explained by the

temperature. Most of the former works were performed only

relationship between Mn precipitation and recrystallization

with a single step homogenization to obtain supersaturated alloy.

behavior during intermediate annealing.

However, the precipitation of Mn-bearing precipitates can be

compared

treated

to

cold-rolling

two

steps

sheets

had

homogenization

quite different according to different homogenization conditions,

Introduction

which could lead to different effects on recrystallization behavior.

In lithium ion secondary batteries, AA1070 & AA1085

aluminum alloy foils are widely used as cathode current

The objective of this study is to investigate the influence of

collectors which act as electrical conductor as well as carrier of

homogenization conditions on the recrystallization behavior of

chemical substances such as LiFeP0 4 • For a battery with

AI-1.2Mn alloy sheets during intermediate annealing process in

constant volume, the power capacity can be raised by increasing

detail.

the amount of chemical substance through reducing the

recrystallization behavior will be discussed.

The

correlation

between

Mn

precipitating

and

thickness of aluminum alloy foil. Nevertheless, the thin foil may

Experimental Procedures

be broken during coating process if its strength is not enough. So, it is very important to develop aluminum alloy foil with high strength and good electrical conductivity.

Preparation of Materials

Addition of Mn is an effective method for obtaining high

The chemical composition of the AI-Mn alloy used in this study

strength foil. This foil is fabricated generally by casting -

is listed in Table 1. Ingots with gauge of 70 mm were made by

homogenization - hot rolling - cold rolling - intermediate

direct chill casting (DC casting).

341

Table 1. Chemical composition of the alloy used in the present

resistivity by equation [9]:

study (mass %). 1.D.

Mn

Fe

Si

Ti

Al

AI-I.2Mn

1.20

0.21

0.22

0.01

Bal.

p=2.7+2.94Mnss%+0.34Mnpre%+2.56Fess%+ 0.058Fepre%+ 1.02Siss%+0.088Sipre%

(1 )

The ingot was homogenized by means of one step or two steps

Where Mnss %, Fe ss % and Siss % are the content ofMn, Fe and Si

homogenization process, as shown in Figure I.

in solid solution, respectively. Mnpre %, Fe pre% and Sipre% are the content of Mn, Fe and Si precipitated from solid solution, respectively. Tn this study, electrical resistivity Pailo, was

620"(' Sit

calculated from electrical conductivity by equation: P alloy(lln'cm) = lOO/K (MS/m) Where

K

(2)

means electrical conductivity which was measured by

Sigma Test eddy conductivity apparatus after buff grinding. As shown in equation (I), the influence of precipitated Fe and Si to

211

on electrical resistivity is very weak. Moreover, most of the Fe

30

Homogenizillion lime, II

formed as intermetallic compounds during solidification, so the

Figure I. Details of homogenization processes used in the

content of Fe in solid solution is very low. Si content in the

present study

present alloy is very low, too. The influence of Fe and Si contents change in solid solution on electrical resistivity during

After homogenization, the ingots were scalped to gauge of 50 mm, reheated to 440 DC and hot-rolled to 6 mm gauge,

annealing could be ignored. Thus the change of electrical

furthermore cold-rolled to 0.5 mm gauge sheets. Two different

resistivity during intermediate annealing can be used to estimate

intermediate annealing processes were adopted after cold-rolling:

the change of Mn content in solid solution by the following equation:

batch annealing and salt bath annealing. The former was performed at 320°C and 380 °C respectively for 4 hours in a

(3)

batch furnace with heating rate of 30 °C/h and followed by air cooling, the latter was performed at 400°C for 30 seconds in a

Where Pbefore and Parter are the electrical resistivity before and

salt bath with heating rate of about 100 °C/s and followed by

after intermediate annealing. The change of Mn content in solid

water quenching.

solution /'., Mn ss % also represents precipitated amount of Mn-bearing particles.

Evaluation of Microstructures

Results and Discussion

The grain structure of longitudinal sections was observed by polarizing microscope after buff grinding and anodizing. After

Microstructures ofTngots after Homogenization

buff grinding and electrolyte polishing, intermetallic compounds were observed by optical microscopy, precipitates and grains

Figure 2 shows microstructures of the alloy after one step and

were examined using electron channeling contrast (ECC)

two steps homogenization, respectively. Large number of

technique in SEM.

AI 6 (Fe,Mn) and AldFe,Mn)3Si intermetallic compounds form during solidification. For one step homogenization ingot only

Measurement ofMn content in Solid Solution

few AlMnSi precipitates can be observed since the temperature is high enough (620°C) to dissolve large number of the

Mn content in solid solution was estimated from the electrical

342

precipitates

which

formed

during

the

early

stage

of

of 0.5-1

can be observed in two steps homogenization ingot.

homogenization, as seen in Figure 2(a). However, as shown in

This implies that AlMnSi phases precipitate and grow during

Figure 2(b), high density of large AlMnSi precipitates with size

cooling from 620°C to 440 °c and keeping period at 440°C.

Figure 2. Microstructures of the alloy after: (a) one step homogenization and (b) two steps homogenization homogenization sheets. As annealing temperature raises to 380

Grain Structures of Sheets after Intermediate Annealing

°C, both sheets have completed recrystallization structures, Grain structures of the 0.5 mm gauge sheets annealed at 320°C

grain size is smaller for one step sheets compared to two steps

and 380 °C for 4 hours are shown in Figure 3. As annealed at

sheets (Figure 3( c) and 3(d)). Figure 4 shows grain structures of

320°C, recrystallization does not happen in the one step

the sheets annealed at 400°C for 30 seconds in salt bath. both

homogenization sheets, but recrystallization grains are observed

one step and two steps homogenization sheets have the same

in two steps homogenization sheets (Figure 3(a) and 3(b)). This

fine

implies that the recrystallization temperature is higher for one

recrystallization behavior significantly.

step

homogenization

sheets

compared

to

two

grains.

This

means

that heating rate

steps

Figure 3. Grain structure after batch annealing: (a) one step homogenization + 320 °C/4h, (b) two steps homogenization + 320 °C/4h, (c) one step homogenization + 380 °C/4h, (d) two steps homogenization + 380 °C/4h

343

affects

the

Figure 4. Grain structure after salt bath annealing: (a) one step homogenization+400 °C/30s, (b) two steps homogenization+400 °C/30s Change ofMn in Solid Solution during Intermediate Annealing

than two steps homogenization sheets during batch intermediate

Table IT shows the variation of electrical resistivity of 0.5 mm

homogenization sheets has higher content of Mn in solid

gauge sheets with different intermediate annealing methods and

solution before annealing. Specially, one step homogenization

corresponding change of Mn contents in solid solution

sheets precipitates more Mn at 380°C than 320 °C. However,

L',Mn ss % which estimated according to equation (3). It is clear

when annealed by salt bath there is not much change in Mnss %

that the electrical resistivity of cold-rolled sheets decreases after

for both one step and two steps homogenization sheets. This

intermediate annealing. From the table, it is also clarified that

means that most of the Mn in solid solution does not have

more Mn was precipitated in one step homogenization sheets

enough time to be precipitated during salt bath.

annealing

(320°C

and

380°C)

because

one

step

Table IT. Electrical resistivity and L',Mn ss % after different intermediate annealing conditions

Electrical resistivity (J.lQ·cm)

L',Mn ss % (mass%)

Cold-rolling

320 °C/4h (batch)

380 °C/4h (batch)

400 °C/30s (salt bath)

One step

5.03

3.94

3.60

4.88

Two steps

4.13

3.52

3.56

4.05

One step

-

0.42

0.55

0.06

0.22

0.18

0.03

Two steps

Influence of Precipitates on Recrystallization Behaviors

occurs

(Figure 3(b)).

Microstructures of the alloy sheets with two kinds of

It is well established that larger non-deformable particles

homogenization conditions after intermediate annealing at 320

(especially those larger than I J.lm) can promote recrystallization,

°C are shown in Figure 5. Deformation structure is clearly

i.e. particle stimulated nucleation (PSN) [10]. In the present

observed after cold-rolling (CR) within both homogenization

study, PSN is assumed to be the dominant mechanism for

condition sheets, as shown in Figure 5(a) and 5(b). For one step

recrystallization nucleation since most of the intermetallic

homogenization sheets, only few precipitates can be observed

compounds are larger than 1 J.lm after cold-rolling. For sheets

compared to two steps homogenization sheets after cold-rolling.

with two steps homogenization, some of the precipitates grew to

After intermediate annealing at 320°C for 4 hours, many fine

large size and thus could act as nucleation sites for

precipitates formed on the deformation structure within one step

recrystallization. Furthermore, from Figure 5 and Table II it can

homogenized sheets. Those precipitates formed preferentially

be seen that less precipitates formed during intermediate

on subgrain boundaries and align along the rolling direction

annealing in two steps homogenization sheets than in one step

(RD), as shown in Figure 5(c). However, only very few

homogenization sheets. According to Nes [11], the precipitates

precipitates is observed within two steps homogenization sheets

with smaller size and higher density are more effective in

after intermediate annealing (Figure 5(d)) and recrystallization

inhibiting the movement of subgrain boundaries. Therefore the

344

Figure 5 Microstructure of the alloy after: (a) one step homogenization + CR, (b) two steps homogenization + CR, (c) one step homogenization + CR + 320 °C/4h, (d) two steps homogenization + CR + 320 °C/4h retarding from precipitates on recrystallization nucleation

intermediate annealing. One step homogenization sheets

should be much weaker in two steps homogenization sheets than

had higher recrystallization temperature and finer complete

in one step homogenization sheets. This can explain that why

recrystallization grain structure compared to two steps

one step homogenization sheets has a higher recrystallization

homogenization sheets.

temperature. Precipitates in the alloy inhibit not only nucleation

(2) This phenomenon can be explained by relationship

but also growth of the recrystallized grains [12]. However,

between Mn precipitation and recrystallization behavior

sheets annealed in salt bath show very similar grain structure

during intermediate annealing. One step homogenization

between one step and two steps homogenization sheets, which

sheets had more Mn content in solid solution after

indicating that the coarse precipitates formed during the second

cold-rolling and precipitated many fme particles during

step

homogenization

do

not

have

obvious

effect

intermediate annealing, which retarded recrystallization

on

behavior.

recrystallization. In one step homogenization sheets where more Mn-bearing precipitates formed during batch annealing, the growth of grains is effectively inhibited, which could result in

Refet'ences

relatively fmer grain structure. [I]

C.S.

Smith,

Interpretation

Conclusions

of

"Grains,

Phase,

Microstructure."

and

Interfaces:

Transactions

An

of the

Metallurgical Society ofAIME, 175 (1948), 15-51.

Influence of homogenization conditions on recrystallization

[2]

behavior of AI-1.2Mn alloy used for lithium ion battery was

behavior of a supersaturated AI-Mn alloy," Materials Letters, 64

investigated, the conclusions are as follows:

(20 I 0), 1829-1832.

(1) Homogenization

conditions

affect

[3]

recrystallization

behavior of the AI-I.2Mn alloy sheets significantly during

W.C. Liu and B. Radhakrishnan, "Recrystallization

1.T. Liu and 1.G. Morris, "Recrystallization textures of

continuous cast AA 3015 alloy: Development of the P

345

orientation

[011],"

Metallurgical

and

(2004),1229-1234.

Materials

Transactions A, 34 (2003), 2029-2032.

[8]

[4]

alloy," Scripta Materialia, 59 (2008), 611-614.

.LT. Liu, S.W. Banovic, R..T. Fields and J.G. Morris,

Y. Birol, "Recrystallization of a supersaturated AI-Mn

"Effect of intennediate heat treatment on microstructure and

[9]

texture evolution of continuous cast AI-Mn-Mg alloy sheet,"

Metallurgy, and Phase Diagrams (Ohio: American Society for

Metallurgical and Materials

Transactions A, 37 (2006),

K.R.

Van

Horn,

Aluminum:

Properties,

Physical

Metals, 1967), 50.

1887-1898.

[10] F.J. Humphreys and M. Hatherly, Recrystallization and

[5]

Related Annealing Phenomena (Oxford, UK: Elsevier, 2004).

M. Somerday and F.l Humphreys, "The Effect of

Dispersoids on the Recrystallization Behavior of a High-Purity

[II] E. Nes, "The effect of a fine particle dispersion on

AI-1.3 Mn Alloy," Materials Science Forum, 331-337 (2000),

heterogeneous recrystallization," Acta Metallurgica, 24 (1976),

703-714.

391-398.

[6]

[12] lE. Burke and D. Turnbull, Progress in Metal Physics

M. Somerday and F.J . Humphreys, "Recrystallisation

behaviour of supersaturated AI-Mn alloys Part I - AI-1.3

(London: Pergamon Press, 1952),220.

wt-%Mn," Materials Science and Technology, 19 (2003), 20-29. [7]

S. Tangen, H. Bjerkaas, T. Furu and E. Nes, "The Effects

of Dispersoids on the Recrystallization Behavior in a Cold Rolled AA3103-Aluminium Alloy," Materials Forum, 28

346

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

Toward a Recrystallized Microstructure in Extruded AA600SA Alloy A. Bahrami1,2, A.I. den Bakker3, A. Miroux 1,2, I. Sietsma2 innovation institute (M2i), Mekelweg 2,2628 CD Delft, the Netherlands 2 Department Materials Science and Engineering, Delft University of Technology, Mekelweg 2,2628 CD Delft, the Netherlands 3 Nedal Aluminium B.Y., Groenewoudsedijk 1,3528 BG Utrecht, the Netherlands 1 Materials

Keywords: Recrystallized microstructure, Extrusion, Al-Mg-Si alloys deformed structure. However, changing the microstructure from a fibrous deformed one to a fully recrystallized one necessities a precise control of chemical composition as well as processing parameters. Non-homogeneous distribution of grain size and partially recrystallized structures are not desirable. Recrystallization behavior during extrusion can be controlled by the alloy chemistry, homogenization treatment and extrusion processing parameters. There is limited available literature regarding the correlation between the alloy chemistry and extrusion parameters in one hand and the grain structure of AlMg-Si extrudate in the other hand. This study aims at understanding the role of chemical composition, homogenization treatment, billet temperature, and quench rate on the grain structure of the alloy AA6005A. The information presented in this paper could be useful to control and optimize the grain structure in other grades of Al-Mg-Si alloys.

Abstract

The correlation between extrusion processing parameters and chemical compositions on the one hand and the grain structure and mechanical properties on the other hand is studied for an extruded AA6005A alloy. Alloys with different chemical compositions and billet homogenization treatments have been extruded under different processing parameters. Optical microscope observations as well as hardness tests show that a wide range of mechanical properties and microstructures can be obtained. The grain structure can vary from a fibrous, deformed microstructure to a fully recrystallized microstructure. In some cases a mixture of the two is observed. The effects of processing parameters and chemical composition on the grain structure is discussed and used to find the optimized combination of homogenization treatment, pre-heating temperature, cooling rate after deformation, and chemical composition, in order to obtain a homogeneous recrystallized grain microstructure and optimum mechanical strength.

Materials and Methods

The chemical composition of the alloy AA6005A is shown in table 1. Two variants of the alloy AA6005A were DC-cast into ingots of length 300 mm and diameter 143 mm: one with Mn+Cr=O.13 (hereafter named 6005A_I) and the other one with Mn+Cr=O.4 (hereafter named 6005A_ll). Mn and Cr are dispersoid-forming alloying elements, which inhibit the recrystallization and grain growth. The ingots were extruded with the speed of 3 mmls to the profile shown in Figure 1. Two homogenization temperatures 530 and 570 DC were used to homogenize the as-cast ingots. Two pre-heating temperatures 480 and 520 DC were also chosen to pre-heat the ingots before extrusion. The ingots were either air- or water-quenched after extrusion, to study the effect of cooling rate after extrusion on the grain structure and mechanical properties of the alloy AA6005A. The grain structure of the extruded profiles was examined with a polarized optical microscope.

Introduction

Aluminum alloys are increasingly used in structural components in automotive and aerospace industries. They offer low weight, relatively high strength, corrosion resistance and good weldability [1-4]. Among different groups of aluminum alloys, the heattreatable AA6xxx series aluminum alloys are the most widely used alloys in various medium-strength structural components, including automobile body sheet, pipes, welded structures, and aircraft components. Extrusion of aluminum alloys is a relatively cheap and fast method of producing complex shapes from AI-MgSi alloys. The flexibility of the extrusion method with respect to the alloy chemistry and profile shapes makes a unique technique for a wide range of components, used in building structures, offshore industry, furniture, aerospace applications, and automotive applications [2]. Control of grain structure in extruded Al-Mg-Si alloys structures is of critical importance in so many applications. For example in aerospace applications, presence of peripheral coarse surface grain size can significantly decrease the fatigue resistance of the components. In most applications for AlMg-Si alloys, a fibrous deformed grain structure is more desirable. This can be achieved by increasing the amount of dispersoid-forming elements (mainly manganese and chromium) in the Al-Mg-Si alloys. However, in most cases the higher the Mg and Cr contents, the higher the extrusion pressure, which could eventually increase the risk of the formation of peripheral coarse grain structure. Coarse recrystallized surface layer in alloy AA6005A could results in orange peel effect during cold forming and poor anodized appearance for decorative applications [3]. So, for some applications it could be more beneficial to have a fully recrystallized microstructure in Al-Mg-Si alloys rather than a

Table 1. Chemical composition (wt%) of the alloy AA6005A 600SA

Elements Si Fe Cu Mg Mn Cr Zn Ti Others each

347

Min

Max

0.5

0.9 0.35 0.3 0.7 0.5 0.3 0.2 0.1 0.03

0.4

Figure 4 shows the effects of cooling rate after extrusion on the grain structure of the alloy AA600SA_1. It is seen that the alloy, quenched in air (Figure 4a), has slightly larger grains, compared to the water-quenched sample (Figure 4b). During slow quenching, the exposure time of the profile at elevated temperatures, where further grain boundary motion could takes place, increases. This could be the main reason for the observed difference between grain size in the water-quenched and aircooled samples. This effect is, however, minimal and obviously the grain structure of AA600SA alloy is relatively stable during elevated temperature exposure after extrusion. It is worth noting that the cooling rate after extrusion does not have any influence on the deformed fibrous structure of the alloy AA600SA_IT either.

Figure 1. The drawing of the die, which is used for the extrusion (dimensions are in mm) Results and Discussions

The grain structures observed after extrusion can be categorized into three categories: fibrous or deformed, fully recrystallized, and mixed recrystallized/deformed. Figure 2 shows examples of these three different microstructures.

Figure 4. Effects of cooling rate on the grain structure of the alloy AA600SA_T after a) air cooling and b) water quenching Figure S shows the effects of pre-heating temperature on the grain structure of the alloy AA600SA_Il. Obviously, the pre-heating temperature does not have any significant influence on the grain structure. This is also the case for the alloy AA600SA_1. Billet temperature influences the stored energy and can drive or inhibit recrystallization. High billet temperatures promote high exit temperatures, allowing grain boundaries to move but also correspond to lower stored energy. In the current case it seems that the difference between pre-heating temperatures 480 and S20 DC is not high enough to make a difference in the grain structure of extrudates.

Figure 2. Examples of a) deformed, b) recrystallized, and c) mixed recrystallized/deformed grain structures Figure 3 compares the microstructures of alloys AA600SA_I and AA600SA_IT,. The microstructure of AA600SA_T, having lower amounts of Mn and Cr, is either completely recrystallized or mixed recrystallized/deformed depending on the process parameters. The microstructure of AA600SA_IT, having higher amounts of Mn and Cr, is always fibrous. This can be rationalized in terms of the concentration of dispersoid forming elements, Mn and Cr. The dispersoids offer a resistance to grain boundary motion and can therefore inhibit the growth of grains during recrystallization. Decreasing the total Mn and Cr content reduces the volume fraction of particles and increases the inter-particle spacing, which decreases the grain boundary pinning effect, resulting in the easier recrystallization.

Figure S. Effects of pre-heating temperature a) 480 DC and b) S20 DC on the grain structure of the alloy AA600SA_II Figure 6 shows the effects of homogenization temperature on the grain structure of the alloy AA600SA_I. At homogenization temperature S30 DC, in most cases a mixed deformed/recrystallized structure is observed (see Figure 6a). However, by increasing the homogenization temperature to S70 DC, a more homogeneously recrystallized microstructure is obtained. This could be attributed to the effects of homogenization treatment on the dispersoid size distribution. Further investigation on the effects of temperature on the dispersoid size distribution is, however, needed to postulate

Figure 3. Effects of Mn+Cr contents on the grain structure in the a) alloy AA600SA_T and b) alloy AA600SA_IT

348

the effects of homogenization temperature on the structure of extrudates.

supersaturation for the material pre-heated at 520°C provides a higher potential for precipitate strengthening during ageing. 140

120

Z

!.

"

.g

l! .t:

,Il :Ii

1llI"Vmu,15° distribution, and an example of the

352

rotation axis distribution at rotation angle of 20±2.So, SO±2.So. and 60±2.So at strain of 67% is shown in Figure 4d. In the grains, the high angle misorientation (> ISO) number fraction increases with strain (Figure 4d). Meanwhile the grains decompose and the size of remain grains decreases with strain (Figures 4a-4c). The HABs in grains mainly locate in the misorientation range of IS-30°, almost no extra high angle boundaries (EHABs) >30° exist in grains. The rotation axis distribution at 20 ° shows that the rotation axes have no preferred clustering (Figure 4d). Since the brass {OIl ) and the copper (112} are the two important orientations in deformed aluminum [17], investigation behavior of the II ND gmins and 15°) distribution and the rotation axis distribution at 20±2.5°, 50±2.5°, and 60±2.5° at 67%. Other oriented grains in Figures Sa, 5b and 5c were excluded (black areas). EBSD step size: 2f.lm.

Figure 6: The highlighted IIND with a tolerance of 15° grains at three strains of (a) 40%; (b) 51 %; (c) 67%, and (d) the corresponding correlated large misorientation angle (> 15°) distribution and the rotation axis distribution at 20±2.5°, 50±2.5°, and 60±2.5° at 67%. Other oriented grains in Figures 6a, 6b and 6c were excluded (black areas). EBSD step size: 2f.lm.

Figure 4: The highlighted IIND with a tolerance of 15° grains at three strains of (a) 40%; (b) 51 %; (c) 67%, and (d) the corresponding correlated large misorientation angle (>15°) distribution and the rotation axis distribution at 20±2.5°, 50±25°, and 60±2.5° at 67%. Other oriented grains in Figures 4a,4b, and 4c were excluded (black areas). EBSD step size: 2f.lm.

Discussion Grain subdivision mechanisms and heterogeneity Based on EBSD results (Figures 4-6), the HABs of IS-30° increase with strain which is consistent with the scaling hypothesis (average misorientation angle increases with increasing strain) for IDB (i.e. cell boundary) misorienation

353

grain and the grain. The thin solid black lines in (a) represent the boundaries with misorientation angle >5° EBSD step size: 0.21.tln.

distribution in deformed fcc alloys based on extensive TEM results [18,19]. The grains only exhibit the cell structure, but other oriented grains hold both cell and extended planar boundaries according to the TEM observations in the compressed aluminum [10]. The misorientation angle induced by the assumed dislocation mechanism may reach up to 15-30° [12]. Activation of multiple slip systems and grain interaction may partially stabilize the grains during channel die compression [18-22]. In addition only cell structure exists in grains [10]. This explains why the grains only show the HABs of 15-30°. The grain subdivision is inhomogeneous in texture point of view. The grains have almost no EHABs, while and grains hold EHABs of 30-60° (Figures 4d,5d,6d). The and in grains locate in the orientation space from a- to fcc aluminum. It is accepted that orientations tend to transform from brass via S to copper in aluminum during plane strain compression. The and grains experience grain subdivision as well, and can only transfer to other deformation texture variants with random orientations. Thus, different deformation texture variants evolve in the original grain and impinge each other where EHAB occurs. The orientation dependence of EHAB of 30-60° distribution could be one reason why the misorientation scaling hypothesis showed a visible deviation when all kinds of dislocation boundaries are considered [12,19]. Thus, the EHAB of 30-60° preferentially existing in and grains with preferred rotation axes (Figures 5d,6d) is attributed to the multiple subdivided deformation texture variants via grain subdivision. Figure 7 shows the substructure and two lines point-point misorientation profiles in AAllOO at strain 40%. Grain subdivision occurrence was confirmed by the substructure and many> 15° subboundaries (Figures 7a, 7b). The substructure and subboundary are heterogeneous in different oriented grains. For example, it seems that only cell structure can be seen in grains, but cell boundaries and MBs coexist in grain; while some other oriented grains hold less substructures (Figure 7a). Moreover, grain subdivision exhibits intragranular heterogeneous. For example, the Line B area (grain boundary area) has dense substructures, but few substructures appear in the interior of the grain (Figure 7a). Figure 7b clearly confirms that HABs and EHABs, indeed are mainly created by grain subdivision [23]. The reason of grain subdivision heterogeneity could be attributed to the combination effect of grain orientation and local grain interaction with its neighboring grains [4,24,25]. (b) ••••••••••••••• LineA

Line B

30 40 50 Distance (micmn)

Grain subdivision effect on texture Due to grain subdivision, texture evolves accordingly. At strain of 40%, large amount of cube has transferred to other orientations including brass, copper and Goss (Figures 3a,3b). The quick decrease of cube texture may originate from three reasons: 1) the initial big grain size quickens up the grain subdivision [7]; 2) high density low angle GBs among cube-oriented grains favors the subdivision of cube texture (Figure 1a); and 3) the most favorable slip system may dominate plasticity at early stage «40%) and result in quick rotation away from cube. Goss is a transition component between cube and brass, thus it is also enhanced at strain of 40% (Figure 3b). As strain increasing up to 51%, however the brass and Goss disappear, while copper slightly increases and cube slowly decreases (Figures 3b,3c). Strikingly, the brass and Goss strengthen again at 67% (Figure 3d). Hence, a fluctuation in the a-fiber texture is observed (Figure 3). Subdivision continues in the deformation texture components. A single brass crystal was proven to be stable under channel die compression because the two slip systems have the same maximum Schmid factor, and simultaneously activate and cancel out the brass crystal rotation [26,27]. However, the active two slips lead to a strong shear strain tRD-TD which renders brass unstable [14,26-28]. Thus in polycrystal case, only the two slips activation in brass are actually impossible. Under plane strain compression, if brass subdivides into brass 1 and brass 2, the signs of the shear strain tRD_TD of brass 1 and brass 2 are opposite and the total TRD_TD is minimized. Similar to brass, copper would subdivide into copper 1 and copper 2 to eliminate the strong shear strain tRD-ND due to the four slips activation in the copper grain [28-30]. Since both brass 1 and brass 2 belong to brass texture, brass still has a strong intensity at 40% (Figure 3b). However, with strain increasing, brassl and brass 2 continue to subdivide. During the breakup of the brass-oriented grain into new non-brass grain (e.g. S texture variants) [27], brass texture quickly weakens due to the grain subdivision (Figure 3c). Copper still slowly increases due to the transformation from S to copper; meanwhile grain subdivision proceeds in copper, thus slows down the rate of copper increasing. The origin of the re-enhancement of the a-fiber with a wide spread at 67% may result from further deformation that causes the subdivided grains to re-aggregate along the a-fiber. The a-fiber texture fluctuation at strains 40%,51% and 67% may presumably result from the cycle of grain subdivision with strain: grain subdivision texture weakness dislocation accumulation - texture enhancement - grain subdivision. In fact, the volume fraction of brass and Goss measured by X-ray diffraction in a rolled aluminum alloy was observed markedly fluctuant in the reduction from 20% to 75% [5]. Only grain subdivision can rationalize this a-fiber texture fluctuation. This texture fluctuation would subside when the grain size reduces down to the critical grain size at high strains where grain subdivision may no longer proceed [7]. The slow decrease trend of cube from 40% to 67% (Figures 3b-3d) confirms that cube indeed has partial stability during compression [20-22, 31]. Grain subdivision effect on extra high angle boundary (EHAB) Because grain subdivision continues during deformation, the deformation texture components evolve to different variants. These texture variants in Bunge Euler angle {lPb 15° boundary within the

354

copper 2 (270,35,135}, S 1 (60,32,65}, S 2 {300, l48,245}, S 3 (120,148,245), and S 4 (240,32,65). The results show that most of misorientation angle locates in 30-60° and the rotation axis preferentially clusters at and .

(270,3S,13S}, S I (60,32,6S}, S 2 {300,148,24S}, S 3 (120,148,24S}, S 4 {240,32,6S}. In a fine EBSD scanning map at 67% (Figure 8a), these different deformation texture variants and the cube in fcc aluminum are colored by their specific orientations (Euler angles) with a tolerance of 20°. Substructures with other orientations (no deformation texture variants and the cube) shown in white are considered as random component (Figure 8a). The orientation map at 67% (Figure 8a) substantially confirms that different deformation texture variants are created by grain subdivision. These subdivided deformation texture variants interweave each other inside a grain (Figure 8a). Moreover, the random texture colored in white in Figure 8a appears to be significant as a result of grain subdivision. Orientation randomization by grain subdivision leads to the lower deformation texture intensity.

Conclusions An annealed cube-textured commercial AA 1100 aluminum alloy was conducted channel die compression to strain of 40%, SI % and 67%, respectively. The microstructure and texture of the AA1100 samples were characterized by EBSD. The misorientation angle and the rotation axis among deformation texture variants were also calculated. Some conclusions were drawn based on experimental and calculation as follows. 1. Grain subdivision occurred in the annealed AA 11 00 aluminum alloy in the strain range of 40%-67%. Grain refinement was carried out by grain subdivision. Both stable and unstable oriented grains experienced grain subdivision during deformation. Grain subdivision proceeded in a heterogeneous manner due to grain orientation and grain interaction. The HABs of IS-30° and the EHABs of 30-60° at medium strains were mainly created by grain subdivision.

Theoretically, there are 36 different neighboring pairs among the deformation texture variants: Goss, brass I, brass 2, copper I, copper 2, S I, S 2, S 3, and S 4. The misorientation and the corresponding rotation axis of the 36 neighboring pairs are calculated by MTEX [16] and plotted in Figure 8b. Note that the misorientation angle (i.e. the disorientation angle) is the smallest misorientation angle among the 24 variants in fcc aluminum. A large number fraction of misorientation angle in 30-60° can be seen, in addition, the corresponding rotation axes preferentially clusters at and (Figure 8b). The calculated misorientation and rotation axis distribution are consistent with the measured EHAB misorientation and the corresponding rotation axis distribution in and grains (Figures Sd,6d,8b). The number fraction of these texture variants increases with strain due to further grain subdivision. This is the reason why the fraction of EHAB of 30-60° misorientation increases with strain in and grains (Figures Sd,6d). The calculated and experimental results confirm that the EHABs indeed result from grain subdivision in the deformation texture components. According to calculation and experimental observations, the HABs of IS-30° mainly result from grain subdivision by dislocation mechanism [12], while the EHABs of 30-60° are mainly induced by grain subdivision proceeding in the deformation textures. It is conceivable that if the deformation mode changes, e.g. simple tension or torsion, then deformation textures will change accordingly. Thus, grain subdivision may proceed differently. Moreover, the effect of grain subdivision may be negligible after high strain deformation because there is little subdivision at high strains [7]. Therefore, typical deformation textures may normally develop and enhance after high strain deformation.

2. The HABs of IS-30° did not have rotation axes clustering in , and grains and the number fraction of high angle boundary of IS-30° increased with strain. It was mainly caused by grain subdivision through dislocation mechanism in , and grains. 3. The EHABs of 30-60° in and grains resulted from deformation texture variants that interweaved each other due to grain subdivision. Few EHABs were observed in the grains due to the partial stability of cube and only cell structures exist in the cube. 4. Grain subdivision smoothed deformation texture intensity, randomized orientations, and resulted in fluctuation in the a-fiber texture in AA 11 00 aluminum alloy at medium strains. Acknowledgements The authors are grateful to the financial support from the Department of Energy, Contract No. DE-FC-26-06NT427SS, and the Center for Advanced Vehicular Systems (CAVS) at Mississippi State University. The authors are also grateful to Robert Malley and Stephen Horstemeyer for their assistance with the experimental works at CA VS. References 1.

2. Figure 8: (a) Subdivided texture variants interweave each other during grain subdivision at 67%. The color rectangles are the orientation legends that represent the specific texture components within 20° of ideal orientations. The solid black lines are the boundaries with a misorientation angle >5°. (b) The distribution of the misorientation angle and rotation axis of the 36 pairs of the ideal deformed texture variants in fcc aluminum. The texture variants include: Goss [0,45,0), brass I (35,45,0), brass 2 (325,45,0], copper I [90,35,45),

3.

4.

355

Q. Liu and N. Hansen, "Geometrically necessary boundaries and incidental dislocation boundaries formed during cold deformation", Scripta Metallurgica et Materialia, 32( 1995) 1289-129S. Q. Liu, et aI., "Heterogeneous microstructures and microtextures in cube oriented Al crystals after channel die compression", Metallurgical and Materials Transactions A, 29( 1998) 2333-2344. G. Winther, X. Huang and N. Hansen, "Crystallographic and macroscopic orientation of planar dislocation boundaries-Correlation with grain orientation", Acta Materialia, 48(2000) 2187-2198. N. Hansen, X. Huang and G. Winther, "Effect of grain

5.

6. 7.

8.

9.

10. II.

12. 13. 14.

IS. 16.

17. 18. 19.

20.

21. 22.

23. 24.

boundaries and grain orientation on structure and properties", Metall Mater Trans A, 42(2011) 613-625. P.l Hurley and F.l Humphreys, "The application of EBSD to the study of substructural development in a cold rolled single-phase aluminum alloy", Acta Materialia, 51 (2003) 1087-11 02. P.1. Hurley, P.S. Bate and F.1. Humphreys, "An objective study of substructural boundary alignment in aluminum", Acta Materialia, 51( 2003) 4737-4750. F.1. Humphreys et aI., "Developing stable fine-grain microstructures by large strain deformation", Philosophical. Transactions of the Royal Society A, 357(1999) 1663-1681. M.F. Horstemeyer and D.L. McDowell, "Modeling effects of dislocation substructure in polycrystal elastoviscoplasticity", Mechanics of Materials, 27(1998) 145-163. F.X. Lin, A. Godfrey and G. Winther, "Grain orientation dependence of extended planar dislocation boundaries in rolled aluminum", Scripta Materialia, 61(2009) 237-240. G.M. Le et aI., "Orientation dependence of the deformation microstructure in compressed aluminum", Scripta Materialia, 66(2012) 359-362. G. Winther et al.. "Critical comparison of dislocation boundary alignment studied by TEM and EBSD: technical issues and theoretical consequences", Acta Materialia. 52(2004) 4437-4446. D.A. Hughes and N. Hansen, "High angle boundaries formed by grain subdivision mechanisms", Acta Materialia, 45(1997) 3871-3886. P.L. Sun, P.w. Kao and c.P. Chang, "High angle boundaries formation by grain subdivision in equal channel angular extrusion", Scripta Materialia, 51(2004) 565-570. L. Delannay et aI., "Quantitative analysis of grain subdivision in cold rolled aluminum", Acta Materialia, 49(200 I) 2441-2451. S.G. Chowdhury, "Development of texture during cold rolling in AA5182 alloy", Scripta Materialia, 52(2005) 99105. R. Hielscher and H. Schaeben, "A novel pole figure inversion method: Specification of the MTEX algorithm", Journal ofApplied Crystallography, 41 (2008) 1024-1037. W. Mao, "Modeling of rolling texture in aluminum", Materials Science and Engineering A, 257(1998) 171-177. D.A. Hughes et aI., "Scaling of misorientation angle distributions", Physical Review Letters, 81(1998) 46644667. D.A. Hughes et aI., "Scaling of microstructural parameters: misorientations of deformation induced boundaries", Acta Materialia, 45( 1997) 105-112. T. Samajdar and R. Doherty, "Cube recrystallization texture in warm deformed aluminum: understanding and prediction", Acta Materialia, 46(1998) 3145-3158. F. Basson and lH. Driver, "Deformation banding mechanisms during plane strain compression of cubeoriented fcc crystals", Acta Materialia, 48(2000) 2101-. P. Mukhopadhyay and S. Badirujjaman, "Relative stability of cube orientation in single crystal aluminum during deformation", Trans Indian Inst Met, 65(2012) 343-353. D.A. Hughes and N. Hansen, "High angle boundaries and orientation distributions at large strains", Scripta Metallurgica et Materiala, 33(1995) 315-321. D. Raabe, Z. Zhao and W. Mao, "On the dependence of ingrain subdivision and deformation texture of aluminum on

25.

26.

27.

28.

29.

30.

31.

356

grain interaction", Acta Materialia, 50(2002) 4379-4394. W.C. Liu and P.P. Zhai. "Characterization of microstructures near grain boundary in hot deformed AA3104 aluminum alloy", Materials Characterization, 62(2011) 81-89. A. Albou, lH. Driver and C. Maurice, "Microband evolution during large plastic strains of stable (l10} Al and AI-Mn crystals", Acta Materialia, 58(2010) 30223034. A. Ray and B.1. Diak, "Grain interaction effect on the stability of the {1l0} "brass" orientation in an aluminum multi-crystal", Scripta Materialia, 62(201 0) 606609. lH. Driver, lH. Jensen and N. Hansen, "Large strain deformation structures in aluminum crystals with rolling texture orientations", Acta Metallurgica et Materialia , 42(1994) 3105-3114. A. Godfrey, J.H. Jensen and N. Hansen, "Slip pattern, microstructure an local crystallography in an aluminum single crystal of Copper orientation {112)", Acta Materialia, 46(1998) 835-848. P. Wagner, O. Engler and K. LUcke, "Formation of Cu-type shear bands and their influence on deformation and texture of rolled fcc (112 } single crystals", Acta Metallurgica et Materialia, 43(1995) 3799-3812. Q. Ma et aI., "Rapid texture measurement of cold-rolled aluminum sheet by X-ray diffraction", Scripta Materialia, 54(2006) 1901-1906.

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

FATIGUE ANALYSIS OF ULTRAFINE GRAINED AL 1050 ALLOY PRODUCED BY CYCLIC FORWARD BACKWARD EXTRUSION Hamid Alihosseini 1 and Mohsen Asle Zaeem 2 * Department of Materials Science and Engineering Missouri University of Science and Technology 1400 N. Bishop Avenue, Rolla, MO 65409-0340 *Email: zaeem(uhnst.edu Keywords: Fatigue, ultratine-grained material, cyclic forward-backward extrusion (CFBE), Al 1050

Abstract

2. Experimental Procedure

In this work, fatigue behavior of ultratine-grained (UFG) Al 1050 alloy produced by a cyclic forward-backward extrusion (CFBE) was studied. Initial average grain size of 120 J.lm was reduced to IJ.lm, 600nm and 320nm using 1, 2 and 3 cycles CFBE process, respectively. After three CFBE cycles, both yield strength and tensile strength increased about 3.5 and 3 times greater than those of as-received samples. Fatigue tests were carried out under loadcontrolled mode at a frequency of 15 Hz. Results indicate that grain refinement of Al 1050 samples improved the resistance to fatigue crack nucleation under predominantly high cycle fatigue loading. To explain the formation process of damage surfaces, microstructure changes in the damage surfaces caused by cyclic stresses were studied by scan electron microscopy (SEM).

2.1. Material and Cyclic Forward Backward Extrusion The material used in this study was Al 1050 commercial aluminum alloy. CFBE consists of forward-backward extrusion, followed by a constrained back-pressing, and completed with conventional extrusion to produce the UFG rods. These steps are performed using a twin punch setup. Cylindrical specimens with diameters of 30 mm and lengths of 50 mm were machined from as-received AA1050 billets. They were then annealed at 400°C for 2 hour and cooled down at furnace atmosphere, which resulted in an average grain size of about 120 J.lm. The CFBE process was carried out using a press with the cross-head speed of 5 mm/min at 25°C. The produced rods had diameter of 10 mm and length of 20 cm.

1. Introduction

2.2. Fatigue Tests

Ultrafine-grained (UFG) materials produced by severe plastic deformation (SPD) processes have received remarkable attentions in the recent years because of their unique properties that make them applicable in ditferent tields [1,2]. Among the SPD processes, equal channel angular pressing [3], high-pressure torsion [4] and accumulated roll bonding [5] are the most commonly used methods. One of the most effective SPD methods was recently introduced by Alihosseini and Asle Zaeem, which is a novel cyclic forward-backward extrusion (CFBE) method to produce UFG materials [6]. This process consists of forwardbackward extrusion, followed by constrained back-pressing and completed with a conventional extrusion. CFBE was used to produce UFG 1050 aluminum rods. Some of the mechanical properties of All050 rods produced by CFBE such as yield strength, tensile strength and microhardness have been studied [6, 7]. In addition to the mechanical properties, fatigue behavior of UFG materials is very important in the usage of these materials. Studies of the UFG materials produced by SPD methods have mainly focused on their microstructure evaluation, and only concerned with the static mechanical properties such as tensile strength and microhardness. Fatigue properties of UFG materials have not been studied profoundly, and only there are a few studies performed to investigate fatigue behavior of pure Al and Cu alloys [8,9]. The fatigue behavior of materials is often controlled by slip bands, grain or twin boundaries, second phases and voids. In the present work, we report the fatigue properties of an All 050 alloy subjected to SPD. By applying different cycles of CFBE, submicron Al alloy samples have been produced with different grain sizes. To study the effects of SPD and resulted microstructures on fatigue behavior of UFG All 050, we performed fatigue tests after different cycles CFBE.

The experimental fatigue tests were carried out on dog-bone shaped specimens. The fatigue test specimen dimensions are shown in Figure 1. The fatigue test samples were cut from the uniformly deformed region of the produced rods by electro discharge machining. After this step the specimens were polished down mechanically to 6 J.lm grid sizes in order to remove the electro discharge machining etfects. All fatigue tests were carried out on a servo hydraulic test. Fatigue tests were done under loadcontrolled mode at a frequency of 15 Hz at load levels of 0.8 times of the yield strength of the as received sample. 6

Figure 1. Samples dimensions for fatigue test specimens (mm). After CFBE, microstructural changes and grain sizes were evaluated and the associated mechanical properties were measured. For detailed investigation of the samples after fatigue test, SEM (JEOL JSM 6490L) was used. Details of fatigue fracture surfaces in ditferent regions were analyzed.

3. Results and Discussion 3.1. Mechanical Properties and Microstructure Evaluation Table I shows the mechanical properties and average grain size of the as-received and CFBE processed aluminum specimens. For

357

processed aluminum, due to the fact that the tensile strength of CFBE processed All050 is about 3 times greater than as received one. For the stress beyond fatigue limit stress, number of cycles to failure (fatigue life) of CFBE processed specimens was larger than as received ones. As it can be seen in Figure 3, by increasing the stress amplitude there is no remarkable difference for fatigue life of processed samples with different CFBE cycles.

the same initial average grain size of 120 !lm, the specimen subjected to 1 cycle of CFBE reached an average grain size of about 980 nm, UFG structures of about 600 nm in size is achieved after 2 cycles of CFBE, and an average grain size about 315 nm is observed after 3 cycles of CFBE [7]. TEM micrographs of samples are shown in figure 2, which show as the number of CFBE cycles increases, the size of grains decreases. The details of the results have been reported in these references [6,7]. Table 1. Mechanical properties of CFBEed Aluminum specimens [6,7]. Yield Tensile Average Number Vickers strength Strength of cycles Grain size Hardness (MPa) (MPa) 0 120 !lm 50 70 30 I 160 980nm 105 55 2 600nm 120 195 62 315 nm 160 215 3 68

(,

.As IlAl'rfrivmi i (",Ie _ 1 Ti7AlsSi12-->Ti(Al,Si)3. Furthermore, this experiment result reveals two two-phase equilibriums. One is the equilibrium between (AI) (its average composition is AI 992 Sios), and Ti7AIsSi12 (its average composition is Ti327AI136Sis37)' The other

392

Conclusions

is between Ti7AIsSi12 and Ti(AI,Si)3 (its average composition is Ti2s6A1623Si12.1 ).

• •

1M Crack Surface ofAI"10\\'1.% Si BulK the Crack Surface of Ti Bulk

•+

••

•• •

*

•

Through AI-lOwt.%Si diffusion couple experiment, the phase equilibriums in the AI-rich corner of the ternary AI-Si-Ti system at 500°C have been investigated. The experiment results can be concluded as follows: 1. The confirmed reaction path way in the diffusion zoon of the sample is (AI) ----> Ti7AI sSi12 ----> Ti(AI,Si)3 . 2. The diffusion couple reveals two two-phase equilibriums existed in the AI-rich corner of the AI-Si-Ti system at 500°C, which are (AI) + Ti7AIsSi12 and Ti7AIsSi12 + Ti(AI,Si)3' This result is consistent with the corresponding calculated isothermal section of the AI-Si-Ti system.

Ti7"!S$i12 Till!;;

TiS'2 AI Si

•

Acknowledgments

+ +

., m

The authors thank Instrumental Analysis and Research Center of Shanghai University for their support of materials testing and research. This work was financially supported by the College Students' Innovative Experiment Project of Shanghai University (CXXJ-Il-060).

+ 00 Diffraction Angle

00

00

References

Figure 3. XRD pattern of crack surfaces of two end-members.

1. N. Saheb et aI., "Influence of Ti addition on wear properties of AI-Si eutectic alloys", Wear, 249 (2001), 656-662.

Fig. 4 is the calculated isothermal section of AI-Si-Ti system at 500°C. Two end points of the red solid line crossing the isothermal section denotes two end-members' compositions. The calculated ternary diagram and related thermodynamics database are extrapolated from binary systems of AI_Si[4 l, AI_Ti[Sl, Ti_Si[6l. The adopted ternary phases in this calculation are Tl (Ti 7AI sSi I4 ) and T2 (Ti 3AI 2Si s), which are reported by J. Grabner, et al.[7l. According to this isothermal section, there are two three-phase equilibriums, including (AI) + (Si) + Ti7AIsSil4 and (AI) + Ti7AI sSil4 + Ti(AI,Si)3, and three two-phase equilibriums, including (AI) + Ti(AI,Si)3, (AI) + Ti7AI sSil4 and Ti7AI sSil4 + Ti(AI,Si)3, in the AI-rich corner. Thus, it can be found that the equilibrium relationships identified by this experiment are consistent with the related equilibrium information from the calculated isothermal section.

2. T. Gao et aI., "Influence of Si and Ti contents on the microstructure, microhardness and performance of TiAISi intermetallics in AI-Si-Ti alloys", Journal of Alloys and Compounds, 509 (2011),8013-8017. 3. M. Zeren and E. Karakulak. "Influence of Ti addition on the microstructure and hardness properties of near-eutectic AI-Si alloys", Journal of Alloys and Compounds, 450 (2008), 255-259. 4. L. Kaufman, "Coupled phase diagrams and thermochemical data for transition metal binary systems-VI", CALPHAD, 3 (1) (1979), 45-76. 5. F. Zhang and Y.A. Chang, "A thermodynamic description of the Ti-AI system", Intermetallics, 5 (1997), 471--482. 6. H.J. Seifert et aI., "Thermodynamic Optimization of the Ti-Si System", Z. Metallkd, 87 (1) (1996), 2-13. 7. J. Grabner, D. Mirkovi'c and S.F. Rainer, "Thermodynamic aspects of grain refinement of AI-Si alloys using Ti and B". Materials Science and Engineering, 395 (A) (2005), 10--21.

40

50

EO

Si Figure 4. Calculated isothermal section of AI-Si-Ti at 500°C

393

Emerging Technology SESSION CHAIR

Subodh Phi nix LLC Lexington, KY USA

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013 Transient microstructural thermomechanical fatigue and deformation characteristics under superimposed mechanical and thermal loading, in AlSi based automotive diesel pistons. 1 Federal-Mogul

Roman Morgenstern l, Scott Kenningleyl Niirnberg GmbH, Nopitschstrasse 67,90441 Niirnberg, Germany

Keywords: Aluminium, Thermomechanical, FEA, Superimposed, Fatigue (TMF) superimposed with high cycle mechanical fatigue (HCMF) loading or TMF/ HCMF has been previously reported [I, S, 12, 13] for similar AlSi based alloys to that used in this study. Detailed modelling of the a-All Si interface has been carried out at an atomistic level [14, IS] and on a microstructural level using FEA [16, 17]. This study provides an evaluation of the micromechanical fatigue damage observed at the primary a-AI! Si matrix interface during semi in-situ isothermal HCMF and superimposed TMF/ HCMF bench testing. The macrostructural fatigue strength gained for IE7 high cycles for superimposed TMF/ HCMF and isothermal HCMF loading are compared. Explanations of deformation under thermal loading, damage initiation mechanisms, and local stress discontinuity effects are supported with a 2D FEA of the experimental region of interest (ROT) under TMF loading.

Abstract Presently AlSi based alloys, consisting up to 12 element systems, are used in the manufacture of automotive pistons for light vehicle (L VD) and heavy (HD) duty diesel engines. The pistons combustion wall is subject to complex superimposed transient mechanical and thermal loading with peak operating temperature representing a homologous temperature range of 0.8-0.9 Thom . Using specialist superimposed thermomechanical bench test apparatus, 'engine like' TMF loading has been reproduced and a number of semi in-situ experiments have been carried out to evaluate key microstructure damage mechanisms. The evolution of microstructural damage at the interface between hard Si inclusions and the softer Al matrix has been documented using scanning electron microscopy. The deformation characteristics at the a-All Si interface have been recreated using FEA techniques incorporating non-linear elasto-viscoplastic properties for the matrix material. Comparisons of bench test fatigue lives for transient superimposed high frequency and microstructural TMF loading, with fatigue lives from isothermal mechanical loading are also made.

Experimental and Material Details Materials The AISi based alloy used for evaluation in the present study is the Federal-Mogul high durability piston material FM-B2, which primarily consists of Si: 12.0-14.Swt%, Cu: 3.7-S.2wt%, Ni: 1.73.2wt%, Mg: O.S-1.5wt%, Fe:

E

nata

"'

00 41l

I.!.

20

Test temperature

Lifetime characteristics of HCMF and TMFI HCMF loading The comparison of stress based fatigue strength at lE7 equivalent high cycles is graphically shown in Figure 7 for isothermal HCMF testing at 350°C, 440°C and for the TMFI HCMF testing. The result shows the superimposed TMFI HCMF loading (consisting of an unconstrained thermal cycle) has an 8% higher and 6% lower mean I E7 high cycle fatigue strength than the isothermal HCMP results for 440°C and 350°C respectively. Damage Mechanisms During fully reversed cyclic fatigue loading, A1Si based piston alloys spend the majority of their total fatigue life in the micro crack incubation stage. The evolution from microcracks to microstructurally small cracks to physically small and then finally to long fatigue cracks occur at a high rate. At high temperatures the long fatigue cracks tend to run along brittle primary alloy phase boundaries with shear across the intervening regions of ductile matrix to final fracture. At lower temperatures the long fatigue crack incident on primary Si can often induce particle fracture. The initiation sites for microcracks are preferentially at stress concentrations. For A1Si based alloys, the stress discontinuities are provided by primary silicon, eutectic Si, intermetallic phases, casting defects (pores and oxides). The presence of pores and oxides in highly stressed regions are generally accepted as being most detrimental to fatigue, reducing lifetimes by in excess of an

micromechauicaJ at the AU Si interfaee taken from dose to n frachlre surface of n TMFI HCMF test with failure

401

the cubic shape of the primary Si particles and by the re-entrant angle created between the two Si particles, as shown in Figure 2f and Figure 3. Regions of interfacial debonding and void growth are essentially stress raising discontinuities acting as initiation sites for microcracks. The propagation of these microcracks is sensitive to the local micro-geometry with not all the cracks necessarily leading to long fatigue cracks [6]. Within the microstructure it is expected that there are many microcrack initiation events occurring at highly stressed locations. As these microcracks propagate, they may coalesce or link to form firstly a microstructurally or physically small crack which then goes onto to evolve into a long fatigue crack eventually leading to final fracture. As microstructurally small and long cracks propagate through the interdendritic a-AI matrix regions, they come into contact with Si particles and intermetallics which generally have the effect of arresting the crack growth rate and deflecting the cracks around them incurring debonding along their interfaces. If the stress concentration is high enough at the interface, Si particle fracture is possible affecting long and microcrack growth characteristics. During thermal cycling, it is more likely that Si particle fracture would occur in the low temperature regime.

2.

3.

4.

5.

6.

7.

Conclusions 8.

It is generally well accepted that in the absence of intrinsic casting flaws the fatigue of AlSi based piston alloys can be dominated by Si particle morphology with initiation more prevalent along the interfaces of larger primary Si cuboids in regions of favorably stressed micro-geometry. The local heterogeneity in highly alloyed AISi microstructures makes it difficult to quantify the effect of any single microstructural feature on fatigue life. In this study micromechanical damage under 'engine like' superimposed TMFI HCMF loading has been identified and discussed with respect to a local microstructural ROT. The semi in-situ tests showed how superimposed TMFI HCMF loading can increase the micromechanical fatigue damage in comparison to the isothermal HCMF loading. This is primarily a function of the repetitive thermal crE tensor induced during heating and cooling. This thermal crE tensor is induced due to the mismatch in thermal expansion between the Si phases and a-AI matrix in the alloy microstructure. The FEA model predictions under thermal cycling conditions showed the heterogeneous state of stress around two cubic clustered primary Si particles. The locations of the high load discontinuities show some correlation to the regions of increased interfacial damage. The TMFI HCMF loading, featuring an unconstrained triangular thermal cycle running between Tmin 200°C and T max 440°C, provided mean fatigue strength results between the values gained for isothermal HCMF test results at 350°C and 440°C.

9.

10.

11.

12.

13.

14.

Acknowledgements 15.

With thanks to the Federal-Mogul Technology team, especially to Robert Willard, Neil Wardrop and Georg Hopp. References 1. Kenningley, S. and R. Morgenstern, Thermal and Mechanical Loading in the Combustion Bowl Region of Light Vehicle Diesel AISiCuNiMg Pistons; Reviewed with Emphasis on Advanced Finite Element Analysis and Instrumented Engine Testing Techniques. SAE Technical Paper 2012-01-1330,2012.

16.

402

Reichstein. S., P. Konrad, and S. Kenningley, Microstructure modification ofpiston materials for high stress and temperature conditions. Aachener Kolloquium Fahrzeug- und Motorentechnik 2007, 2007. Reichstein, S., et aI., High-Performance Cast Aluminum Pistons for Highly Efficient Diesel Engines. SAE Technical Paper 2007-01-1438,2007. Barnes, S. and K. Lades, The Evolution of Aluminium Based Piston Alloys for Direct Injection Diesel Engines. SAE Technical Paper 2002-01-0493,2002. Humbertjean, A. and T. Beck, Effect of the casting process on microstructure and lifetime of the Al-pistonalloy AISi12Cu4Ni3 under thermo-mechanical fatigue with superimposed high-cycle fatigue loading. International Journal of Fatigue, 2011. McDowell, D.L., et aI., Microstructure-based fatigue modeling of cast A356-T6 alloy. Engineering Fracture Mechanics, 2003. 70(1): p. 49-80. Jordon, J.B., et aI., Microstructural Inclusion Influence on Fatigue of a Cast A356 Aluminum Alloy. Metallurgical and Materials Transactions a-Physical Metallurgy and Materials Science, 2010. 41A(2): p. 356-363. Wang, Q.G., D. Apelian, and D.A. Lados, Fatigue behavior of A356-T6 aluminum cast alloys. Part 1. Effect of casting defects. Journal of Light Metals, 2001. 1(1): p. 73-84. Wang, Q.G., D. Apelian, and D.A. Lados, Fatigue behavior of A356/357 aluminum cast alloys. Part II Effect of microstructural constituents. Journal of Light Metals, 2001. 1(1): p. 85-97. Mbuya, T.O., et aI., Micromechanisms offatigue crack growth in cast aluminium piston alloys. International Journal of Fatigue, 2012. 42(0): p. 227-237. Han, S.-W., S. Kumai, and A. Sato, Effects of solidification structure on short fatigue crack growth in AI-7%Si-0.4%Mg alloy castings. Materials Science and Engineering: A, 2002. 332(1-2): p. 56-63. Beck, T., et aI., Damage mechanisms of cast Al-Si-Mg alloys under superimposed thermal-mechanical fatigue and high-cycle fatigue loading. Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing, 2007. 468: p. 184-192. Beck, T., 1. Henne, and D. Loehe, Lifetime of cast AlSi6Cu4 under superimposed thermal-mechanical fatigue and high-cycle fatigue loading. Materials Science and Engineering a-Structural Materials Properties Microstructure and Processing, 2008. 483-84: p. 382-386. Gall, K., et aI., Atomistic simulations on the tensile debonding of an aluminum-silicon interface. Journal of the Mechanics and Physics of Solids, 2000. 48( I 0): p. 2183-2212. Ward, D.K., W.A. Curtin, and Y. Qi, Aluminum-silicon interfaces and nanocomposites: A molecular dynamics study. Composites Science and Technology, 2006. 66(9): p. 1151-1161. Gall, K., et aI., Finite element analysis of the stress distributions near damaged Si particle clusters in cast Al-Si alloys. Mechanics of Materials, 2000. 32(5): p. 277-301.

17. 18. 19.

Fan, J., et a!., Cyclic plasticity at pores and inclusions in cast Al-Si alloys. Engineering Fracture Mechanics, 2003.70(10): p. 1281-1302. Moffat, A..T., Micromechanistic analysis of fatigue in aluminium silicon casting alloys, University of Southampton: Southampton, 2007. p. 296. Moffat, A.J., et a!., The effect of silicon content on long crack fatigue behaviour of aluminium-silicon piston alloys at elevated temperature. International Journal of Fatigue, 2005. 27(10-12): p. 1564-1570.

403

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

MECHANICAL BEHAVIOUR OF COLD FORMED METAL-POLYMER LAMINATE AND THE INTERACTION OF ITS LAYERS I

Feidhlim 6 Dubhlaing l , David J. Browne l , Robin Rennicks2, Connor Rennicks2 School of Mechanical and Materials Engineering, University College Dublin, Belfield, Dublin 4, Ireland 2Prodieco Ltd, Unit 30, Cookstown Industrial Estate, Tallaght, Dublin 24, Ireland Keywords: Metal-Polymer Laminate, Tensile, Erichsen, AA8079 However this formability is not comparable to that of polymers [I], [2]. The addition of the polyamide layer provides greater elasticity to the aluminium, while the PVC provides rigidity to the formed part and aids in preventing springback [3]. The metalpolymer laminate exhibits mechanical properties that are superior to a metal sheet of equivalent thickness meaning lower forming forces needed during part production. To avoid both bending and extension forces being induced during and after forming the laminate, symmetry of the laminate properties and geometry is usually created about the middle layer. In some cases, unsymmetrical laminates in terms of geometry and/or properties are produced to meet a given need [4]. It is therefore of great interest to establish the formability of the laminate and its individual layers, with an understanding of the laminate failure in terms of the failure of the laminae [4], [5]. As an example, the mechanical properties of a tibremetal laminate are observed to display the plasticity of the metal sheet under tensile loading but with a much higher strain-tofailure created by the fibres than would be observed for the metal on its own [6]. In the consideration of laminates it is assumed that the individual laminae are perfectly bonded and that the adhesives and primer layer thicknesses are negligible, discounting the occurrence of shearing in these regions [4].

Abstract

This research is aimed at characterising and comparing the mechanical properties of a polymer/metal/polymer laminate and its individual layers. The laminate is cold deep drawn for pharmaceutical packaging, thus the properties were characterised at room temperature. The investigation was needed to understand how the layers contribute to formability in terms of elastic and plastic deformation. Particular attention was given to how the properties of the aluminium alloy layer were enhanced by the addition of each polymer layer. Tensile tests were carried out to determine material strengths, anisotropy and strain-hardening. These properties vary for each layer and the dominating properties in the complete laminate are of strong interest. Cup forming tests were performed to establish the forming depth of the constituent layers, carried out in line with the Erichsen cupping test. Wrinkling of the specimen during these cup forming tests had to be avoided, placing strong emphasis on understanding the stretching mechanism. Introduction

The metal-polymer laminate under investigation in this study is illustrated in Figure I and is a PA (Polyamide 6) Aluminium (AA8079) - PVC (Rigid Polyvinyl Chloride) laminate with layer thicknesses of 25,47 and 60 microns respectively. The three layers are bonded using appropriate adhesives and primers to prevent delamination during packaging production and its shelf-life. Each layer of the laminate contributes in its own way to the formability of the laminate during cold forming by punch and die tooling. To understand fully the mechanical properties and formability of the laminate, it is necessary to evaluate the limits of formability of each of the layers. Tensile testing of the layers and the laminate allow for the determination of these properties and how the layers may interact. Uniaxial testing of materials does not necessarily indicate the limit of formability in multiaxial deep drawing operations however, so the limit of formability was determined using a punch and die also. To aid in the determination of the properties, the various attributes of the layers that contribute to or are detrimental to the formability are discussed. Such effects as the water absorptivity of polyamide, the aging or strain-hardening effects of PVC or the annealing of the AA8079 aluminium alloy are quantified. The storage and handling history of the layers is unknown and thus these effects cannot be accounted for, but by comparing and contrasting results in relation to work previously carried out the formability of the laminate can be reasonably determined.

In a complete study of the mechanical properties of this laminate it is necessary to understand the mechanical properties of each layer and how they change with forming environment or pretreatments. Polyamide 6, while increasing the elasticity of the aluminium layer, also has high strength, excellent corrosion resistance, good wear resistance and self-lubricating properties [7], all of which are advantageous for cold formed pharmaceutical packaging. The disadvantage of PA6 is that it has a high water absorption rate, and the mechanical properties of the polyamide change with water content. Previous work has been carried out on

Theory

The aluminium layer in the laminate provides the barrier properties necessary to protect the pharmaceutical product being packaged, and the specific alloy is chosen for its high formability.

405

this [6], where accelerated conditions were used to determine the water absorptivity of the polymer and the resulting change in mechanical properties tested. An increase in water content did not affect the yield strength to a great degree, with the yield strength measured in the region of 44MPa, but the strain at fracture was reduced by 55.1% in the water-absorbed samples [7]. This indicates that the formability of the laminate is highly dependent on the production, transport and storage environments. particularly with respect to the humidity of the environments. PVC on the other hand has no noticeable water absorptivity but it exhibits yield and necking phenomena typical of tough thermoplastics [8]. Its stress-strain curve is similar to that for a metallic material but the transition between elastic and plastic deformation includes upper and lower yield strengths between which strain softening occurs, similar to that in Figure 2 [9]. Once the upper yield strength is reached shear bands form resulting in neck initiation, which leads to varying levels of stress and strain across the specimen being loaded [8]. The stress-strain curve determined from testing the polymer may not then truthfully represent the behavior of the polymer due to the varying stress levels across the neck and other sections of the specimen [10]. This neck then has a tendency to propagate through the specimen until it has completely necked [10]. It has been found that the necking behavior of PVC can be controlled by the thermal pre-treatment of the polymer, in terms of either quenching or annealing [8][11], [12]. The curves in Figure 2, representative of illustrations in [9], are examples of pretreatment effects. Quenching the PVC in iced water for example could eliminate a stress overshoot and strain softening, but as time elapses and the PVC ages, it will return to its original structure and a stress overshoot will result [8]. This increase in yield stress can exceed 10% if the PVC is annealed, with a combination of previous studies indicating an increase from 38MPa for quenched PVC to approximately 50MPa for commercial PVC [11]. As is evident in Figure 2 however, the post yield behavior of the PVC such as flow stress and strain hardening is irrespective of the thermal pre-treatment [9], [11].

occurs at yield [8]. This necking, caused by highly localised levels of strain [13], can then lead to thermal rupture of the neck at sufficiently high oS, thought to be greater than 100h,l [8]. The thermal fracture is as a result of a localised increase in heat in the neck not having sufficient time to dissipate [8], [12]. Another aspect of the plastic deformation of PVC and that can be observed throughout the duration of material testing and forming, is the stress whitening of the polymer. The polymer becomes opaque as it deforms, due to the formation of microvoids throughout the material [13]. The dimensions of these microvoids are on the same scale as the wavelength of light and thus change the refractive index of the polymer. It must be taken into account that this change is more a visual alteration than a change in properties, as the stress-whitened PVC can withstand much the same load as PVC absent of microvoids [13]. This stress whitening occurs once the polymer begins to yield and the polymer chains begin to reorient. A stress 'plateau' is observed during plastic deformation then during which the polymer chains begin to align, but once a significant amount of alignment is achieved and thermal fracture is avoided, strain hardening of the polymer occurs [9], [13]. Aluminium, like many metallic materials, does not display this type of yield and necking phenomena, rather once yielding occurs the material will tend to deform uniformly while strain hardening develops. Once the metal reaches its tensile strength it may then neck [13], but necking is usually negligible [6]. The AA8079 alloy used in this case is extensively used in the packaging industry as a whole [14], due to its excellent elongation at failure and tensile strength versus other available alloys [3]. The alloy is an AI-Fe-Si ternary alloy, amongst other elements in minor proportions [14], and ASTM B479 outlines the acceptable percentage of each alloying element as in Table 1 [15]. Examining the elements present in the alloy it has been observed that the presence of an Mg solute has detrimental effects on the speed at which the alloy can be formed, while the presence of Si atoms increases this speed. This is due to the negative heat of formation between Mg and Al creating immovable clusters that restrict grain boundary sliding, while the Si atoms weaken the aluminium bonds promoting grain boundary sliding [14].

(%)

Table 1 A:\8079 Softening 1.8

0.00

IMiO

OAO

0.1

0.05

0.1

0.08

0.05 each

0.15 total

98 min

In forming parts from this alloy in its annealed state its formability can be taken full advantage of: with the annealing temperature also playing a role in this formability [14], [16]. The effects of annealing are graphed in Figure 3, determined from previous work [16]. If low temperature annealing is carried out, a decrease in the ductility of the alloy is observed due to the removal of dislocations [16], but at higher annealing temperatures greater control over recrystallisation takes place increasing the deformation potential of the alloy [17]. The necking behavior is also dependent on annealing with a localized neck occurring in the low-temperature annealed state and more diffuse necking in the high-temperature annealed state [16], [17]. Through higher temperature annealing greater control can be obtained on the grain size in the alloy. A differential grain size across a section of material causes non-uniformities in the flow stress of the material [18], whereas a fine, uniform grain size is preferred for sheet metal forming [19]. An ASTM grain size of

1.00

Strain curves for a

with and effect of heat tbe PVC stress-strain curve straill

treatment Oil

0.2

Similar to the heat treatment of PVC, the strain rate has an influence on the stress-strain curve and the eventual fracture of the polymer [13]. Three different ranges for strain rate have been proposed, with differing stress-strain curves and failure mechanisms associated with each. It is deemed that below a strain uniform deformation of PVC can be rate (oS) of Ih'l induced with minimal stress overshoot, while above this necking

406

7 or finer is sought, which corresponds to a grain diameter of 31.S!lm for a uniform, randomly oriented, equiaxed structure [19], [20]. The relative size of the grains compared to the sheet thickness may have an effect on the bulk properties of the material [21], [22], as illustrated in Figure 4, similar to that in previous studies [14].

separation. As the layers are very thin, the determination of strain throughout the test was accomplished by means of a non-contact Zwick/Roell VideoXtens extensometer. The VideoXtens requires that the gauge marks are at a slight angle to the transverse axis of the specimen to allow the marks to be located more easily by the camera.

1111 dimensions in mm

o

100

200

300

400

In determining the elastic modulus, and the yield and tensile strengths the engineering stress (a) - strain (s) graphs are used. To then determine the strain-hardening exponent (n) it is necessary to plot the true stress (s) - strain (e) curves. For a material with uniform deformation in the gauge n is determined by conventional true stress-strain techniques, as in Equations I -3 [19], [26], where n is the slope of the log-log plot of the true stress-strain curve and C is the intercept of the line [19].

Annealing Temperature fOC) .3 Effect of

temper11hm::

Oil

of

the

45 41.1

120 100

'1 ! '"'I!:!"

...

VI

35

80

31.1

60

-aT

25

"AI

20

i

41.1

15

21.1

11.1 W 5

0 0,00

0.11.1

0.20

O.SO'

0.40

0,50

0.60

s=Cell

o

=>

s= a(l +s)

(1)

e=ln(J+s)

(2)

In s = In C + nln e

(3)

These equations can be applied to the laminate, aluminium and PA but due to the localized necking behavior of PVC other methods have to be adopted. Possible methods have been studied previously such as the Gaussian model in Equation (4) [S], [II], where Y is the yield stress, A is the extension ratio 1/10, and K is the strain hardening modulus. On occasion it may be appropriate to select a small volume of the material under testing for which uniform deformation can be assumed but there is also potential for issues with the strain softening and strain hardening overlapping [9]. For PVC however it has been observed on occasion that the strain softening can disappear at low strain levels of A=I.2 [II]

Sheet Thickness (mm)

Experimental To determine the specific mechanical properties of the materials, tensile tests were conducted on both the laminate as a whole and its individual layers. In addition to this, an Erichsen cupping test [23] was carried out to determine the material forming depths.

(4) Normal anisotropy of the layers was measured using Equations 5 and 6. Generally the anisotropy is calculated using the width and thickness of a specimen but measuring the thickness can be troublesome so the length can be used instead by adjusting the equations accordingly [19], [24]. The anisotropy is usually calculated at a strain of 10, 15 or 20% [24].

In conducting uniaxial tensile tests the elastic and plastic deformation of the layers could be characterized along with the transition between the two types of deformation. For the cold deep drawing of a material it is of great importance to understand the work-hardening effects, the anisotropic behavior and the failure limit of the material [24]. To carry out the testing efficiently a suitable tensile specimen should be used and as the material under study is in the form of thin sheet a flat dog-bone specimen was chosen. The laminate in this case is designed to exhibit high levels of strain at failure and thus a tensile specimen from [25] was deemed appropriate. The specimen is illustrated in Figure 5, with the specific dimensions of [25] labeled. The specimen includes a 25mm gauge length, 6mm gauge width and an SOmm grip-to-grip

(5) (6) Finally the Poisson's ratio was calculated as the ratio of lateral strain Sy against axial strain Ex given in Equation 7, and the strain sensitivity (e) was determined by analysing the tensile strength of the laminate against the strain rate. The tensile strength was

407

being obtained most often. As Young's modulus is an intrinsic property of a given material it is thought that the deviation in this case is as a result of grain size effects [21], [22] and possible surface defects on the samples. The Young's modulus for the laminate has been determined also, but the accuracy of this value is uncertain and will have to be reevaluated once more certainty is gained as regards E for the aluminium.

chosen for the determination of (E) as the failure of the material is of greater interest in cold forming than the yield strength. (7) Erichsen Test The Erichsen test is a simple forming limit test designed for metallic sheets [23]. The forming depths of the laminate and the three layers were all determined using the tooling illustrated in Figure 6, where the forming limit is denoted by the Erichsen Index (IE). It must be noted that minor deviations to the test method had to be made such as (i) the maximum blank holding force of the machine being used was 8kN while the standard specifies lOkN, (ii) the safety requirements of the machine did not allow for close observation of the test so a series of samples were tested interpolating between forming depths until the depth at failure was found, and (iii) the punch was made from PTFE where its self-lubricating properties [7] were utilised rather than using graphite grease. The latter deviation was due to the extensive use of PTFE punches in cold forming of the laminate in the pharmaceutical packaging industry.

100 30

I III It!

III

"Aluminium

-laminate

60 40 20 0 0.00

1:1.40

0.60

Stmin 7

stress-strain for eacn of tne ami tile laminate as a wl1ok'

From initial observation on the plasticity of the layers on the other hand, it is evident that the strain-to-failure for both the PA and PVC far exceed that of the aluminium, as expected. The strain-to-failure of the laminate is thus greater than that of the aluminium by a factor of 3. The yield strengths for each layer are similar, indicated in Table II, and thus the yield strength of the laminate is also comparable. The tensile strength on the other hand is strongly influenced by that of the PVC. When the stressstrain curves for the laminate are determined the cross-sectional area used is that of the entire laminate, i.e. thickness of 137f.lm, but a large proportion of this is the thickness of the PVC which has the highest strain-to-failure ratio. This in addition to the assumption of perfect bonding of the layers [4] could indicate the reduction in tensile strength of the laminate compared to that of the aluminium.

Results

4

Illustrated in Figure 7 are typical engineering stress strain curves for each of the three layers and the laminate as a whole. These curves were detennined using various strain rates, but as the strain rate increases the accuracy of results decreases. This is due to the sampling rate of the extenso meter eventually reaching a maximum and becoming unable to satisfactorily plot a smooth curve above this rate. For this reason the lowest strain rate of 0.0025s· 1 is used to understand the fundamental properties and interactions of the layers. The foremost property of the materials being tested is the Young's Modulus (E). E was determined for each of the four samples, with the values for the polymer layers given in Table II. In determining the Young's Modulus for the aluminium layer issues arose, with the result deviating from the inherent value of 70GPa [19]. The Young's modulus in fact tended to vary slightly from one sample to the next, with values of approximately 30GPa

-Aluminium -PA ·····Lamlnate

-s

-6

-4

In (True Strain)

material The strain hardening of the aluminium and PAis evident in the laminate, without any noticeable contribution by the PVc. Up to the failure strain for the laminate, the PVC exhibits

408

only strain softening followed by a plateau stress. The minor strain-hardening of the PVC occurs at strains significantly higher than the failure of the laminate, thus having no obvious effect on the laminate properties. The strain hardening curves derived from Equation 3 are illustrated in Figure 8 and their respective slopes given in Table II. The strain hardening curve for PVC is not illustrated due to the issues with determining its true stress - strain curve previously discussed [8], [11]. It is also of interest to note that the assumption of perfect bonding of the layers is strengthened by the observation that there is no localized necking of the PVC layer at yielding, which would occur if the adhesive bond strength was not sufficient.

reference, due to the water absorption tendencies of the PA, but the prior history of the PA is unknown and previous conditions may not have been suitable to prevent excessive water absorption.

140 120

.-

:.

100

UI UI

80

60

f! 40 2()

their

Aluminium E (GPa) Yield (MPa)

UTS

PA

PVC

2.74

2.36

56.43 91.20

43.39 110.75

49.06 44.78

0.25 0.26 0.47

0.42

Laminate

v r

if

49.65 66.33

0° 0.282

PA

0.46

45° 0.360

1:1.10

0 . 20

0.30

0.40

0.50

stness-strain for mEed «(10} and transverse directions of 1'/\

As the laminate is used in a high speed forming process it is of interest to establish the strain rate sensitivity of the laminate, with a number of strain rates (E) tested (0.0025, 0.025, 0.0505, 0.10 1 and 0.2525s· I ). Currently the laminate is formed with a punch speed of 7000mm/min which corresponds to E=4.667s· 1, but it was found that no great variation existed in the failure strain above 0.025s· 1 with a minor increase of 1-2MPa in the tensile strengths of the laminate and aluminum, from one E to the next. It is important to carry out a strain rate sensitivity test to determine whether a low cycle time would be of signiticant importance in the formability of the laminate. Finally, the Erichsen cupping test was conducted, with a trend in the results similar to that for the strain-to-failure of the different layers, as in Figure 7. The tests were carried out three times for each of the four specimens, and were carried out in random order. This randomization was introduced to avoid any errors associated with the test environment, inhomogeneities in the materials and in particular any wear of the PTFE punch during forming. The maximum forming depths without failure are collated in Table IV.

0.21 0.29 0.63

to the Tab!e lJJ Strains at fail!lre for different directioll of tile lamillate and aluminium

Laminate

0.00

Strain

(MPa)

n

()

90° 0.275 0.36

As the laminate approaches failure, there is no significant necking present. It has a tendency to fail rapidly once the UTS is reached, with the strains to failure listed in Table Ill. The strains for 0°, 45° and 90° to the rolling direction are listed, along with those for the aluminium. It is apparent that the strain to failure is higher in the 45° direction, indicating anisotropy of the aluminium, which is evident in the laminate also. The normal anisotropy parameter (r) is given in Table II for both the aluminium and the laminate. The values for both samples differ, which indicates there may be some influence of the polymer layers on the anisotropy. The normal anisotropy is used to give a reasonably good approximation to the limiting drawing ratio of a material [19]. After the tensile specimens fail there is significant elastic recovery of the laminate in the rolled direction, with the specimens curling up tightly on the PA side of the laminate and out to the side of the specimen for the 90° direction. This is thought to be due to bending forces discussed previously [4], but to try and quantify this behavior the PA was tested in both the 0° and 90° directions, resulting in the curves in Figure 9. It is clearly evident that there is more strain-hardening of the PAin the 90° direction (n = 0.62 as opposed to n = 0.42 for 0°), explaining the curling and recovery for 0°. This characteristic could have a significant effect when cold-forming the laminate and in spring back of the formed pocket. When these tests were being carried out the temperature and humidity were recorded for future

Table lV

Forming Depth (mm)

11.95

7.113

13.907

18.407

Discussion and Conclusions In conducting this study it was hoped that an insight would be attained into the formability of the laminate and how each of the layers contribute to this formability. Initial analysis was carried out to determine how each of the materials behaves under testing in terms of test environment and pre-treatments. This analysis concluded that significant variation in results can be expected from each of the three layers, in terms of waterabsorptivity for PA [7]. strain rate and thermal pre-treatment for PVC [8], [11] and annealing temperature and sheet thickness for AA8079 [14], [16]. Each of the materials was subject to tensile testing to determine and compare the properties of each.

409

The aluminium alloy exhibits good strain-to-failure compared to that for conventional aluminium alloys, with this formability extended considerably by laminating it with polymer layers. The tensile strength of the laminate is lower than that for the aluminium layer, which allows for lower forming forces needed during cold deep drawing of the laminate than would be needed for an aluminium layer of equivalent thickness. The yield and tensile strengths and the plasticity of the aluminium layer also correlate well with previous studies [14], [16]. It was observed that the plasticity determined during testing correlated to a high in Figure 3. This then annealing temperature of strengthens the idea of grain-refinement contributing to the formability of the alloy. The elongation-to-failure of this aluminium layer is enhanced by the addition of the polymer layers. The polymers possess greater elongation-to-failure than the aluminium on its own, and due to exceptional bond strength between the layers the high elongation attributes of the polymers are attained by the aluminium. Anisotropy and strain hardening are evident in the laminate, which can be detrimental to the cold deep drawing and stretching of the laminate, thus the need to quantify them. The multiaxial stretching of the laminate was also tested by means of the Erichsen test and good correlation between these results and those from the tensile tests exist. As the storage and transport environments of the laminate and layers is not precisely known, the exceptional properties of the laminate could in fact be greater than found here, through stringent control of the storage and forming environments. If excessive moisture absorption of the PA and rapid quenching of the PVC were controlled the maximum formability of the laminate could be determined. The results in this case though are deemed to be representative of those expected in industry as aging and water absorption of the layers is inevitable over time.

[9]

K. Chen and K. S. Schweizer, Macromolecules, vol. 44, no. 10, pp. 3988-4000, May 2011

[10]

R. N. Haward, Polymer, vol. 28, no. 9, pp. 1485-1488, 1987

[II]

R. N. Haward, Macromolecules, vol. 26, no. 22, pp. 58605869, 1993

[12]

A. Cross et ai, Polymer, vol. 20, no. 3, pp. 288-294, 1979

[13]

A. S. M. International, Characterization and Failure Analysis of Plastics. ASM International, 2003

[14]

K. Deliji6 and D. Radonji6, 'The Mechanical Properties and Corrosion Behavior of an AI-Fe-Si (AA8079) Twin Roll Cast Foil", 2nd International Conference on: Deformation Processing and Structure of Materials, Belgrade, May 2005

[15]

B07 Committee, "Specification for Annealed Aluminum and Aluminum-Alloy Foil for Flexible Barrier, Food Contact, and Other Applications," ASTM International, 2006

[16]

M. Aghaie-Khafri and R. Mahmudi, JOM Journal of the Minerals, Metals and Materials Society, vol. 50, no. II, pp. 50-52, 1998

[17]

R. Mahmudi, Scripta Metallurgica et Materialia, vol. 32, no. 12, pp. 2061-2065, Jun. 1995

[18]

H. Wielage et ai, Journal of Materials Processing Technology, vol. 212, no. 3, pp. 685-688, Mar. 2012

[19]

References

S. Kalpakjian and S. Schmid, Manufacturing, Engineering & Technology, 5th ed. Prentice Hall, 2005

[ I] [2]

D. A. Dean. Pharmaceutical Packaging Technology. Taylor & Francis, 2000

[20]

M. Weiss et aI, Journal of Engineering Materials and Technology, vol. 129, no. 4, p. 530, 2007

[21]

E04 Committee, "Test Methods for Determining Average Grain Size," ASTM International, 20 10 P. J. M. Janssen et aI, Materials Science and Engineering: A, vol. 419, no. 1-2, pp. 238-248, Mar. 2006

[3]

[4]

Alcan Packaging University, Cold Formable High Barrier Laminates, Oberlingen, 5th _6th March 2008

[22]

H. S. Kim and M. B. Bush. Nanostructured Materials. vol. ll, no. 3, pp. 361-367, May 1999

R. M. Jones, Buckling of Bars, Plates, and Shells. Bull Ridge Corporation, 2006

[23]

Erichsen Cupping Test, BS EN ISO 20482, 2003

[24]

C. O. Osueke et aI, International Journal of Advanced Engineering Sciences and Technologies, TJAEST, Vol 9, NO.2, pp. 280-288, 20 II

[25]

Plastics - Determination of Tensile Properties, BS EN ISO 527-3. 1996

[26]

P. Han, Tensile testing. Asm IntI, 1992

[5]

Z. Hashin, Journal of Applied Mechanics, Vol. 50, pp. 481505, Sept. 1983

[6]

P. Iaccarino, A. Langella, and G. Caprino, Composites Science and Technology, vol. 67, no. 9, pp. 1784-1793, Jul. 2007

[7]

R. ZHAO et ai, Transactions ofNonferrous Metals Society of China, vol. 16, pp. s498-s503, 2006

[8]

A. Cross and R. N. Haward, Polymer, vol. 19, no. 6, pp. 677-682, 1978

410

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

MECHANICAL AND TRIBOLOGICAL PROPERTIES OF AA2124-GRAPHENE SELF LUBRICATING NANOCOMPOSITE A. Ghazaly, B. Seif, and H.G. Salem* Mechanical Engineering Dept., YousefJamil Science and Technology Research Center (Y.TSTRC), American University in Cairo-Egypt

Key words: AA2124-Graphene-Extrusion-self lubricant-Milled powder.

and hence cost effectiveness compared to other conventionally reinforced composites. Due to the ease of fabrication and the promising results, graphene has been investigated as reinforcement in polymer matrix composites [5, 6]. Only few researches investigated the effect of graphene in metal matrix composites due to the complex microstructure and difficulties in dispersion. Aluminum alloy 2xxx series are used in high technology areas such as the construction of the internal structure of aircrafts and tanl Tj I T phase are developed, Fig. (5), as listed in table (I). It is observed from XRD that he phase is not pronounced in the specimens aged at 473K, whereas it is clearly pronounced as the specimens are aged at 673K and AI 12Mg I7 phase also detected. Therefore, it may be

Table (Ia): The identified phases for Alx at. %Mg- 2 at. %Zn alloys at 473 K. Aging Temp. X at.% phases

Mg Mg 4Zn 7

Mg;,(AI,Zn)49

A1 12Mg 17

-------

cubic a=I.422

-------

2

hexagonal a= 0.50 c= 0.87

Monoclinic a= 2.596, b=I.428, c=0.524 y =1 02.5°

cubic a=1.422

cubic a=I.0553

2.4

hexagonal a= 0.50 c= 0.87

-------

cubic a=1.422

-------

3

hexagonal a= 0.50 c= 0.87

Monoclinic a= 2.596, b=I.428, c=0.524 y =102.5°

cubic a=1.422

-------

4.2

hexagonal a= 0.50 c= 0.87

-------

cubic a=I.422

-------

1.8

433

473 K

MgZn2 hexagonal a= 0.50 c= 0.87

Table (lb): The identified phases for AI- x at. %Mg2 at. o/oZn alloys at 533 K. Aging Temp. X at.% phases

1.8

2

2.4

Table (2a): The Precipitate's size of phases in AIx at.%Mg- 2at. %Zn alloys at 473K.

533 K

Aging Temp. X at.% phases

Mg MgZn, hexagonal a= 0.50 c= 0.87 hexagonal a= 0.50 c= 0.87 hexagonal a= 0.50 c= 0.87

3

hexagonal a= 0.5225 c=0.8568

42

hexagonal a= 0.5225 c=0.8568

Mg 4Zn7 Monoclinic a= 2.596. b=I.428. c=0.524 y =1 02.5°

Mg,,(AI.Zn)49 cubic a=1.422

AI 12Mg17

1.8

-------

2

-------

cubic a=1.422

cubic a=1.0553

-------

cubic a=1.422

-------

cubic a=1.422

-------

cubic a=1.422

cubic a=1.0553

Monoclinic a= 2.596, b=I.428, c=0.524 y =102.5° Monoclinic a= 2.596, b=I.428, c=0.524 y =102.5°

2.4 3 42

MgZn,

Aging Temp. X at.% phases 1.8

2.4

Mg

3

Mg 4Zn7 Monoclinic a= 2.596, b=I.428, c=0.524 y =102.5°

AI 12 Mg 17

1.8

hexagonal a= 0.5225 c=0.8568

2

hexagonal a= 0.5225 c=0.8568

-------

hexagonal a= 0.5225 c=0.8568

Monoclinic a= 2.596, b=I.428, c=0.524 y =102.5°

cubic a=1.422

hexagonal a= 0.5225 c=0.8568

-------

cubic a=1.422

-------

hexagonal a= 0.5225 c=0.8568

Monoclinic a= 2.596, b=I.428, c=0.524 y =102.5°

cubic a=1.422

cubic a=1.0553

2.4

3

42

2

673K

Mg;,(AI,Zn)49 cubic a=1.422

-------

cubic a=1.422

-------

Precipitate Size (nm) MgZn, 30.13 ± 1.52 18.39* 47.97 ± 1.37 32.34 + 077 40.62*

-------

9.02* -------

3139* -------

MgdAI,Zn)49 52.50 ± 0.89 43.96 ± 0.91 63.48 ± 0.49 47.60 + 0.44 70.56*

Al 1,Mg17 -------

25.22* -------

-------

-------

Table (2b): The Precipitate's size of phases in Alx at.%Mg- 2at. o/oZn alloys at 533K.

Table (I c): The identified phases for AI- x at. %Mg2 at. o/oZn alloys at 673 K. Aging Temp. X at.% phases

473 K

42

533 K Precipitate Size (nm) MgZn2 29.42 ± 1.34 56.03 ± 0.54 46.45 ± 0.26 30.56 ± 0.64 28.79*

727* -------

-------

26.98* 19.72*

Mg32(AI,Zn)49 49.02 ± 1.01 52.95* 52.35

± 1.01 50.44* 47.04*

AI 12Mg17 -------

60.34

± 0.70 -------

-------

52.92

± 1.06

Table (2c): The Precipitate's size of phases in AI- x at.%Mg- 2at. o/oZn alloys at 673K. Aging Temp. X at.% phases

cubic a=1.0553

1.8 2

Calculation of precipitate size forAI- x at.% Mg- 2 at.% Zn, (x=1.8, 2, 2.4,3 and 4.2) alloys, by using scherrer equation (4-1) are listed in Table (2 a-c), at different aging temperature (473 K, 533 K and 673 K).

673K Precipitate Size (nm) MgZn, 36.51 ± 1.32 28.19 ± 1.54

7.22* -------

2.4

3822*

13.72*

3

38.78 ± 0.37

-------

4.2

15.97*

6.66*

MgdAI,Zn)49 43.54 ± 0.53 56.07 ± 0.71 57.74 ± 078 59.07 ± 1.12 51.79 ± 112

Al 1,Mg 17 -------

-------

49.93* -------

5115*

Scherrer equation in the form:

KA

D=--f3 COS e

(1 )

where K is a constant approximating to unity; its value is related both to the particle shape and to the

434

way in which p and Dare defmed. p is the width of the powder reflection peak free from all broadening due to the experimental method employed in observing it. Most investigators define P as the angular width at half maximum intensity FWHM. Bragg gave a simplified derivation for the Scherrer equation and found that K=0.89, the patterns were run with Cu as a target, and graphite as a monochrometor 00.=0.154178 nm [16]. As observed by SEM, the precipitates of 11' and T' developed in the present work are similar in shape but rather larger in size in comparison with those detected in AI- 2.1 at.% Zn - 1.7 at.% Mg [3].This may be due to the difference in heat treatment of the material and the lower Mg:Zn ratio. They were aged their specimen at two different temperatures and times. Namely at 363 K for 168 h and at 423 K for 45 h. On other hand the particle size of our work is also slightly different from that obtained for AI- 2.3 at.% Zn - 1.38 at.% Mg aged at 373 K for 5 h + 6 h at 423 K. Our specimens Al- 2.0 at.% Zn - 1.8 at.% Mg and AI- 2.0 at.% Zn - 4.2 at.% Mg are aged at 473 K and 673 K for 0.5 h. Table (3) shows the particle size for the different alloys and different heat treatment.

(a) (b) Fig. (6) SEM micrograph of AI- 2 at.% Zn - 1.8 at.% Mg at 473 K at 673 K for 30 min.

(b) (a) Fig. (7) SEM micrograph of Al- 2 at.% Zn - 4.2 at.% Mg alloy, (a) aged at 473 K (b) aged at 673 K for 30 min.

Tn conclusion the surface density of precipitates (number of precipitates per unite area) is decreased and the particle size increased with increasing the aging temperature. Tn all studied alloys AI- 2 at.% Zn - x at.% Mg, (x=1.8, 2and 4.2), the particle size of grown precipitates increased with increasing the aging temperature as shown in Fig. (8).

Table (3): Particle size for the different alloys and different heat treatment

Alloy AI- 2.1 at.% Zn1.7 at.% Mg AI- 2.3 at.% Zn1.38 at.% Mg AI- 2.1 at.% Zn1.7 at.% Mg AI- 2.0 at.% Zn1.8 at.% Mg AI- 2.0 at.% Zn4.2 at.% Mg

Particle size (nm)

Heat treatment

3

168 h at 363 K

5

+ 6 h at 423 K

10

Ref X=4.2

5 hat 373 K

45 h at 423 K

lOa

0.5 h at 473 K

ISO

0.5 h at 673 K

[3] ru "CC 2000

[5]

'iii

1

1000

[3]

X=2

0

This work This work

0.99979

"oj

It can be noticed that the temperature is more effective than time in the coarsening of the precipitates. It is noteworthy to mention that the precipitates are coarser in the specimen which contains higher Mg content. To confirm the developed processes at the DSC reaction peaks, the SEM specimens are aged at each reaction peak temperature. As AI- 2 at.% Zn - 1.8 at.% Mg and Al- 2 at.% Zn - 4.2 at.% Mg as example, are aged at 473 and 673 K. The SEM examinations are shown in Figs. (6, 7 (a, b)).The density of precipitates decreased with increasing aging temperature due to coarsening of the precipitates on the expense of their number [5]. It is shown from the SEM electron micrographs that as the Mg concentration increases in the alloy the density of the precipitates and size are larger after aging for 30 min at 473 K, Figs. 6(a) and 7(a). While after aging at 673 K, the precipitates begin to dissolve, Figs. 6(b) and 7(b).

450

/ 500

550

600

T [K]

I

, 650

700

Fig. (8): Particle size of precipitates vs. aging time for AI-2 at.%Zn- x at.% Mg. The behaviour of particle size as a function of aging temperature is found to fit exponential growth as [17]:

T

D = a exp( - ) + DO b

(2)

where a and b the fitting parameters and Do is the particle size at 473 K of exponential formula which differ in the alloys as shown in Table (4). Table (4): The fitting parameters for Al- 2 at. %Znx at %Mg alloys X

1.8 2 4.2

435

Fitting parameters a 0.01023 0.002 22.70861

b

58.80533 47.52602 192.49639

Do 185.17416 90.17734 -15.0539

Dependence of the Volume Density on Aging Temperature:

6: At lower aging temperature 473 K for 30 min, the concentration and size of developed precipitates is higher for the specimen containing higher Mg concentration.

Tn the following, we tried to calculate the volume density of the particles as a function of aging temperature in the considered alloys. Tn this concern, calculating the linear density of the particles NL/.lm· 1 and the surface density NA/.1m· 2 are necessary. The volume density Nv can be calculated from [17,18]: . tv V

References

[IJ [2]

7rN A

= - - (3) 4N L

[3J

It can be easily noticed that N v decreases with increasing aging temperature, because of its growth with raising the aging temperature. The variations of Nv as a function of temperatures for all specimens AI- 2 at. %Zn -x at. %Mg, (x= 1.8, 2, and 4.2) alloys are represented in Fig. (9). The obtained values of Nv - 10 18 m· 3 are in good agreement with those obtained by Celotto [19] as 109 mm· 3 in the same range of temperature. The behaviour of N v as a function of temperatures IS presented in Fig. (9).

[4] [5J

[6J [7]

[8J ge.QOE+018

[9] .. 1

TIKI

[10]

Fig. (9) Volume density vs. aging temperature for AI- 2 at.% Zn - x at.% Mg. Conclusion

[llJ

I: The alloys can be described as having typically very fine precipitates microstructure. The G.P. zones, YJ' and l' phases are the most effective hardening phases. The reason for the hardening effect is the interface coherency in Al planes. A high number density of precipitates is responsible for high hardness, which leads to improvement of the mechanical properties. 2: The precipitation sequence of the supersaturated AI- 2 at.% Zn - x at.% Mg, (x=1.8, 2, 2.4, 3 and 4.2)alloys, based on the combined results ofHY, and DSC, confirmed by SEM and XRD can be written as follows: Supersaturated solid solution--+ solute - vacancy clusters + G.P. zones--+ intermediate YJ' and l' phases--+ equilibrium YJand T phases. 3: From XRD analysis, the crystal structure of both YJ' (MgZn2), YJ (MgZn2) are hexagonal, l' (Mg32 (AI, Zn)49) and T (Mg32 (AI, Zn)49) phases are identified as a cubic structure. A few additional precipitates of AI 12Mg l7 and Mg4Zn7 are also existed. 4: From SEM, precipitates size increases, while particle density decreases with increasing heat treatment. 5: The volume density of the precipitated particles has an average Nv of _IOIR m- 3, and the N v decreases with aging temperature.

[12]

[13]

[14J

[15]

[16J

[17J

[18]

[19J

436

C. Jihua, C. Zhenhua, Y. Hongge, Z. Fuquan and 1.. Kun, J. of Alloys and compounds, 461 (2008) 209-215. 1.. Bourgeois, B. C. Muddle and J. F. Nie, Acta Mater., 49 (2001) 2701-2711. S. K. Maloney, K. Hono, 1. 1. Polmear and S. P. Ringer, Scri. Mater., 41 No. 10 (1999) 1031-1038. J. C. Werenskiold, A. Deschamps and Y. Brechet, Mater. Sci. and Eng., A293 (2000) 267-274. T. Engdahl, V. Hansen P. J. Warren and K. Stiller, Mater. Sci. and Eng., A327 (2002) 59-64. T. Ma and G. den Ouden. Mater. Sci. and Eng., A266 (1999) 198-204.

X. Z. Li, V. Hansen, J. Gj0nnes and 1.. R. Wallenberg, Acta Mater., 47 (1999) 26512659. 1.. Hadjadj, R. Amira, D. Hamana and A. Mosbah, J. of Alloys and compounds, 462 (2008) 279-283. G. W. Lorimer, Precipitation Processes in Solids, ed. K. C. Russel and H. 1. Aaronson, chapter 3, Met. Soc. AlME, New York (1978). P. Liang, T. Tarfa, 1. A. Robinson, S. Wagner, P. Ochin, M. G. Harmelin, H. J. Seifert, H. 1.. Lukas and F. Aldinger, Thermochimica Acta, 314 (1998) 87-110. N. Afify, A. Gaber, M. S. Mostafa and Gh. Abbady, Journal of Alloys and compounds, 462 (2008) 80-87. A. Gaber, N. Afiry, M. S. Mostafa and Gh. Abbady Journal of Alloys and Compounds, 477 (2009) 295-300. TCDD card No. 85-1327 [Ref H. E. Swanson and E. Tatge, Nat!. Bur. Stand. (U.S. ), Circ. 539, I (1953) 359]. W. 1. Kim. J. D. Park and U. S. Yoon, Journal of Alloys and Compounds, 464 (2008) 197-204. T. Sumitomo, C. H. Caceres and M. Veidt, Journal of Light Metals, 2 (2002) 49-56. H. P. Klug and 1.. E. Alexander, "X-Ray Diffraction Procedures", J. Wiley and Sons Ltd, New York (1959). N. Afify, A. Gaber and Gh. Abbady, Materials Sciences and Applications, 2 (2011) 427-434. M. Kiritani, Y. Shimomura and Sho Yoshida, J. Phys. Soc. Japan, 19 (1964) 1624-1631. S. Celotto,Actamater., 48 (2000) 1775-1787.

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013 Design and Development of a Permanent Mould for the Production of Motor-Cycle Piston in SEDI-Enugu. Ilochonwu C.E. I Nwonye E.l 2 ., Scientific Equipment Development Institute,( SEDT-E), Box 3205, Enugu, Nigeria. Key words: Casting 1, Heat treatment2, Motorcycle piston3

process because the shape and attributes of the mold will

Abstract

directly affect the final product. Mould design affects the

This paper demonstrates the possibility of designing and

shape, configuration, quality, and uniformity of a product

developing a four-cavity low pressure die casting mould in

created through the die casting process.

SEDI-Enugu. This research and development included the design of the die blocks and the core box for core production.

1.

Draft

Furthermore, the die blocks were cast, machined and later heat treated to obtain the desired properties of a mould

Draft is the degree to which a mould core can be tapered. A precise draft is needed to smoothly eject the casting from the die, but since draft is not constant and varies according to the angle of the wall, features such as the type of molten alloy used, shape of the wall, and depth of the mould can affect the process.

material. After which the components were assembled and mounted on the Low pressure die casting machine. Emphasis was made on the coating of the mould and crucible pot internal surfaces with a solution of zinc oxide and sodium silicate. This is done prevent iron entrapment in mass of the molten metal as well as to minimize the sticking of castings on the mould during casting operation. Introduction

Figure 1: Draft angles sample [1].

There is hardly any street corner in Nigeria where one cannot find tricycles and motorcycle used as a means of transportation. For this reason, the research and development of a four-cavity low pressure die casting mould for piston production became imperative for the technological and economic development of the transportation sector in Nigeria. Since there is high unemployment rate in the country, so many people depend on this means of transportation both businesses for their daily survival and easy movement of goods and services.

2.

Fillets

A fillet is a concave junction used to smoothen an angled surface. Sharp corners can hinder the casting process, so many moulds have fillets to create rounded edges.

Nigeria's GDP would have been greatly enhanced if the volume of importation of pistons is reduced drastically through local production. In addition, employment opportunities would have been created, and vices would have be curbed in Nigeria. Figure 2: Radii design sample

This need prompted us to design and develop the four-cavity die casting mould for motorcycle and tricycle piston production in Nigeria.

3.

Parting Line

Design Considerations The parting line connects different sections of the Mould design is one of the most important steps in the

mould together. If the parting line becomes

437

deformed from work strain. material may seep through the gap between the mould pieces, leading to non-uniform moulding and excessive seaming. Figure 6: Hole design sample 7. Sprues, runners and gates should be so provided that they facilitate the removal of the casting. Sprue holes are always tapered. A sprue pin is usually fixed at the inner end of the

Figure 3: (a) Parting line and (b) Mismatch[l].

sprue to deflect metal into the runner. Runners or runners are cut at the die parting. Gates join the runner with the die

4. Bosses

cavity [3]. Bosses are die cast knobs that serve as mounting points or stand-offs in mould design. Manufacturers often add a hole to the interior structure of the boss to ensure uniform wall thickness in a molded product.

Figure 4: Boss design sample 5. Ribs: Die cast ribs can be used to improve material

Figure 7: Four-cavity Motor cycle piston die mould designed

strength

with Solid Works package

in

products

lacking

the

wall

thickness

Sink

required for certain applications.

Ejecting system

Figure 5: Ribs design sample

Ejecting pin

6. Holes and Windows Holes or windows in a die cast mould directly affects the ease of ejecting a completed moulding and enables the creation of substantial drafts.

Figure 10: Die Blocks Production Development of die blocks involves a lot of manufacturing

438

operations and they are as follows:

We have seen that machining processes involve high local temperatures and high friction at the chip-tool interface. Thus

•

Gravity Casting of Cast iron blanks.

•

F ettling of the Cast.

•

Annealing the Blank.

•

Milling Operation.

chips from the cutting area, protect the newly machined

•

Turning Operation.

surface against corrosion.

•

Boring Operation.

Non-traditional Machining Processes

•

Drilling Operation

•

Threading Operation.

•

Surface Grinding.

•

Heat Treatment Operation.

most practical machining operations use a cutting fluid designed to ameliorate these effects. The primary functions of a cutting fluid are: to decrease friction and wear, reduce temperature generation in the cutting area, wash away the

Table 1: Non-traditional machining processes. Source of energy

Name of processes

Thermal

Electrical discharge machining,

energy

processes

EDM Laser-beam machining, LBM

Temperature In Metal Cutting

Plasma-arc machining, PAM Electrical processes

Although the vast majority of machining operations are conducted with the work piece at ambient temperature,

Electrochemical machining, ECM Electrochemical grinding, ECG

because of the large plastic strain and very high strain rate

Chemical processes

Chemical machining, CHM

there is a significant temperature rise. This has an important

Mechanical

Ultrasonic machining, USM

bearing on the choice of tool materials, their useful life and

processes

on the type of lubricant system required. As a result of the

[7].

rubbing action a secondary deformation zone develops in the Heat Treating of Die Blocks

chip adjacent to the chip-tool interface and this also contributes to heat generation.

Heat treatment of die blocks involves the following steps: 1. Heating of the die blocks slowly to austenitizing

Much data on temperature generation in metal cutting may be correlated with a dimensionless number R t K

is the thermal diffusivity

=

=

temperature.

Kiud, where

2. The dies blocks are soaked at this temperature for an

k/pc, u is the cutting speed and

d is the depth of cut. Rt is called the thermal number. If all of

extended period to achieve uniform austenitic structure.

the heat generated goes into the chip, the adiabatic

3. After soaking, the dies are quenched in a medium to temperatures below the transformation temperature to

temperature is given by

achieve martensitic structure.

fl

4. Tempering is the next stage of heat treatment. Here,

pC

martensite formed as a result of quenching is tempered to a

T ad = -

tougher structure. The stages in heat treating a tool steel die Where [J

=

the specific cutting energy,

p

=

is illustrated below in Figure 27.

the density of

the work piece material, c = specific heat of work piece[5]. Cutting Fluids

439

[13]. Among these factors, only a few can be changed in the heat treatment shop. The selection of optimum quenchant and quenching conditions both from the technological and

A.e

economical point of view is an important consideration [13]. Brine solution, oil, polymer etc. are used as conventional quenching media. Brine solution is restricted to quenching simple shapes and steels of comparatively low hardenability because of the occurrence of intolerable distortion, and quench cracks [14]. Quenchant

Workpiece / COITlpOnent

i.P\lloy grade

ii .[Vlass/weight iil.Geometry iv. Surface rougnness \ j . Surface

Figure 27: Heat treatment cycle of hot working steels (Krauss 1995).

vi·load alTangelllent

Quench Harden iog .-/

"

Iron Atoms

Figure 30: Factors influencing the metallurgical transformation during quench hardening.

Interstitial Table 2: Approximate Soaking Periods for Hardening,

Carbon Atom

Annealing, and Normalizing Steel [9].

Before heat treatment

Thickness

Empty Interstitial Space

of

Metal (mm).

Time of Heating

Soaking

to

(hr)

Required

Time

Temperature (hr). 25.4

0.75

0.5

25.4 - 50.8

1.25

0.5

50.8 - 76.2

1.75

0.75

76.2 - 101.6

2.25

I

After quenching

101.6-127

2.75

I

Figure 28: Harden Iron-Carbon Atomic Structure [10].

127 - 203.2

3.5

1.5

Quenching during heat treatment involves simultaneous

Table 3: Heat treating of die blocks (shock resisting

occurrence of different physical events such as heat transfer,

material)[15].

phase transformation and stress-strain evolution, but heat transfer is the driving physical event as it triggers other processes [11]. When the hot metal is submerged into the liquid pool, heat transfer is controlled by different cooling stages known as vapour blanket stage, boiling stage, nucleate

Rate

Prehe

Hard

Tim

Med

Temp

Soaking

of

at

ening

e of

IUm

ering

Time

Hea

Temp

Temp

Que

at

Temp

ting

eratur

eratur

nchi

Que

eratur

e

e

ng

nchi

e

boiling stage and convective or liquid cooling stage [1-3].

ng

The important factors, which influence the metallurgical

Ver

transformation during quenching, are shown in Figure 30

440

650

0

845

0

5

-

Oil

250 0

60min(t

y

C

C

-

Slo

950 0

wly

C

30

C

7.

emperin

Machining,

Discharge

"Metals

Handbook", pp.227-233

g)

min

Electrical

45min(

(5°C

hardeni

/min

ng)

8.

S. Babu. D. Ribeiro and R. Shivpuri, " Material and Surface Engineering for Precision Forging Dies". NCMS, 1999.

)

9.

J.L. Dossett and H. E. Boyer, "Practical Heat Treating" 2nd Edition, 2006. ASM International Ohio USA.

10. G. Ramesh and N. K. Prabhu, " Review of Thermo-Physical Properties, Wetting and Heat Characteristics Applicability

of III

Nanofluids Industrial

and Quench

Their Heat

Treatment", Spring Open Journal 2011. 11. C. H. Gur and J. Pan," Handbook of Thermal Process Modelling of Steels", New York CRC Press Publication. 2008. 12. H. J. V. Hernandez and B. H. Morales," A novel

Figure 19: Drilling, Milling, Ejection system assembly and Assembled die blocks

probe design to study wetting front kinematics during forced convective quenching", Exp Therm Fluid Sci 2009, pp. 797-807.

References

13. B. Liscic, " State of the art

III

quenching", In

1.

G E Plastics Injection Moulding Design Guidelines

2.

1. H. J. Bast and L. A. Mohammed, " Building

international seminar of the international federation

Integrated CAD/CAM/CMM System

for heat treatment and surface engineering. The

quenching and carburizing proceedings of the 3rd

Institute of Materials, 1993. 3.

P. L. Jain, " Principles of Foundry Technology" 3rd 14. ASM Handbook: In Heat Treating, volume 4

edition, Tata McGraw Hill Publishing Company

materials

Limited, New Delhi. Pg 127-142

park.

ASM

international

handbook

committee,1991. 4.

G. E. Dieter, " Mechanical Metallurgy" 3rd edition, 15. Heat Treating Data Book, lOth edition E-book.

1978

Published 5.

Addison -Wesley

Publishing

Company

SECO/Warwick

corporation.

Inc ... Hi. Bob McClintic, "The do's and don'ts of design" ejournal, Die casting engineer, 2007.

Reading Mass .. , 1966. 6.

by

N. H. Cook, " Manufacturing Analysis", p.47,

R.

K.

Springborn

(ed.),

"

Non-Traditional

Machining Processes", Society of manufacturing Engineers, Dearborn, Mich .. , 1967

441

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

DEVELOPMENT AND RESEARCH OF NEW ALUMINIUM ALLOYS WITH TRANSITION AND RARE-EARTH METALS AND EQUIPMENT FOR PRODUCTION OF WIRE FOR ELECTROTECHNICAL APPLICATIONS BY METHODS OF COMBINED PROCESSING l.Matveeva l , N.Dovzhenk0 2 , S.Sidelnikov2, L.Trifonenkov 3 , V.Baranov 2 , E.Lopatina2 IUC RUSAL, 13/1, Nikoloyamskaya st., Moscow, 109240, Russia 2Siberian Federal University, 79 Svobodny Prospect, Krasnoyarsk, 660041, Russia 3LLC RUSAL ETC, 37 bId. 1 Pogranichnikov St., Krasnoyarsk, 660111, Russia Keywords: wire rod, wire, cable, electrotechnical, conductivity, tensile strength, thermal resistance, alloying, zirconium, rare-earth metals, mishmetal Two groups of aluminium alloys were proposed to till this need in accordance with Russian Federation Patents No 2458151 and No 2458170 [1,2].

Abstract Development of electrical alloys of system aluminium - rare-earth of metals and aluminium-zirconium for production electrotechnical application wire rod. Design of technological line for their manufacturing. The effect of rare-earth and transition metals on the properties of the aluminium alloys containing such metals is analysed. New alloys with different content of zirconium, cerium, and other components featuring enhanced mechanical and electrophysical characteristics have been proposed. New technologies for production of long round-in-section items involving combined processing methods have been developed. The effects of the processing methods on the structure and properties of semifinished products made of new alloys have been studied and recommendations for the modes of preparing alloys, casting, shaping, and thermal processing have been made for the set of the studied alloys. The method of combined casting and drawingextrusion is shown to ensure, in laboratory conditions, improved

Properties of alloys of Al-Zr system The first group includes Al-Zr low alloys because zirconium alloys are the most frequently used kind of thermal resistant alloys worldwide. One characteristic that is unique to this group of alloys is that the content of zirconium and iron in them is 0.100.19 wt.% and 0.21-0.35 wt.%, respectively, while the content of silicon is 0.11-0.15 wt.%. These amounts of alloying elements optimise the mechanical strength of the alloy (ultimate tensile strength of up to 150 MPa) and its electrical resistance (no more than 0.0285 Ohm.m/mm 2 ) while also giving it acceptable thermal resistance. At the same time, it has to be noted that high strength characteristics contributed by these alloying elements are not affected by high temperatures. Features of AI-rare earth metals alloys The second group of alloys are AI-REM (rare earth metals) alloys. These are characterised by high strength (ultimate tensile strength of up to 220 MPa) and thermal resistance. One feature of rare earth metals is that they do not dissolve easily in solid aluminium, which means that they do not significantly reduce the electrical conductivity of aluminium. On the other hand, rare earth metals actively react with aluminium and transition metals forming intermetallic compounds in the eutectic and in highly disperse form. The resultant AillMe3 (AI4Me) intermetallic transition phases, which have a very high melting point (over 1,200°C), ensure consistency of the mechanical and electro-physical properties of the alloys at temperatures of up to 200°C. The specific electrical resistance of these alloys does not exceed 0.0295 Ohm·m/mm2 while the ratio of specific strength to ultimate tensile strength to the density of the wire material is 92.6 N·m-g- l , whereas even the best grade plastically deformed and hardened 1050 aluminium alloy wire have a ratio of no more than 60 N·m.g- I. Analysis of the mechanical and electro-physical properties of these allows has shown that as the content of cerium and lanthanum is increased, the ultimate tensile strength, hardness and specific electric resistance also increase.

mechanical properties and the required level of electric resistivity. Introduction

Recent years saw a noticeable increase in the use of aluminium alloys in electrical engineering. This has been primarily the result of relatively high prices of copper and copper wire rod. Aluminium is, on average, almost 4 times cheaper than copper while aluminium wire rod is also more than 3 times lighter than copper wire rod. The main drawback of technical grade aluminium wire rod made of 1050 and 1070 alloys sold in the market today is that it has relatively low strength (its ultimate tensile strength is 80-110 MPa). Silicon and magnesium are added as alloying elements to aluminium (AVE alloys), which increases tensile strength to 120-130 MPa, but silicon and magnesium significantly reduce the electrical conductivity of the wire. Analysis of published scientific and technical materials has shown that aluminium alloys made with transition group metals and rare earth metals have increased heat resistance because they have a lower diffusion coefficient. In particular, adding minor amounts of zirconium or cerium significantly increases the tensile strength and thermal resistance of aluminium alloys; however, such alloys also have higher electrical resistance. This means that we need to develop alloys that would combine the mechanical and electrophysical properties required by our customers.

Equipment for combined processing Equipment most often used in production of semi-finished aluminium products for the electrical engineering industry is comprised mainly of casting and rolling machines. However, wire

443

rod that can be produced on currently available casting and rolling machines does not have the required mechanical properties for it to be used as wires in power transmission lines. Thus, additional measures still have to be taken to increase the strength of the cable by wrapping it up in various materials, complicating the design of the cable, or by using various cores etc., with all such methods increasing the production costs of finished cables. In addition, the casting and rolling equipment currently available at the production sites can only process aluminium and soft aluminium alloys. The task of processing the new alloys using more energy efficient production processes and equipment can be accomplished by using combined processing methods [3,4], the most efficient of which is the combined casting, rolling and extrusion process (CREP). To implement this method and research the technologies for processing the new alloys, a prototype CREP machine was designed and manufactured [5]. You can see what the new machine looks like in Figure 1.

Table 1 Technical Specifications of the CREP 2 5 Machine Parameter

Value

The initial diameter of the roll, mm

400

The length of the side of the roll, mm

350

Number of spins of the roll: -mInImum - maxImum

1 15

Gear ratio

40

Power of the electric motor, kWt

45

Operating pressure of the hydraulic plant, MPa

200

Capacity, tonnes per hour

2,5

Dimensions, mm

3400 x 2350 x 627

Results of examinations of Al-Zr alloys For the first of of AI·Zr system, technology of the and extrusion process was implemented metal behveen the rolls, which was consistc:ntlly vl'X"''"'U'''''''' in the groove of rotating rolls, pressed out by them and out a matrix in the form of a hotpressedproducl (wire rod) of!) 111m in diameter. In order to check the technojogicalefTectiveness of the made semi-finished products, their drawing on a Chain drawbench without intermediate wascamed oUI, and a wire of 2 mm in according to recommendations 3I1ineWlIullg of thc \vire was done at a fixed holding of 300e C, and then at a temperature m()chani.cal and electrical resistivity of products were measured at each the LFM400 testing machine with a the performed studiesimrcstiQatc influence of technology features for semi-finished products with combined extrusion method on mechanical and ele:ctr'OpIlY:;lclal Prof)cl'1ties of of Al-Zr system alloys, in 1-0.3 wt % was chosen as a

Table 2.

of the deformed semi-finished products of Al_. Yo raoy o i"O/Z n

Mechanic.::!l properties of wire rod, 09 mm

Figure 1. Industrial Prototype of a CREP Machine. CREP-2.5 machine is a full scale prototype of an experimental combined casting, rolling and extrusion processing machine for producing electro-technical aluminium wire rod and is part of a combined processing line. The technical specifications of the CREP 2.5 machine are presented in Table 1.

UIS, MPa

A research was carried out into producing electrotechnical wire rod from the new aluminium alloys using the combined casting, rolling and extrusion method and the newly developed equipment. Some of the unique findings of the research are presented below.

El., %

121

21

106

29

I properties of wire 0 2 mm in strained and annealed state UIS, El., MPa % 194 ,2 73 36 212 80

,2 34

Microhardness of wire rod and wire

Electrical resistivity of wIre

kg£'mmL

Ohm·mm2/m

35.9±0.7 41.9±1.4

0.0284 0.0275

35.9±0.7 41.9±1.4

0.0294 0.0282

Metallographic studies showed that the structure of all samples made with combined casting, rolling and extrusion method is characterized by non-uniform distribution of phases across a section of aluminum solid solution (Figure 2), and also more rough conglomerations of iron-containing particles and zirconium aluminides are observed here. In addition, small particles of Al3Zr

444

phase which are extended along the deformation direction are found in the samples. Precipitation of some quantity of Al3Zr inclusions is apparently connected with technology features used for melt preparation and its pouring between the rolls. Cold drawing when making a wire leads to crushing of Al3Zr particles, therefore they form lines of small particles similar to a roundish form whereas in the samples of a wire rod made with combined casting, rolling and extrusion method, the majority of particles has a needle form. Unevenness of structural components distribution is inherited by the structure of a wire as well.

Uti

120 100

80 60 40

20 0 0,00

0,05

Figure 3. Dependence of deformation degree on yield strength for samples made with combined casting, rolling and extrusion method from Al - O.IS%Zr alloy (pouring temperature is no°C), at different test temperatures: 1 - 20°C; 2 - 200°C; 3 - 3S0°C A rather uncommon shape of the graphs means that when the temperature is increased at the initial stage of deformation, it barely affects the tension 0.2% YS and more significant effects can only be observed at later stages. This can be put down to the very specific way in which the structure of the rods forms during accelerated crystallisation and deformation of the metal being processed using the combined casting, rolling and extrusion method. Earlier studies [9] allowed us to establish that using combined casting, rolling and extrusion ensures the formation of a stable sub-granular structure that is related to the temperature at which recrystallization of the metal begins. The experimental research that we carried out demonstrated that 99.7% grade aluminium extrusions produced using the CREP method begin and stop recrystallising at the temperatures ofTrb = 290°C and Trs = 3S0°C respectively, which is 40-70°C higher than the recrystallization temperature of wire rod produced using the traditional process of making rolled bars and rods. This is an indication of a more stable sub-granular structure of extrusions produced using the combined casting, rolling and extrusion process. The same effect can probably be achieved by using combined processing of Al-Zr aluminium alloys.

a

b

Figure 2. - Microstructure of wire rod (a) and wire (b) ofa test sample of Al-O.lS%Zr, made with combined casting, rolling and extrusion process, xSOO In order to evaluate properties of wire rod of aluminium alloy containing O.IS % zirconium, received at a pouring temperature of no°c, high-temperature tests were performed at the LFM400 universal testing machine with a tensile load of 400kN. Test results are shown in Figure 3.

Assessment of the strength properties when the test temperature is increased from 20°C to 3S0°C is shown in Figure 3. It should be noted that the tendency for the temporary tensile strength to fall by 3-7% is present regardless of the casting method (the melt temperature and the temperature at which the melt is poured between the rolls of the casting, rolling and extrusion machine) used to make wire rod from O.IS% zirconium alloy (Figure 4).

445

OS. )a

120 liS

110

lOG

20°C

e

350°C

.1.-_--1_ _ _-'-_ _ _' - -_ _-1.._ _

b

Figure 4. Tensile strength properties of wire rod made from AI-0.15Zr alloy using the combined casting. rolling and extrusion method at different test temperatures The tests also demonstrated that as the zirconium content in the alloy increases from 0.15 to 0.25%, the temporal tensile strength increases by 5-7% reaching UTS= 125-130 MPa, but the electric conductivity of the material decreases at the same time. Results of examinations of AI-REM alloys For the second group of aluminium alloys with a content of rare earth metals of up to 5.0% of the mass we researched the effect the temperature and the rate of deformation have on the structure and properties of wire rod produced using the combined casting, rolling and extrusion process at melt temperatures of 750°C and 780°C and deformation rates of s= 0.74 S-I 11 S=1.49 S-I [10]. The structure of deformed semi-finished samples is shown in Figure 5. Deformation at 750°C at different rates causes the formation of a structure that varies greatly between the middle of the wire rod and its edges. Along the edges small particles are evenly distributed over the solid solution. In the middle, there are lighter areas a of the solid solution and darker areas of eutectic (a+AI4Mm), where Mm represents a mishmetal comprising cerium, lanthanum, praseodymium and other metals. At 780°C the lack of uniformity of structure in various layers of the metal is less pronounced. In the samples of 5 mm in diameter there are large lumpy particles of 9x3 flm to 15x7 flm in size. This probably has to do with the high rate at which the metal cools off as it crystallises in the rolls and the high degrees of deformation that the metal undergoes as it is extruded through the calibration hole of the matrix. The optimal metal structure is achieved at 780°C and a high rate of deformation of 1.49 S-I.

c

d Figure 5. The microstructure of wire rod of9 mm in diameter (a,b) and 5 mm in diameter (c,d) produced using combined casting, rolling and extrusion process at temperatures of 750°C (a,c) and 780°C (b,d) at a deformation rate of 1.49 S-I x 500 Table 3. Mechanical Properties of Wire Rod made of AI-REM All oys

The mechanical properties of wire rod with a diameter of 5 mrn, 7 mm and 9 mm produced using the combined casting, rolling and extrusion process at different deformation rates and at melt temperatures of 750°C and 780°C are presented in Table 3.

T=750 Parameters

s= 0,74 sI

s= 1,49 sI

°e

T=780

°e

d=9 mm

d=7 mm

d=5 mm

d=9 mm

d=7 mm

d=5 mm

UTS, MPa

212,3 7

237, 92

253. 27

212, 78

232, 55

241, 92

El.. %

13.34

13,4

12,8 2

UTS, MPa

198,5 8

13,5 3 191, 56

218, 45

11,8 6 231, 72

El.,%

10,2

11.7 2

10,6 2

10.2 9

9 226, 01 10.5 2

246, 78 9,64

12,8

As the table shows, the ultimate tensile strength of wire rod samples made using the combined casting, rolling and extrusion

a

446

9. E.Sidelnikova, A. Klimko, V. Biront et al. Studying the temperature at which aluminium wire rod produced using the combined casting, rolling and extrusion process begins to crystallise// Materials Science and Modern Technologies: Interregional Collection of Research Studies, edited by Yu. Balandin. Magnitogorsk, 2002. Part II, pages 15-18. 10. S. Sidelnikov, N. Dovzhenko, D. Voroshilov etc. Studying the structure of metal and assessing the properties of samples made from AI-REM alloys to determine whether they can be used in the production of electrical conductors using casting and extrusion. Herald of G. Nosov Magintogorsk State Technical University- Magnitogorsk 2011. No 2, pages 23-28.

process is on average between 190 and 250 MPa, while the relative elongation is 9-l4%. Values of electrical resistivity were within 0,0294-0,0310 Ohm-mm2/m.

Conclusion I.New aluminium alloys with transition and rare earth metals were developed that have better mechanical and operational characteristics and the correlation between various modes of combined processing and the properties and structure of wire rod made from these alloys was studied. 2.A prototype combined casting, rolling and extrusion machine was built for making elongated products with improved strength properties from the new aluminium alloys. 3.Aluminium alloys with 0.1-0.3% zirconium content can be used for making electrical conductors, while the temporal tensile strength of hot-deformed semi-finished products made by using the combined casting, rolling and extrusion method reaches up to 120-130 MPa, while the tensile strength of cold extrusions deformed to a degree of 95-98% reaches 200-210 MPa, in the meantime their specific electrical resistance is within 0.02850.02950hm·mm2/m. 4.Wire rod produced by using combined processing from aluminium alloys with rare earth metals has high tensile strength (up to 200-250 MPa in a state of hot deformation) and can be used for production of electrical conductors to be operated at increased temperatures. 5.The estimated mechanical properties of the new alloys and the way they change depending on the processing parameters were used in the design of prototype combined processing equipment and in the development of production processes for making elongated electrical engineering products. References 1. Patent 2458151 Russia, MPK 7 S21C 1/02. Aluminium Alloy.

N. Baranov et al. (Russia). - published August 10th 2012, Bulletin No 22. 2. Patent 2458170 Russia, MPK 7 SS22SC 1/02. Aluminium Alloy. N. Baranov et al. (Russia). - published August 10th 2012, Bulletin No 22. 3. S. Sidelnikov, N. Dovzhenko, N. Zagirov. Combined and Mixed Methods for Processing Non-ferrous Metals and Alloys, a monograph// M.:MAKS Press, 2005, - 344 pages. 4. S. Belyaev, N. Dovzhenko, S. Sidelnikov et al. Increasing the Efficiency of Producing Extrusions from Aluminium Alloys by Controlling the Temperature of the Extrusion Process. Siberian Federal University Journal, No 4, 2009, pages 418-426. 5. Patent 2457914 Russia. MPK 7 V2lS 3/00. B22D 11100. Continuous Casting, Rolling and Extrusion Machine for Nonferrous Metals and Alloys. / V. Baranov et al. (Russia). published August 10th 2012, Bulletin No 22. 6. N. Belov, A. Alabin Prospective Aluminium Alloys with Zirconium and Scandium/ Non-Ferrous Metals. 2007. No 2. 7. N.Belov. Impact of Annealing on the Structure and Mechanical Properties of Cold Rolled Sheets Made from Al-Zr Alloys/ N. Belov, A. Alabin, V. lstomin - E. Kastrovsky, E. Stepanova// Non-Ferrous Metallurgy - 2006. No 2. 8. S. Sidelnikov, N. Dovzhenko, L. Trifonenkov etc. Studying the structure of metal and assessing the properties of samples made from Al-Zr alloys to determine whether they can be used in the production of electrical conductors using casting and extrusion. Herald of G. Nosov Magintogorsk State Technical UniversityMagnitogorsk 2012. No 1, pages 51-55.

447

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

INFLUENCE OF MACHINING PARAMETERS ON AL-4.SCu-TiC IN-SITU METAL MATRIX COMPOSITES Pradeep Kumar .Tha*, Anand Kumar, M M Mahapatra Department of Mechanical and Industrial Engineering Department Indian Institute of Technology Roorkee, Uttarakhand- 247 667, India *Email: [email protected] Keywords: In-situ MMCs, Turning, Cutting force, Surface roughness, ANOV A therefore, a need to investigate the machninability of in-situ TiC reinforced composite materials. N. Muthukrishanan et at. [6] reported in their studies on continuous turning of A356/SiCp/lOp composite by medium grad polycrystalline diamond (PCD 1500) inserts at various cutting condition. The response parameters of machining such as specific power consumed, surface roughness, and tool wear were considered. It was concluded that the formation of built up edge (BUE) and tool wear was less at lower cutting speed. At higher cutting speed, less consumption of specific power and less tool wear was observed. Quan Yarning et at. [7] had investigated the tool wear in machining of SiC reinforced aluminium based metal matrix composites. The material structures and tool wear mechanism were investigated. They stated that the tool life was strongly depended on the volume fraction and size of reinforcement in composites materials. Hooper et al. [8] had studied machinability of aluminium metal matrix composite reinforced with SiC particles with polycrystalline diamond (PCD) and conventional tungsten carbide (WB) tools. It was concluded that the PCD tool was more suitable for the MMCs as compared to WB tools. X Ding et al. [9] also reported on the superior performance of polycrystalline diamond (PCD) tools in terms tool life over poly crystalline boron carbides (PCBN). It was observed that the formation of BUE tendency was more in PCBN tool as compared PCD tools. L.T.Lin et al. [lO] had studied the machining of AI356/SiCI20p with poly crystalline diamond (PCD) tools under various cutting speed, feed rate and at constant depth of cut. The tool life data had been analyzed using regression and general form of the Taylor's equation was developed to describe the tool performance for the AI356/SiCI20p composite. The time required to reach the tool wear limit decreased with increases of speed and feed. However, the volume of material removed before reaching the wear limit increased with the higher feed rate. A Manna et al. [II] investigated the machining of AI/SiCI20p composite material with uncoated tungsten carbide (WB) K-I 0 type tools. Analysis of variance was employed to investigate the influence of cutting speed, feed and depth of cut on surface roughness. It was observed that the surface roughness affected by cutting speed, feed rates and depth of cut, while the feed rates and depth of cut affected the maximum peak-to-valley roughness height. M Seeman et at. [12] studied the machinability of AI/SiCI20p composite material with uncoated carbide tool (K-lO) inserts. Response Surface Methodology (RSM) was employed to investigate the machining parameters including cutting speed, feed rates, depth of cut and machining time on the basis of performance characteristics such as tool flank wear and surface roughness. K. Palanikumar [13] developed a model for surface roughness prediction through response surface method (RSM) for machining of glass fiber reinforced (GFRP) composites. Four factors, five level, central composite rotatable design

Abstract With the advent of number of composite materials, a systematic study of machining characteristics of these new materials is necessary for their rapid adoption in the actual engineering applications. Al-Cu-TiC metal matrix composite is widely used in aeronautical and automobile industries due to their excellent mechanical and physical properties. However machining of these composites is difficult because of the harder reinforcement particles. This paper presents an experimental investigation on the machinability bahavior of AI-4.5Cu-TiC insitu cast metal matrix composite reinforced with weight percentage of 10% of Titanium Carbide. The experimental studies were conducted under varying process paranleters e.g. cutting speed, feed rate and depth of cut. The optimization of machining parameters was done by designing a full factorial matrix using Taguchi method. The analysis of variance (ANOV A) is employed to investigate the influence of used parameters on surface roughness Ra.

Introduction A number of processing routes have been developed for production of aluminium based metal matrix composites like powder metallurgy, liquid metal infiltration, compocasting, squeeze casting method, stir casting and spray decomposition method etc. Among these, in-situ stir cast route offers more advantages when compared to others conventional routes. Small particles size (lJ.lm to 3J.lm), greater bonding, high specific strength, high specific stiffness, better wear resistance, good thermodynamic stability, and uniform distribution are the advantages of produced in-situ composite materials. Clustering and improper wetting of reinforced particles can be avoided by using in-situ stir casting method [1-3]. The major typical application of in-situ composite materials serves in the aerospace and automotive sectors include piston, engine block, brake component (discs and rotors), valves, liners, drive shaft [4]. The component produced by composite materials requires a machining process to achieve a required shape with dimensions with requisite tolerance. The machining of composite materials is associated with interaction of hard ceramic reinforced particles with the tool. These characteristics lead to difficulties in machining of composite materials. The cutting force, tool wear, surface finishes are the important indices to assessing the machinability behavior. Hence, the selection of proper cutting tool and machining condition is essential to produce a better finished metal matrix composites product. From the literature review it is observed that much works have been done on machining and machinabilty of ex-situ SiC reinforced composites, however reported works related to machining of insitu TiC reinforced composite materials are limited [5]. There is,

449

matrix was employed to carry out the experimental investigation. Analysis of variance was also used to check the validity of the model. E.L. Gallab et al. [14] had studied PCD tool performance during high-speed turning of AI/SiC/20p MMC and reported that PCD tools suffered excessive edge chipping and crater wear during the machining of the MMC. From the literature it is observed that the machining of Al based MMC is an important research area but only very few research have been carried out on machining of in-situ TiC reinforced composite materials.

Mathematical modeling of experimental data Machining experiment was conducted using the full factorial design matrix as given in Table I. Twenty seven experiments with replication was done and the average experimental data for the full factorial experiment is given in Table II. The responses of the experiment such as cutting force and surface roughness values for the twenty seven experiments are also presented in Table II.

Experiment Details

Table 1. Process parameters (control factors) for turning operation

In-situ as-cast AI-4.5%Cu-TiC metal matrix composites were fabricated with weight percentage of 10 by stir casting method. The cylindrical bar specimen of 45 mm diameter and 300 mm length were turned on NH-22 self centered three jaw chuck based lathe made by Hindustan Machine Tools (HMT). The uncoated cemented carbide inserts were used for turning of AI4.5%Cu-l0%TiC composites materials. The specification of cemented carbide inserts was ISO coding CNMG 120408 grade HI3A and tool holder specification of ISO DCLNR 2525M12. The cemented carbide inserts were rigidly mounted on tool holder. The angles set with inserts and tool holder were at rake angle of - 6°, clearance angle of 5°, negative cutting edge inclination angle of - 6°, approach angle of 95°, and nose radius of 0.8 mm respectively. A Kistler Pizoelectric Dynamometer Kistler (Type 9257B) loaded with multi charge amplifier of Kistler (Type 5070) was connected with the tool holder. Data acquisition of machnining was carried out by appropriate software (DynaWare Kistler type 2825A-02).The surface roughness (Ra) was measured by Veeco optical profiling system (WYKO NTlIOO) instrument with the scan length of 50 flm. Data acquisitions system for machined surface was carried out by Vision 32 software. The microstructure of AI-4.5% Cu/IO% TiC cast composite material is shown in Figure2. The full factorial machining parameters (control factors) and their levels are presented in Table 1.

Sl. No.

Parameters

Unit

Level I

Level 2

Level 3

1

Cutting Speed

mm/min

40

80

120

2

Feed

mm/rev

0.12

0.24

0.36

3

Depth of Cut

mm

0.5

0.75

1.00

Table II. Turning conditions and machining responses of MMC Sl. No. 1 2 3 4 5 6 7 8 9 10 11 12 13

14 15 16 17 18 19 20 21 22 23 24 25 26 27

Figure 1. Experimental setup of turning operation

Figure 2. SEM micrograph of AI-4.5%Cu/IO%TiC composite material

450

Cutting Speed (m/min) I I I I I I I I I 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3

Feed (mm/rev.) I I 1 2 2 2 3 3 3 I 1 I 2 2 2 3 3 3 I I I 2 2 2 3 3 3

Depth of Cut (mm) I 2 3 1 2 3 1 2 3 I 2 3 1 2 3 1 2 3 I 2 3 I 2 3 1 2 3

Cutting force, Fz (N) 47.22 84.68 119.35 108.82 149.75 189.67 182.35 224.56 256.85 45.67 87.59 126.56 119.62 153.55 184.22 191.81 226.34 267.65 42.04 75.97 116.89 107.68 140.09 177.66 169.87 208.54 322.39

Surface roughness Ra(flm) 2.670 3.140 3.590 3.094 3.571 4.023 3.510 3.890 4.340 1.960 2.460 2.935 2.391 2.780 3.120 2.560 2.970 3.699 1.380 1.792 2.296 1.895 2.260 2.734 2.430 2.934 3.294

The full factorial experimental data was used for building the mathematical model for the turning operation of Al4.5%Cu/lO%TiC composite. Analysis of Variance (ANOVA) was used to investigate the effect of the control factors on the responses such as cutting force and resulting surface roughness. ANOV A is a method of apportioning variability of an output to various inputs. Table Ill& IV shows the results of ANOV A analysis from MINITAB 13.1 software [15]. The purpose of the analysis of variance was to investigate which machining parameters that significantly affect the performance characteristics such as cutting force and resulting surface roughness. Initially the experimental data was analyzed using the ANOVA. Table TIT and Table IV, shows the ANOV A for responses such as cutting force, Fz and surface roughness, Ra respectively. From Table III the significance of control factors such as cutting speed (V), feed (F) and depth of cut (DOC) and their interactions were summarized in terms of "P" value. As indicated from Table ITT. the "P" values of feed (F) and depth of cut (DOC) are less than 0.05 indicating significance in the regression relation. The "P" value for other terms in Table TIT such as for cutting speed and interactions is more than 0.05, indicating their non-significance in the regression relation. This is also exhibited in the main effect plots for cutting force, Fz in Figure Ill.

-116 -O.0029*cutting speed+604* feed + 166*DOC

Fz

b

(1)

$I0T"""""-------y;

no

101! 200 Measufl1d \'allw

250

300

Figure 4. Measured and predicted value of cutting force Fz (N) from the regression equation: (a) For the full factorial experimental data; (b) For the test cases Measured and predicted value of cutting force Fz (N) from the regression equation(l) for the full factorial experimental data given in Table II and for the test cases are plotted in Figure IV. It can be observed from Figure IV that there is close agreement between the measured and predicted values of cutting forces for the experimental and test cases. Table IV. Analysis of Variance for surface roughness, Ra

Table TIT. Analysis of Variance for cutting force, Fz

Source

DF

Ad'.I SS 6.67

Ad'.I MS 3.37

F

P

480.47

0.000

Source

DF

Seq SS

Adj SS

AdjMS

F

P

V

2

Seq SS 6.67

F

2

94857.6

94857.6

47428.8

204.9

0.00

F

2

3.05

3.04

1.53

219.31

0.000

DOC

2

31032.7

31032.7

5516.4

67.04

0.00

DOC

2

3.68

3.68

1.84

265.26

0.000

V

2

123.7

123.7

61.8

0.27

0.77

V*F

4

0.16

0.16

0.04

5.69

0.018

F*DOC

4

932.2

932.2

233.1

1.01

0.46

V*DOC

4

0.01

0.01

0.002

0.22

0.921

F*V

4

395.2

395.2

98.8

0.43

0.79

F*DOC

4

0.01

0.01

0.003

0.48

0.751

DOC*V

4

926.6

926.6

231.7

1.00

0.46

Error

8

0.06

0.06

0.007

Error

8

1851.7

1851.7

231.5

Total

26

13.63

Total

26

130119.8

The data of the surface roughness obtained from the machining of AI-4.5%Cu/l0%TiC composite for the twenty seven experiments together with the input control factors presented in Table II were also investigated for the interaction effects of the process variables and ANOVA results are given in Table IV. The significance of control factors such as cutting speed (V), feed (F) and depth of cut (DOC) and their interactions were summarized in terms of "P" value in Table IV. The "P" values of cutting speed (V), feed (F) and depth of cut (DOC) are less than 0.05. The "P" value for other terms in Table IV such as for the interactions is more than 0.05, indicating the non-significance interaction. The main effects of the process variables for surface roughness are presented in Figure 5. From Figure 5 it can be inferred that with increasing the cutting speed the surface roughness of AI4.5%Cu/lO%TiC composite decreases. For example at cutting speed of 120 m/min a surface roughness value of 2.4 Jlm is achieved. The increasing depth of cut and feed rate have detrimental effects on the surface roughness.

Cultng speed (mfmin)

Figure 3. Main effects plot data means for cutting force Fz (N) Based on the ANOVA Table 3 and the main effect plot for the cutting force, Fz in Figure 3, the regression equation is developed for predicting the cutting force and given in equation 1.

451

Cutting speed (m/min) Feed mm!

experimental and regression equation predicted values of cutting force and surface roughness.

Depth of cui (mm)

Reference

/ I.

2.

Figure 5. Main effects plot data means for surface roughness, Ra

3.

Based on the ANOV A data and interaction effects of the process variables, the regression equation to predict the surface roughness of AI-4.5%Cu/lO%TiC composite is stated in equation (2).

4.

Ru

=

1.90-0.0150*cutting speed +3.43* feed + 1.81 * DOC

5.

(2)

a 4S ,...--------"71 S 4.11

1

3.5

l,'!:U

r'

l:U

iu 2.1)

\\l.MI

15 lAl

6.

O

L() 1.5 Ul

3A) 35 4.0 4.5 •.lmud \·"IIlI.

1.5

3.5

4:0

Figure 6. Measured and predicted value of surface roughness, Ra (J.lm) from the regression equation: (a) For the full factorial experimental data; (b) For the test cases

7.

8.

The regression equation (2) was also tested for a number of test cases, the parameters of which are excluding those of given in Table II. The measured and predicted value of surface roughness, Ra(J.lm) from the regression equation for the full factorial experimental data and for the test cases are shown in Figure 6. It can be observed from Figure 6 that there is close agreement between the experimental and predicted values of surface roughness, indicating the adequacy of the regression model.

9. 10.

II.

Conclusion From the machining experiment it is observed that the in-situ AI4.5%Cu/lO%TiC composite is machinable with the use of cemented carbide inserts with appreciable surface roughness and tolerable cutting force. The machining force generated did not lead to detrimental chattering and vibration of the machine tool. It was also observed that the cutting force was directly proportional to feed and depth of cut. The cutting speed had marginal effect on the cutting force. However, the cutting speed was having significant effect on the surface roughness of in-situ AI4.5%Cu/lO%TiC composite. The length of removal chips from composite material was longer at higher cutting speed and low depth of cut. Feed and depth cut were having similar effect on both cutting force and surface roughness. The full factorial experimental data was utilized in developing the mathematical model for predicting the machining responses such as cutting force and surface roughness. The regression equations were developed based on the significance of control factors and interactions. Close agreement was observed between the

12.

13.

14.

M.Gupta and M.K. Surappa, "Processing-MicrostructureMechanical Properties of Al Based Metal Matrix Composites Synthesized Using Casting Route", Key Engineering Materials, Part I, 259 (1995),104-107. S.Prabu Balasivanandha , L. Karunamoorthy , S. Kathiresan , B. Mohan. "Influence of Stirring Speed and Stirring Time on Distribution of Particles in Cast Metal Matrix Composite". J. of Materials Processing Technology, 171 (2006), 268-273. Jasmi Hashim,"The Production of the Cast Metal Matrix Composites by a Method of Stir Casting Method", Journal Teknolgi, Vol. 35(A) (2001), 9-20. D.B. Miracle"Metal Matrix Composites-From Science and Technological Significance". Composites Science and Technology, 65 (2005), 26-2540. A.Mahamani," Machinability Study of AI-5Cu-TiB2 In-situ Metal Matrix Composites Fabricated by Flux-assisted Journal of Minerals & Materials Synthesis", Characterization & Engineering, Vol. 10, No. 13(201 1),1243-1254. N. Muthukrishnan & M. Murugan & K. Prahlada Rao "Machinability Issues in Turning of AI-SiC (lOp) Metal Matrix Composites". int .J Adv A1anuf Techno/ , 39(2008), 211218 Quan Yanming, Zhou Zehua, "Tool Wear and its Mechanism for Cutting SiC Particle-Reinforced Aluminium Matrix Composites". Journal of Materials Processing Technology, 100 (2000), 194-199. R.M. Hooper, J.L. Henshall, A. Llopfer, 'The Wear of Polycrystalline Diamond Tools Used in the Cutting of Metal Matrix Composites". Int J Refract Met Hard mater 17( 1999), 103-109. X. Ding, W.Y.H. Liew, X.D. Liu "Evaluation of Machining Performance of MMC with PCBN and PCD Tools". Wear 259 (2005) 1225-1234. J.T. Lin, D. Bhattacharyya, and C. Lane "Machinability of a Silicon Carbide Reinforced Aluminium Metal Matrix Composite". Wear 181-183 (1995) 883-888. A. Manna, B. Bhattacharyya, "Investigation for Optimal Parametric Combination for Achieving Better Surface Finish during Turning of Al ISiC-MMe". Int J Adv Manu! Technol 23(2004),658--665. M. Seeman & G. Ganesan & R. & A. Vclayudham "Study on Tool Wear and Surface Roughness in Machining of Particulate Aluminum Metal Matrix Composite-Response Surface Methodology Approach". Int J Adv 48(2010), 613-624. K.Palanikumar, "Modeling and Analysis for Surface Roughness in Machining Glass Fibre Reinforced Plastics using Response Surface Methodology". Mater Des 28(2007), 2611-2618. EI-Gallab M, M.Sklad, "Machining of AI/SiC Particulate Metal-Matrix Composites part 1 & 11", J Mater Process Techilo183(1998), (151-158):277-285.

15. Minitab Inc User manual of MINITAB Statistical Software. Release 13.31 (2000) State College PA 16801 USA.

452

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

EFFECT OF Mg CONTENTS ON FLUIDITY OF AI-xMg ALLOYS Nam-SeokKim, Seong-Ho Ha, Young-Ok Yoon, Gil-Yong Yeom, Hyun Kyu Lim, Shae K. Kim KITECH(Korea Institute ofTndustrial Technology); 156, Gaetbeol-ro, Yeonsu-gu; Incheon, 406-840, Korea South Keywords: AI-Mg system, Fluidity, Molten state, Liquidus temperature Abstract

determined with adding 40°C to liquidus temperature obtained with JMatPro. This is to make the same condition in superheat, which is an factor to affect the fluidity [15]. Drawing and photograph of fluidity mold with spiral type are shown in Fig. 1. Fluidity test was carried out after the holding for 10 minutes. The determined pouring temperatures are summarized in Table 1. The mold temperature was held to 250°C being controlled by the inserted heating bars. The flow lengths with the gain of the Mg content were measured with the samples obtained after the pouring into the mold. The microstructure observation by OM (optical microscopy) were carried out about the cross section at I cm distance from the fore end. The samples were grinded and

This study is focused on a basic approach for the fluidity of the AI-Mg binary system. The objective ofthis study is to investigate the fluidity change of the AI-Mg alloys with increasing Mg content. As a result of fluidity test, pure Al showed the highest value in all the examined alloys. With 2.5%Mg addition, it decreased rapidly. On the other hand, the gain of the flow length was shown with increasing Mg content from 5%Mg addition. The change of the fluidity shown in this study is roughly similar to that reported previously. However, they also showed the difference in the Mg content which has the most viscous fluidity. It is considered that it is attributed to the different experimental conditions. In microstructures, with increasing the Mg content, the dendritic a-AI was developed and the existence of precipitation (MgsAIs) is shown in the grain boundary. regarded as This tendency became conspicuous from the 5wt%Mg addition. And then. the grains were refined with the formation of the precipitation.

polished, finally with 1 /Jm. And also, the samples were cleaned using pure alcohol and etched with keller's etchant.

Introduction AI-Mg based alloy has attracted considerable attention because of its superiority in mechanical properties, corrosion resistance, weldability, as well as low density. With this promising potential, the demand for AI-Mg based alloy in vehicle and aircraft industries has increased [1-3]. On the other hand, further improvement of strength is required for a practical performance on a par with steel. The reinforcement by Rib had been considered for additional strengthening of the alloy [4, 5]. The addition of Rib, however, leads to the complication of product shape, causing a defect in mold filling. Therefore, material design for good fluidity is an important task in Al alloy process. Oxide layer and inclusion formed in melting process could be one of factors to reduce melt fluidity [6]. It is well-known that the Mg in AI-Mg system causes the formation of the oxide inclusion in melt due to its affinity to oxygen. The oxide clusters affect melt fluidity and mechanical properties of the final product [7, 8]. To avoid such a problem, the use of S02 gas and Be addition have been considered [9, 10]. Due to the limitation in utilizing S02 and toxicity of Be, the eco-friendly alternative measure is required [2, 11-13]. In an earlier study, it was reported that the oxidation resistance of the molten AI-Mg alloy was improved by the addition of Mg in the form of the master alloy including AI 2Ca [14]. The present study focuses on a basic approach for the fluidity of the AI-Mg system. The objective of this study is to investigate the fluidity change of the AI-Mg alloys with increasing Mg content. Experimental procedure Alloys were prepared by electric resistance furnace with 2.5 to 10wt%Mg addition. Pouring temperature of each alloy was

Figure 1. Drawing and photograph of fluidity mold.

453

Table 1. Liquidus and solidus temperatures with Mg contents Mg contents

Liquidus

Solidus

CC)

("C)

Pouring temperature (liquidus + 40, "C)

660.4

660.4

0

700.0

2.S

648.1

623.3

24.8

690.0

S.O

63S.7

S83.9

S1.8

67S.0

10.0

609.9

S12.2

97.7

6S0.0

(%)

CC)

0

AT

700

Results and Discussion Figure 2 and Table I show liquidus and solidus temperatures calculated from JMatPro. Pure Al showed 660.4 DC, close to the melting temperature of Al reported generally. With increasing Mg content, the liquidus temperature decreased to 609.9 DC, about SO DC lower than that of pure AI. As shown in Figure 2, the liquidus temperature linearly decreased with increasing the Mg content. The results of fluidity test are shown in Table IT and Figure 3. The results of each alloy have low standard deviation. Pure Al showed the flow length of 44.8 mm, the highest value in the examined alloys. This arises from the different solidification mode with that of alloy [IS]. Pure metal solidifies with planar front, while alloy does with mushy front [IS]. With 2.S%Mg addition, it decreased rapidly, to 40.4mm. And then, the gain of the flow length was shown with increasing Mg content. From the L'.. T values in Table T, it was confirmed that the mushy zone was extended with increasing the Mg content. Therefore, it was considered that the extension of the mushy zone led to the fluidity improvement. According to the previous report [IS], fluidity decreased significantly in the solid solution portion of the system up to 4.9%Mg and increased as the fraction of eutectic formed. And then, the fluidity increase was reached to its peak in the hypereutectic region. The change of the fluidity shown in this study is roughly similar to that explained above [IS]. It is considered that the difference in the Mg content with the most viscous fluidity is attributed to the different experimental conditions. Figure 4 shows the microstructures of the fluidity tested samples. In the case of pure AI, it is thought that the existence of grains was not clear. With increasing the Mg content, the dendritic a-AI solid solution was developed and the existence of precipitation regarded as (MgsAIg) was shown in the grain boundary. This tendency became conspicuous from the Swt%Mg addition. And then, the grains were reiined with the formation of the precipitation. Based on the AI-Mg phase diagram [16], from the 2.Swt%Mg, the solid solution would be supersaturated with Mg solute atoms, because the Mg content is higher than 1. 9wt%Mg, which is the solubility of Mg in Al at room temperature [17]. As the solidification proceeded, the precipitation, an equilibrium (MgsAIs) would be formed as along the grain boundaries. And then, the grain refinement would be occur by the pinning effect.

:2

4

6

8

'10

Mgcontents (%J

Figure 2. Liquidus and solidus temperatures with Mg contents. Table II Results of fluidity test and standard deviation Fluidity Alloys Standard deviation (cm) 1. Pure Al

44.8

1.7

2. AI-2.SMg

40.4

0.8

3. AI-S.OMg

41.8

0.6

4. AI-IO.0Mg

43.0

1.8

-e

u

"'3 u:::

46

40

246

e

Mgcontents (%) Figure 3. Result of fluidity test for AI-Mg alloys.

454

10

(a)

(b)

(c)

(d)

Figure 4. Microstructures of (a) pure AI, (b) AI-2.5Mg, (c) AI-5Mg and (d) AI-IOMg after fluidity test.

According to the previous report [15], the grain refinement can also have a positive effect on fluidity. Therefore, it was thought that the grain refinement also contributed to the increase of the fluidity.

4. D. Shan et a!., "Research on Local Loading Method for an Aluminium-Alloy Hatch with Cross Ribs and Thin Webs", Journal of Materials Processing Technology, 187-188 (2007), 480--485. 5. B.S. Kang, 1.H. Lee, and B.M. Kim, "Process Design in Flashless Forging of Rib/Web-Shaped Plain-Strain Components by Finite Element Method", Journal of Materials Processing Technology, 47 (3) (1995), 291-309. 6. C.R . Loper Jr., "Fluidity of Aluminum-Silicon Casting Alloys", AFS Trans., (1992), 533-538. 7. C.N. Cochran, D.L. Belitskus, and D.L. Kinosz, "Oxidation of Aluminum-Magnesium Melts in Air, Oxygen, Flue Gas, and Carbon Dioxide", Metallurgical Transcation B, 8B (1) (1977), 323-332. 8. Y.D. Kwon, and Z.H. Lee, 'The Effect of Grain Refining and Oxide Inclusion on the Fluidity of AI-4.5Cu-0.6Mn and A356 Alloys", Materials Science and Engineering: A, 360 (1-2) (2003), 372-376. 9. C. Houska, "Beryllium in Aluminum and Magnesium Alloys", Metals and Materials, 4 (2) (1988), 100-104. 10. L.F. Mondolfo, Aluminum Alloys: Structure & Properties (London, Butterworth & Co. Ltd., 1976), 11. D.L. Belitskus, "Oxidation of Molten AI-Mg Alloy in Air, Air-S02' and Air-H2 S Atmospheres", Oxidation of Metals, 3 (4) (1971),313-317. 12. Y. Wang, and Y. Xiong, "Effects of Beryllium in AI-Si-MgTi Cast Alloy", Materials Science and Engineering: A, 280 (1) (2000), 124-127. 13. David H. DeYoung, and 1. Peace, "Beryllium in Dross Produced during Aluminum Melting", Light Metals 2009, (2009), 659-664. 14. lK. LEE, and S.K. KIM, "Effect of CaO Composition on Oxidation and Burning Behaviors of AM50 Mg Alloy",

Conclusions As a result of fluidity test, pure Al showed the highest value in all the examined alloys. With 2.5%Mg addition, it decreased rapidly. On the other hand, the gain of the flow length was shown with increasing Mg content from 5%Mg addition. The change of the fluidity shown in this study is roughly similar to that reported previously [15]. However, they also showed the difference in the Mg content which has the most viscous fluidity. It is considered that it is attributed to the different experimental conditions. In microstructures, with increasing the Mg content, the dendritic GAl was developed and the existence of precipitation regarded as phase (MgsAIs) is shown in the grain boundary. This tendency became conspicuous from the 5wt%Mg addition. And then, the grains were refined with the formation of the precipitation. References 1. O.A. Kaibyshev, Superplasticity of Commercial Aluminium Alloys (Moscow, Metallurgiya, 1984),37-51. 2. O. Ozdemir, lE. Gruzleski, and R.A.L. Drew, "Effect of LowLevels of Strontium on the Oxidation Behavior of Selected Molten Aluminum-Magnesium Alloys", Oxid Met, 72 (2009), 241-257. 3. F.H. Samuel et a!., "Influence of Composition, Sr Modification, and Annealing Treatment on the Structure and Properties of Cast AI-4 pct Mg Alloys", Metallurgical and Material Transcation, 34A (1) (2003), 115-129.

455

Transactions of Nonferrous Metals Society of China, 21 (2011), s23-s27. 15. K.R. Ravi et aI., "Fluidity of Aluminum Alloys and Composites: A review", Journal of Alloys and Compounds, 456 (2008), 201-210. 16. T.B. Massalski et aI., eds., Binary Alloy Phase Diagrams, vol. 1 (Metals Park, OH: American Society for Metals, 1986), 130. 17. Q. Han, and H. Xu, "Fluidity of Alloys under High Pressure Die Casting Conditions", Scripta Materialia, 53 (1) (2005), 7-10.

456

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

EFFECT OF PROCESS PARAMETERS ON CENTRIFUGALLY CAST AI-Si FGM Kiran Aithal Sl, Vijay Desai2, Narendranath S2, P G Mukunda 1 iNitte Meenakshi Institute of Technology, Yelahanka, Bangalore, Karnataka, 560064, INDIA 2Nationaiinstitute of Technology Karnataka, Surathkal, Mangalore, Karnataka, 575025, INDIA Keywords: Gradient, Centrifugal, Hardness, Rim thickness segregate towards the axis of rotation, while the denser particles segregate away from the axis of rotation [2].

Abstract Functionally Graded Materials (FGM) are such kind of materials wherein the properties and structure are varied from one end of the cast to the other intentionally. Centrifuge technique has been used in this study to produce Al-Si FGMs. Several process parameters determine the microstructure and the distribution of phases in the FG casting. These parameters include the size and initial concentration of alloying element, the centrifugal force, solidification rate, cooling rate. In this work an attempt has been made to produce FGMs using three different process variables such as mold temperature, melt temperature and mold rotational speed, their effect on the structure and properties. For this study AI-17wt%Si is used. From the results it is seen that for a particular melt and mold temperatures by increasing the mold rotation speed enhances the segregation of the Si particles at the one end of the casting. Similarly increasing mold or melt temperature only, increases the segregation.

In this instead of traditional centrifugal process where in the molten metal is poured into a rotating mold a different process is used where the mold is stationary and liquid metal is poured into it. This technique is basically used to produce solid cylinders compared to hallow cylinders in horizontal centrifugal technique. The set up designed and developed in house. The work attempts to find the effect of the process parameters: mold temperature, melt temperature and mold rotational speed on structure and hardness and also on the gradation along the length of the specimen.

Experimental Details In this process the centrifugal force magnitude 'G' is given by the Eq.l

Introduction

(1)

Functionally Graded Material (FGM) is a novel concept for the realization of innovative properties and functions that cannot be achieved by conventional homogeneous materials. FGM is a material that shows change in magnitude of property values from one end of a specimen or component to the other end. FGM has an intermediate layer whose structure, composition and morphology vary smoothly from one end of the specimen to the other end. These transition profiles are predesigned and intentional in order to achieve the desired properties [1]. FGMs multifunctional behavior and performance, enable wide scope for applications in aerospace, automotive, electronics, and biomedical sectors. The Functional gradient can be tailored to the specified service conditions, thus ensuring the best response of the system. One of the advantages of continuously varying volume fraction of the constituent phases is the elimination of stress discontinuity that is often encountered in laminated composites. Accordingly delamination problems are avoided and further the gradual transition also allows the creation of superior and multiple properties without any mechanically weak interface. Moreover, the gradual change of properties can be tailored to suit different applications and service environments.

where 'R' is the radius of the arm in meters, 'w' is the arm rotational speed in rad/sec and 'g' is the acceleration due to gravity. The 'G' plays an important role in positioning the reinforcement during solidification. FGMs with ex-situ and in-situ reinforcements can be processed by this technique. When particlecontaining slurry is subjected to centrifugal force, two distinct zones, one with enriched and the other with depleted particles are formed, separated by an intermediate graded zone. It is reported [3] that the extent of particle segregation and relative locations of enriched and depleted particle zones within the casting are mainly dictated by the relative densities of the particle and liquid, teeming temperature, melt viscosity, cooling rate, particle size and magnitude of centrifugal acceleration. The lighter particles segregate towards the axis of rotation, while the denser particles move away from the axis of rotation. The particles such as SiC, alumina and zircon in Aluminum alloy system will settle away from the axis, while the lighter Si, graphite, mica will drift towards the axis. In this work the forced segregation of hard insitu Si particles towards the upper regions (towards the axis) of the casting by centrifugal forces provides a unique approach to production of the FGMs. This region has higher surface hardness and wear resistance in the casting, while retaining high levels of toughness in the rest of the regions.

In centrifugal casting process for producing FGM the alloy melt or alloy melt with reinforcement particle melt is subjected to centrifugal force. Two distinct zones, one with enriched and the other with depleted primary alloy phase and particles, with an intermediate graded zone are formed. The extent of particle segregation and relative locations of enriched and depleted particle zones within the casting are mainly dictated by the densities of the particles, melt temperature, melt viscosity, cooling rate, particle size and magnitude of centrifugal acceleration. Depending on the density of particles, the lighter particles

In this study the centrifuge technique is used to process the functionally graded alloy and composites. The centrifugal force progressively increases the volume fraction of the Si reinforcement within the liquid Al matrix along the radial direction, owing to the density difference between the materials (PA1=2700 kg/m3, PSi=2300 kg/m3). A centrifuge machine was built for this purpose and tested for reproducibility. The centrifuge

457

machine is shown in Fig. I. The difference between this machine and the commonly used machine is that, in this machine the pouring is done while the mold is stationary and machine operates on vertical axis. Thus, centrifugal forces are not applied immediately as in the traditional casting methods since the mold takes some time to reach its casting speed. The principal advantage of this is good mold filling combined with microstructural control, which usually results in improved mechanical properties. Apart from high production rate, time saving, and ability to cast different shapes on reproducible basis, it can produce a stable system based on the in situ nucleation and growth of the reinforcement from the parent melt.

between the Si and the melt. The Si crystals are pushed to the top by the centrifugal force acting downward. The mold rotational speed, the mold and teeming temperatures have strongly influenced the structure development and the distribution of the primary Si. During mold rotation, the particle suspended in the liquid is subjected to centrifugal force acting on a particle given as mroJ and the gravitational force given by mg. The ratio of the centrifugal force to the gravitational force is called the gravitational coefficient (G) or G number. The rim thickness is defined as the thickness of the Si rich (primary) layer which indicates the extent to which particles have segregated along the length of the specimen. The rim thickness decreased with increased mold speed for a given mold and teeming temperature. This is attributed to the increase in cooling rates at higher rpm. It can be noted that, as the angular velocity increases, the G factor increases. Since the centrifugal force acting on the particle is G times higher than the gravitational force, the role of gravitational force can be ignored. Thus as the rotational speed increases, the force acting on the particles to segregate is expected to increase, as is observed in the study. For the casting produced under the lower levels of teeming and mold temperatures i.e 800°C and room temperature respectively, the specimen centrifuged at 22.3G (200rpm) showed no remarkable gradation of Si. The rim thickness of Si is observed to be 26mm. The maximum primary Si volume % obtained at this 'G' force is 8%, slightly higher than normal primary Si in this alloy. Similarly we can see from Figs. 4.41and 4.42 that, for casting produced at 300 rpm speed, the transition from hypereutectic microstructure to that of fine eutectic takes place at about 23mm (rim thickness) from the top surface. The primary Si is about 9%. At 400 rpm we can observe a rim thickness of l6mm and a volume fraction of 14%.

Fig. 1 Figure showing arm, mold and casting of the centrifuge machine Several parameters determine the microstructure and the distribution of phases/ particles in the FG casting. These parameters include the size and initial concentration of particles / alloying element, the centrifugal force, solidification rate, cooling rate, which in turn are controlled by: temperature of the mold and pouring temperature [4]. The process parameters used in the present work are provided in Table 1. Table I Process parameters for centrifuge casting Process parameters Levels Melt Temperature (T p), °C 800,900 Mold Temperature (Tm), °C 30 (room temperature), 180 200,300,400 Rotational Speed, rpm 'G'Number 22.35, 50.3, 89.42 Silicon wt.% 17 In the present work Brinell Hardness tester (model BV-120) is used which confirms to IS-Specification 1754. Prior to the testing, the specimen is cleaned to remove dirt and oil on the surface. The type of indentor and the load to be applied are selected in accordance with ASTM E-I O. The hardness is measured along the length of the cast specimen for every 4mm from top to bottom of the casting. At least five readings are taken perpendicular to the longitudinal axis. The final BHN value for each specimen is arrived at considering the statistical variation. Hardness is calculated for all the cast FG under different process parameters. Results And Discussions

Fig. 2. The microstructure of the AI-17wt%Si FGM cast at, T p =800°C, Tm=Room temp., G=22.3(200rpm). a) Top end showing primary Si, b) Bottom end showing primary a-AI dentritic structure

Microstructure And Gradation Of Si

The segregation and precipitation of Si at the top of the casting during the centrifuge casting is attributed to the density difference

458

For the FG alloys cast at 800a C pouring temperature, the intluence of higher mold temperature is studied at ditferent rpm's for evaluating the percentage Si segregated and the rim thickness. In the present work mold temperature of I80 a C is chosen. It was observed that at 22.3G (200rpm) the volume fraction of Si was 9% with a rim thickness of 24mm. At 50.3G(300rpm) and 89.42G(400rpm) the primary Si volume fraction was found to be 14% and 18%, whereas the rim thickness reduced to 21 mm and 14mm respectively. This gives an indication that the segregation of the primary Si forming a narrow Si rich region is influenced by the mold rpm and also the solidification rate (influenced by the mold temperature).

2S

,-----------------'""'""1 AI·17%SI N"'''OOrlllll Tr"'SOO"C T,.",30,,('

Q

rOI'

The optical micrographs of specimens, show microstructures with a non uniform distribution of needle-like Si particles in the matrix of a-AI (eutectic) at the bottom of the casting. As we move towards the top region, micrographs showed similar microstructure, except with increasing primary Si concentration. At the top region of the casting, the FG alloy exhibits not only needle-like eutectic Si phase but also large faceted massive primary Si crystals that signify a high silicon hypereutectic microstructure. The bar graphs for volume fraction of primary Si against normalized thickness are shown in Fig. 3. The rpm influences primary Si free zone (compliment of rim thickness) and the PFZ increases as the speed (G force) is increased.

(I

(U

0.2

(M

0..4

0.5

0.,6

0.1

GAl

1 Bottom

0.9

NOrillalized 25

AI-17%SI

N"'''OOrlllll

TI'",800"('

1',.=1110·('

;

if' il'g

Ai;

100

th

(,)

t

!i J!

ei g

til

""'"

:>

r-:

Q

Figure I. Energy balance relationship[l] Figure I shows that the heat input/output may be divided into the above items based on pot energy balance principle. The object of voltage reduction is the voltage combination in the heat input, the majority of which is voltage drop between anode-cathode. It should be pointed out that the high-temperature production

This will soon raises Chinese aluminum reduction technology to the world advanced level. Moreover, the consumption of energy and raw material for aluminum reduction production has been very high in recent years, especially power consumption. With the energy crisis, the aluminum reduction production cost must be reduced without delay. For this, the most efficient method is to reduce the DC consumption by increasing current efficiency (CE) and reducing pot voltage.

during aluminum reduction mainly depends on the louie heat generated in the bath in which the current passes from the anode to the cathode. The normal production shall be kept through the dynamic balance of heat output and heat input during operation. If the louie heat generated from heat input is not enough to maintain the heat output, the pot shall get cool gradually, and the process system shall be damaged. Therefore, the energy balance of the pot is maintained by reducing the heat dissipation in heat output combination as well as the voltage in heat input combination so as to reduce the voltage.

Analysis of mechanism and nature of pot work voltage reduction based on energy balance principle The pot energy balance was summarized by Warren Haupin as shown in Figure I.

537

Where:

Analysis of potentialities and approaches on voltage reduction by voltage composition of heat input

la

is the Anodic Current Density (A/cm2);

X 8 ta

is the Electrical Conductivity of the Bath (l/ncm); is the Bubble Layer Thickness (cm); is the Adhering Bubble Thickness (cm); is the Gas Fraction in the Bath (0.02 . %AI 20 3 ); is the Fraction of the Anode Covered with Gas

Elctrochemical Voltage

O.

Bubble Vol tage drop - - slotted anode block

-------I

bath vol tage drop

From Equations 2 and 3, we can see that the mechanism of bubble voltage drop reduction lies on improving the bubble release capacity, hence reducing the bubble coverage fraction and reducing the bubble thickness.

-- L

elrop - - ca.st iron bet("Ieen ca.thode carbon blD,::k and

Until now, the measures taken by the industry mainly include the slotting of anodes, the control the length-width ratio of anode, the improvement of bath composition, etc.

Figure 2. Heat input structure and approaches of voltage reduction

The slotted anode can make the bubble release from the anode bottom more efficient to reduce the bubble coverage fraction and the bubble thickness, so as to reduce the bubble voltage drop of anode, thus reducing the pot voltage.

The total pot voltage is just the sum of the parts as follows[2,3l: V pot = Vanode +

IEel +

11 sa + 11ca

+ Vbub+ VACD+ 11cc+ Vcath+ Vext

(I)

There is deeper research on the slotted anode technology abroad, including slotting location, slotting width, slotting depth, slotting process, etc. At present, the more mature slotted anode structure in China consists of 2 slots (at trisection location), slot width around I 1.5 cm, slot depth or height of half of anode consumed in anode change cycle as per the anode height generally. The slotting process is generally vibrating forming plus slot cleaning. As per a lot of on site tests in China, using slotted anode can reduce the pot voltage by mY.

Where: V pot

is the Pot Voltage (V);

Vanode

is the Anode Voltage (V);

IEel

is the Equilibrium Potential (V);

l1sa

is the Anode Surface Overvoltage (V);

11ca

is the Anode Concentration Overvoltage (V);

Vbub

is the Bubble Voltage (V);

VACD

is the Voltage Across the ACD (V);

11cc

is the Cathode Concentration Overvoltage (V);

Vcath

is the Cathode Voltage (V);

Vext

is the External Voltage (V).

Bath voltage dropl31

ia

= -(ACD - 0) X Where:

Figure 2 shows that if the design dimensions of the pot are determined, the object of pot voltage reduction is mainly the voltage drop reduction between the anode and the cathode (industry term: active voltage). If the CE is fixed, the object of voltage reduction is mainly the voltage drop of bubble (Vbub) and

ACD

Perspective of changing the bath composition

Bubble voltage dropl31

ia [

--

OIlO -

is the Anode to Cathode Distance (cm).

Equation 4 shows that the mechanism of ACD voltage drop reduction lies on changing the bath composition and reducing the ACD itself

in the bath across the ACD (VACD).

V"

(4)

Up to now, the measures taken in industry are as follows: the bath conductivity is increased by adding the additive, in which the most effective method is to add the LiF. and there will be significant effect combining with low bath ratio technology.

«(j - fa) . ' - -ta- - -

X (1 - e) 1.5 (l

From the present calculation and statistical data, it is seen that for every increase ofLiF by 1%, the voltage drop of unit ACD (cm) mY. For a pot with an ACD of 5 cm will be reduced by about for example, every 1% LiF addition can reduce the voltage by mY, and every 3% LiF addition can reduce it by mY, which is a considerable effect.

(3)

538

The main designs tried in the aluminium industry in China have been a stepped surface cathode, a sloped surface cathode and a flow resistance block.

Perspective of ACD reduction The ACD reduction is theoretically divided into: (1) effective ACD reduction; (2) non effective ACD reduction (see Figure 3).

Stepped surface cathode metal flow model

Anode

Bath

\Ietal

/iilliiili /'

6

4-

/

$'

/'\ \

7

a - Bubble layer t,- Effective ACD layer

Figure 4. Stepped surface cathode design

c - Noneffect i ve AeD 1 ayer

Figure 3. 3-layers structure model of ACD[4] As shown in Figure 3, the ACD consists of 3 parts including a- a bubble layer, b- an effective ACD layer and c- a non effective ACD layer. Zone a depends on the width of the anode, the specific gravity and viscosity of liquid bath, the surface tension of bath to carbon dioxide gas, the alumina concentration, etc.; zone b is a heating area for maintaining the high temperature production of the pot, as well as an isolation layer for making the wave crest of metal away from the lower edge of bubble to avoid the back reaction; zone c depends on the MHD cell stability. For the conventional pot, if the ACD is 5 cm, as per the calculation and averaged measurement, generally zone a (bubble layer) is about O.5cm, zone c (non effective ACD) is cm (it has relationship with the pot stability), hence, zone b (effective ACD) is cm. The irregular cathode technology and the horizontal current reduction technology are decreasing the height of zone c (non effective ACD) to reduce the pot voltage; and the current intensification technology is decreasing the height of zone b (effective ACD) to reduce the pot voltage, i.e. the lowest voltage of current intensification selected in order to satisfy the heat balance, thus obtaining the lowest height of zone b assuming no CE loss. Therefore, if the pot with bad stability has current intensification to reduce the voltage, it is highly possible that it will reduce the height of zone c and bring about more back reaction, thus the pot will experience CE loss and overheating.

Figure 5. Model of metal flow velocity of stepped surface cathode Sloped surface cathode metal flow model

Figure 6. Sloped surface cathode design j,;1f'!l:"'l!l

We can divide the type of applications used to reduce the ACD in three categories: (1) irregular cathode technology; (2) horizontal current reduction technology; (3) current intensification.

Ukl< tAft'!,·t2

Irregular cathode technology In 1994, Vittorio de Nora put forward the thinking of the irregular cathode structure. The irregular cathode structure is adopted to change the metal and bath flow state and reduce the melt flow velocity and the interface wave range of metal surface (reduce the non effective ACD), thus improving the pot stability in order to gain the option to reduce the ACD. Such technology is a kind of method to reduce the non effective ACD. Figure 7. Model of metal flow velocity of sloped surface cathode

539

Special insulation[5] between cathode carbon block and collector bar

Comparison of results obtained Table 1 is the comparison of calculation and measurement between irregular cathode and standard cathode in a plant in China. Max. flow velocity (cm/s)

This new kind of design is a cathode assembly which reduces the horizontal current by adding an electrically insulated region between the cathode carbon block and the collector bar, as shown in Figure 8.

Max deformation of metal surface (cm) CalculaMeasuretion ment

Calculation

Measurement

Standard pot

15.73

14.99

1.82

1.97

Irregular cathode

7.47

8.24

0.65

0.51

Variation percentage

52.50%

45%

64.30%

74.10%

AN

Table 1. Comparison of calculation and measurement between irregular cathode and standard cathode in a plant

Insula.ted ma.terial

Compared to the standard cathode, for the irregular cathode the flow velocity is reduced by about 50%, the maximum deformation of metal surface by 75% and noise by 15%. At present, the voltage of the most of irregular cathode pots in China is about V, based on the calculation of of CE loss. The power consumption can be reduced by 560 k WhiT Al compared to the standard cathode pots.

Figure 8. Cathode assembly for restraining the horizontal current ( lY) This design has been modeled using a 3D generic parametric whole pot model, based on ANSYS® 13.0, as shown in Figure 9.

Horizontal current reduction technology It has been proved by the long-term practice that the fluctuation of

metal liquid layer and bath liquid layer has close relationship with the horizontal current and the vertical magnetic field, which combined brings about the pot voltage fluctuation. So for a given vertical magnetic field, a reduction of the horizontal current in the metal can make possible a significant reduction of the height in the metal pad, reducing the cell heat loss and so provide an opportunity for pot voltage reduction, while maintaining the pot production and increasing the CE, all for the purpose of reducing the specific energy consumption. The horizontal current has relationship with the following factors: 1) 2) 3) 4)

Geometric dimensions, such as width and length of cathode, width and height of collector bars; Material of cathode, such as material of cathode carbon block, connection method between the cathode carbon block and the collector bar; Geometric dimensions of the pot, such as dimension of thermally insulating pier, hence the position of the ledge toe; The location of the collector bars exit out of the pot (side wall or otherwise).

Figure 9. Geometry of the ANSYS® based 3D generic parametric whole pot model The comparisons of simulation results between the cathode assembly without restraining horizontal current and that with restraining horizontal current are shown in Figures 10 and 11 respectively. From the above analysis and comparisons it shows that the cathode assembly with insulation has a good effect on the reduction of the horizontal current; from the curve distribution, it shows that the curve of no sloping pasting presents the raised parabola (Figure 10) with a maximum value of 0.26 A/cm2 . However, the curve of the anode bottom middle of sloping pasting presents a leveled curve (Figure II) at a value of 0-0.04 A/cm2 .

The up to date prototype tests were designed to: 1) 2)

Increase the electrical insulation between cathode carbon block and collector bar; Try a cathode design with bottom exit collector bar.

540

mid

iI·max: 257'1,9849 jymin: -25635899 jysvg: 1444,8881

4000, 3000 2000 '1000,

",:

", :

::;-

e.o II: cJf

018

Difference

Net Voltage

V

4.065

4.689

-0.624

Current Efficiency

%

95.10

92.34

2.76

DCkWhr/kg

12.74

15.14

-2.40

mV

3

17

-14

Net Specific Energy

"::.-

018+

I;::.c

14

'2o.u

Noise

... -

VI £:)

12 !II

Net Carbon

kgC / kgAI

0.420

0.446

-0.026

10

Anode Effect Frequency

#/cell/day

0.015

0.44

-0.43

Anode Effect Duration (>8V)

seconds

57

31

26

PFC Emissions*

C0 2 eq. kg/mt AI

16

247

-231

z

92 1

3 5

7 9 :U 13 15 17 19 21 23 25 Weeks Since Start-Up

Figure 13: D 18+ Current Efficiency and Specific Energy. After adjustment of the thermal balance, the bath temperature and AlF3 concentration are now stable and on target.

* CO2 equivalent is calculated as in reference [3], using the Tier 2 method and SAR (Second Assessment Report).

564

Conclusion The successful test and validation of the D18+ cell technology has proven that it is both technically and practically possible to update and replace the cell technology within an existing operating potline. Study of the feasibility and optimal engineering pathway is currently in progress to enable replacing the remaining 513 D 18 cells with the D 18+ technology in Lines I & 3 at DUBAL. Acknowledgements We are most thankful to Abdulmunim Bin Brek for the exemplary work that has been carried out by the DUBAL Major Projects team to construct the seven D 18+ cells well within the project target limits. The DUBAL Performance Improvement team and who were responsible for the conception and design of the D 18+ technology have played a major role in the success of this project. Acknowledgement is also given to Syed Fiaz Ahmed and the DUBAL Technology Development engineering team for their proactive approach during the design stage and timely response to the design modifications. The effort and dedication of the potroom operations personnel of Kamel Alaswad, Jose Blasques, Handerson Dias and their team have been crucial for the successful start-up, stabilisation and optimisation of the D 18+ design. Also the invaluable work and support of the D 18 Process Control team have been indispensable for the progress achieved with this project.

I.

References D.Whitfield, T. Majeed, S. Akhmetov, M. AI-lallat: K. AIAswad, 1. Baggash, A. AI-Zarouni, "Update on the Development of DI8 Cell Technology at DUBAL", Light Metals 2012, 477 - 48l.

2.

M. AI-lalla!: A Mohamed, Kumar A, M. Ali, "Evolution of CD20 Reduction Cell Technology Towards Higher Amperage Plan at DUBAL", Light Metals 2009, 451 - 454.

3.

A. Al-Zarouni, A Zarouni, N. Ahli, S. Akhmetov, 1. Baggash, L. Mishra, A Al-Jasmi, M. Bastaki, M. Reverdy, "DX+ an Optimized Version of DX Technology", Light Metals 2012. 697 - 702.

565

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

INDUSTRY TEST OF PERFORATION ANODE IN ALUMINIUM ELECTROLYSIS TECHNOLOGY Yingfu Tian 1, Hesong Li 2, Longhe Wei 2, Xi Cao

2,

Jianguo Yin

3

I Chongqing Tiantai Aluminum Industry Co., Ltd. Chongqing, China 401328;

2 School of Energy Science and Engineering, Central South University, Changsha, China 410083; 3 Chongqing University of science and technology, Chongqing, China 401331 Keywords: perforated anode, aluminum electrolysis, anode bubble, cell voltage, energy saving

T]-----current efficiency.

Abstract

In above equation, there are two ways to reduce the energy consumption Wpractical' One is increasing current efficiency 11 in the cell operation, and another is decreasing the average cell voltage V.

In recent years aluminum industry in China has been developed rapidly. In 2011 China's electrolytic aluminum output reached 18.06 million tons and has topped the list in the world for II consecutive years. However, the energy and environmental issues restrain development of aluminum industry. So it is necessary to promote the use of energy saving technology in the aluminum industry. The composition of the cell voltage is analyzed in this paper. It is reasonable that the cell voltage of ordinary flat cathode cell is about 4.05V. The approximate linear relationships between the bubble layer thickness and the anode width is analyzed. Changing the structure of the anode and adding two rows of the vent on the common anode can reduce the gas running distance below carbon anode, anodal bubble voltage and the work voltage of aluminum reduction cells under the same current efficiency of aluminum reduction cells. Perforated anode technology was tested in small-scale and large-scale industrial cells and the experimental results show that average current efficiency of the test cell is 91.447%, the direct current consumption reaches 12337kwh/T-AI. The anode effect coefficient of the test cell with good thermal balance reduces to an average of 0.185 time Icell-day. The results are consistent with the theory and a comprehensive promotion goal is achieved.

After lots of studies on how to improve the current efficiency, the current efficiency has been effectively improved from 90 percent in 1980s to 95 percent now. It is very difficult to further improve the current efficiency. In recent years reducing the average cell voltage V without loss of current efficiency has become the major research goals for energy-saving in aluminum reduction industry. The average cell voltage V consists of following sections: V=Vpolarizaqtion +Vanodc +Vcathodc +Vdc=< 71 . V"Xg > where, for any j, a(locg)1(locg) =< X(locg)j >. Pg(OtagVg+

(Vg Q9Vg))

With this assumption (10) becomes

Approximations and modeling

(15)

In order to have a tractable model we make the following assumptions. P2 = 0; the Reynolds tensor is included in the viscous term a(1'l with a change of the viscous constant which then becomes a diffusivity, (the notation will not be changed); we assume that (in T l , Tg, Tl and Tg) PI = PI = Pg = Pg· In order to handle the averaging appearing in the equations we consider the different terms separately (without the indices). We also assume that Maxwell tensor is not affected by the averaging process. Setting (6) v = v + W, so that < X1W >= 0, one gets

From (3) and (13) one draws divVl Otag

=

+ div(agvg) =

0, O.

(16) (17)

Constitutive equations We will now make the assumption that the field (14), i.e. f int , is a function of Vslip only. Following [5] we assume that

(18) where A is a constant. Introducing this expression into (15) yields, since by assumption PI = P2,

(7)

586

(19)

so that

_

VsZip

Vih

=

(20)

-T'

The current j = (jl,j2,j3) is then computed in the following way to avoid rough approximations.

1·

The coefficient A has to be determined experimentally. The model With the above results we are now ready to give the equations on which our model is leaning. Fluid averaged equations (21 ) &iVVl = 0,

+ V· (aZVI Q9 VI)) -Vih + V· (azh) + azV· MI + pzazg·

pz(Ot(aZVl) =

(22)

II

J z1/;dx = -

(29)

UXz

In this study, it is assumed that the alumina feeding is known, and that the dissolution is instantaneous. The alumina distribution in the bath rlb is given by the following partial differential equation 8c( x, t) -dw . (( a x,t )Vc (x,t )) +

ot

+ agAVsZip =

(v(x, t)IVc(x, t))

O.

(24)

=

0

(30)

where • c is the alumina concentration in mallm3 ,

Jump averaged conditions

• a is the anisotropic diffusion coefficient. A value of 0.5m 2 I s was determining the Reynolds mean tensor. It also leads to an average alumina concentration reflecting industrial cells.

(25)

+ ag =

o¢ 'ljJdx (5 (c) --;:;:-

Alumina distribution

(23)

az

II

for suitable functions 1/;, and 1 = 1,2,3. The model will take into account the analysis of the parameter (5 as a function of the bath composition.

Gas averaged equations

a gVih

1

1.

(26)

Boundary conditions on the different fields have to be added. Alumina diffusion and convection: Theory

Let rl C Jl{3 be the domain representing the cell, rlb be the bath, and rla the anodes.

• v is the velocity field in rlb induced by the bubble motion, and the MHD, The boundary and initial conditions have the following form. • The concentration c is given on the feeder. • The concentration flux on the anode and the

Current density

oc

The current density distribution in the bath is a function of alumina concentration, and is given by

bath-aluminum interface is a On + f3(jln) = 0; 1 the coefficient f3 verify f3 = if [c] = mallm 3 ,

j = -(5(c)V¢

with z is the valence, and F is the Faraday constant, in our case z = 6.

zF

(27)

where ¢ is the potential, and (5 is the electrical conductivity, which depends on the alumina concentration c. Since &iv j = 0, we get

-div((5(c)V¢) = 0

(28)

with following the boundary conditions (jln) = jo on the anodic rod, ¢ = 0 on the bath-metal interface, o¢ and (5(c) On = 0 elsewhere.

8c • a On

=

0 elsewhere,

• c is given at time t = O. Weak formulation Let us use the following notations.

587

• rb is the bottom of the cell (the metal-bath interface),

• ra

c) The electric current jh = (jh,1,jh,2,jh,3) is obtained as, Vl = 1,2,3, and VIjJ E Zh,

is the top of the anodic rod (the entrance of the current),

• rj

1

is the feeder domain,

• r ab

o

is the interface between the anode and the bath.

=

{1jJ E Hl(rl), 1jJ = 0 on

rd '

(31)

equipped with the norm IV)I = IV)lv = 11\71jJIIL2(0)' (32)

=

jo1jJdiJ.

/'

Jr.

1

/' (8c

m+l

0"

(33)

T

1

Then the computation of the current density j = (jl,j2,j3) is obtained from equation (29) \f1j) E L2(rl). Finally, the weak formulation for the alumina concentration is: Find C E L2(0, T; Hl(rlb)) with C = Cj on r j, such that, V1jJ E W,

Jo" at

0

a¢h iJ(C)-a 1jJdx. Xl

(36)

Find ch'+l, solution of

equipped with the norm IV)I = IV)lw = 11\71jJlbco,,)· The weak formulation of problem (28) is: Find ¢ E V such that, VIjJ E V, /' (iJ(c)\7¢I\7IjJ) dx

1

d) A BDF scheme of order 2 [6] is used for the time discretization of the concentration equation (34). Moreover a Petrov-Galerkin streamline diffusion method is applied for the advection term [7, 8]. We get the following equation (5 small parameter) .

With these notations, set

v

jh,l1jJ = -

0"

1

+ 5(\7ljJlv)) dx+ (0:\7

+ 5(\7v)lv)) dx =

-(31

raburb

/' 2ch' -

Jo"

dx+

0"

1(jln)lljJdiJ+ (1jJ

+ 5(\71jJlv)) dx.

(37)

T

+(v l\7C))V)dX+ /' (o:\7cl\7V))dx=

-(31

Jo"

rnhurb

1(jln)l1jJdiJ.

(34) Alumina diffusion and convection: Industrial cell

Numerical methods Let us decompose the domains rl, and rlb into classical tetrahedral finite element mesh. For the numerical simulations of problems (33), (29), and (34), the following algorithm is used.

• Initialization An initial alumina concentration distribution Co is given at time t = O. Let T be the time step.

• Iterations For time tm at time tm

= Tn . T,

if ch' is the concentration

a) Compute the electrical conductivity iJ iJ( ch').

=

b) Find ¢h E Vh such that, VIjJ E Vh, /' (iJ\7¢hl\71jJ) dx =

Jo

/'

Jr.

jo1jJ diJ.

(35)

In this section some numerical result for the computation of the alumina distribution in the bath are presented. On the feeders, the alumina concentration is set to 5% of the bath weight. As a stationary solution is presented, continuous feeding is assumed. The impact of dump feeding could easily be analysed. Figure 1 correspond to the stationary alumina distribution, when the velocity is neglected. The concentration is shown under the anodes. The two feeders locations appear clearly in the figure. The asymmetry of the diffusion pattern reflects the larger channel width at the feeders. Figure 2 shows the alumina concentration under the same conditions at metalbath interface level. Away from the feeders, at a distance larger than about one anode width, the concentration is close to 2.55%. The vertical variation of the alumina concentration is 0.5% under the feeders. It is negligible away from the feeders.

588

Figure 1: Alumina concentration in the bath when the velocity is zero, under the anodes

Figure 4: Alumina concentration in the bath when the velocity is induced by the MHD

The previous cases did not take the bubbles into account. It is well known that they have an important effect on the velocity field. Moreover the considered cell has slotted anodes. This also has an impact on the velocity. Figure 5 considers the case when the velocity field consists of the effects of the MDH, the bubbles, and the slots in the anodes.

Figure 2: Alumina concentration in the bath when the velocity is zero, bath-metal interface

In figure 3, the velocity streamlines induced by the MHD are presented. Figure 5: Alumina concentration in the bath, velocity induced by MHD, bubbles, and slots

1

velocity

2169

0.2000

w-

From the different figures, the alumina concentration field appears as slightly modified by the velocity field. However, when considering the concentration evolution, the time needed for reaching the stationary state is reduced by a factor 2 in any situation when the velocity field is acting. Therefore the velocity field plays an important role in the feeding process (alumina dumps).

I

0.000

Figure 3: MHD velocity in the bath (streamlines)

The impact of this velocity field is shown in figure 4.

To highlight the role of the velocity field, figures 6 and 7 show the difference between the alumina concentration field due to the diffusion only and in presence of MHD velocity, resp. total velocity field.

589

• The velocity field has an important effect for the alumina distribution under the anodes. It helps to homogenize the alumina concentration . • Bubbles and slots modify the velocity field which generate turbulences leading to increased homogenizing effects.

Figure 6: Alumina concentration variation due to MHD [%]

References

[1] O. Kobbeltvedt, S. Rolseth, J. Thonstad. The dissolution behavior of alumina in cryolite bath on a laboratory scale and in point fed industrial cells. Department of Electrochemistry, Norwegian Institute of Technology, N-7034 Trondheim, Norway SINTEF Materials Technology, N7034 Trondheim, Norway [2] RG. Haverkamp. PhD Thesis, University of Auckland (1992).

Figure 7: Alumina concentration variation due to MHD, bubbles, and slots [%]

The highest differences are observed at the ends of the cell, due essentially to the MHD effects. High negative values relate to high alumina concentration difference. The effect of bubbles and slots generate turbulence, homogenizing the concentration distribution. Conclusions

A new model for the velocity field in presence of MHD, and small bubbles is developed. This velocity field is used to determine the evolution of the alumina concentration using a non-stationary convection-diffusion equation. This equation takes into account the feeding, and the Faraday law at the anodes and cathode. The application to an existing cell with two point feeders demonstrate the following: • The alumina concentration can vary up to 2.5%. Typically a variation 1% can be expected between anodes. • The time needed to reach the stationary state due to the diffusion process only is twice the one for the case with MHD and bubbles effects velocity fields. It was found around two minutes.

590

[3] O. Kobbeltvedt, S. Rolseth, J. Thonstad. On the Mechanisms of Alumina Dissolution with relevance to Point Feeding Aluminium Cell. Light Metals, TMS, 1996. [4] D.A. Drew and S.L. Passman. Theory of Multicomponent Fluids, Spinger 1999 [5] F.RG. Panescu. Modelisation eulerienne d'ecoulements diphasiques a phase dispersee et simulation numerique par une methode volume-elements finis, Inria Sophia Antipolis 2006 [6] E. Hairer, S.P. N0rstett, G. Wanner. Solving Ordinary Differential Equations 1. Springer-Verlag, 1987. [7] C. Johnson, J. Saranen. Diffusion Methods for the Incompressible Euler and N avier-Stokes Equations. Math. of Compo 47, (1986), pp. 1-18. [8] L.P. Franca, G. Hauke, A. Masud. Revisiting stabilized finite element methods for the advectivediffusive equation. Comput. Methods Appl. Mech. Engrg. 195, (2006), pp. 15601572.

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

INVESTIGATION OF ELECTROLYTIC BUBBLE BEHAVIOUR IN ALUMINUM SMELTING CELL Morshed Alam), Yos Morsi), William Yang2, Krishna Mohanarangam2 Geoff Brooks) and John Chen3 lFaculty of Engineering and Industrial Sciences, Swinburne University of Technology, Hawthorn, Victoria, Australia. 2CSIRO Process Science and Engineering, Clayton, Victoria, Australia. 3Department of Chemical and Materials Engineering, University of Auckland, Auckland, New Zealand. Keywords: Hall Heroult cell, Electrolytic process, Geometric and Dynamic Similarities.

Abstract

at the porous sections on the anode surface and undergo spherical growth, lateral spread, mutual impingement and coalescence to form a big bubble as shown schematically in Figure 1. The bubbles are then roll along the anode surface and escape around the anode edge. The measured gas bubble layer thickness was approximately 5mm for a horizontal anode. The effects of current density (CD), anode-cathode distance (ACD) and anode inclination angle on the gas layer geometry, anode coverage, bubble velocity and gas release frequency were investigated. ACD had no effect on gas bubble behavior. An increase in CD increased the bubble size, thickness of bubble front, average fraction of anode surface covered by bubbles and the bubble velocity while an inclined anode was found to decrease these parameters. The bubble release frequency was found to vary from 0.2 to 3.3 Hz depending on the anode inclination angle and was not influenced by the CD.

A 1I4th scale low temperature electrolytic model of the HallHeroult cell was constructed to investigate the electrolytic bubble formation mechanism, coalescence and movement under the horizontal anode surface. Geometric and dynamic similarity between the model and real cell was maintained through using similarity criteria. A 0.28M CuS04+20%H2S04 solution was selected as an electrolyte where Cu was deposited at the cathode and O2 bubbles were generated underneath the anode, similar to the phenomena of real cell. The bubble generation mechanism, movement, coalescence and detachment under the electrolytic medium were observed using a high speed camera. It was found that electrolytic bubbles generate uniformly under the whole anode surface and grow through gas diffusion and coalescence. At higher current density and higher anode inclination angles, bubbles escape quickly from underneath the anode surface. The bubble layer thickness and bubble sizes were also found to decrease with an increase in anode inclination angle. Introduction

Aluminum is produced by Hall-Heroult electrolytic process which was invented independently by Hall and Heroult in 1886. In this method, alumina CAI 2 0 3) is dissolved in a molten cryolite (Na3AlF6) bath at around 950°C where it is reduced to produce liquid aluminum metal and oxygen ions. The liquid aluminum metal is slightly denser than the electrolyte and is continuously deposited at the bottom of the cell while the oxygen reacts with the carbon anode to form CO 2. The overall cell reaction is: 2AI 2 0 3(solution) + 3C(s) = 4AI(l) + 3C02 (g)

Figure I Bubble shape according to Fortin et al.[I] Solheim and Thonstad [2] found that bubble size decreased with the addition of i-propanol which inhibits the coalescence. It was reported that the smaller bubble size results in higher accumulated gas volume as well as higher resistivity in the bubble layer. Shekhar and Evans [3] observed that the bubble layer becomes thinner in the case of a tilted anode. Xiang-peng et a1. [4] reported that the bubble detachment volume decreases and bubble sliding velocity increases with an increase in anode inclination angle. The bubble velocity was found to decrease when ACD was less than 4cm which is in contrast with the results of Fortin et al. [1] where it was reported that ACD has no effect on bubble behavior. Che et al. [5] observed that the bubble shape changes from ellipsoid to crescent with an increase in gas flow rates. In a later studies, Perron et al. [6] observed the existence of two distinct bubble regimes under the anode surface: "creeping bubble (a and b)" and "the bubble on a wetting film (c and d)" as shown in Figure 2.

(1)

The gas bubbles induce flow in the cell which plays an important positive role in homogenization of the alumina distribution and the temperature field in the electrolytic bath. Conversely, the gas bubbles increase the ohmic voltage drop underneath the anode surface which in turn results in higher energy consumption for the smelting process. The phenomenon of bubble formation and sliding underneath the horizontal surface is complex due to the bubble shape, surface tension and the anode surface characteristics. A number of studies have been carried out in the past on the bubble behavior under the anode surface and its effect on the electrolyte flow. Fortin et a1.[I] used a full-scale water model where anodic gas evolution was simulated by passing air through a micro-porous polyethylene plate. The flow rate of air was selected from the current density and gas evolution correlation (four electrons are necessary to produce one mole of CO 2) which is 10 kAm-2 = 2.71 Lm- 2s- 1• The gas bubbles nucleate

a

b

c

Figure 2 Regimes of movement of the bubbles[6].

591

d

(Inertial force )/(Buoyancy force)

Aussilllous and Quere [7] also reported the formation of a liquid film between the bubble and anode surface. During creeping motion, the longer axis lies in the direction of the displacement while for wetting film bubbles the longer axis lies perpendicular to the displacement.

Modified Froude number =(p,,'(/)I(PrPg )gL

Das et al. [S] observed that bubbles lose their symmetric shape immediately after the detachment when sidewall is at a close proximity. In their other studies [9, 10], it was reported that bubble size increases with an increase in liquid surface tension and decreases at higher anode inclination angle.

Modified Reynolds number p,)qL)l1'

Modified Weber number = (Pgc/ L)I(J

Eotvos number = (g(Prp,)L2)1(J

Cooksey and Yang [11, 12] measured bubble induced liquid flow in a full scale water model of aluminum reduction cell using Particle Tmage Velocimetry (PTV) technique. A recirculation zone was detected in both the center and side channel of the cell. The area of high turbulence was located in the gas plume region near the end of the anode and at the liquid surface. Wang et al. [13] observed similar phenomenon using the Laser DopIer Velocimetry (LDV) technique in their physical modeling study.

=

(Inertial force)/(Surface tension force)

(Inertial force )/(Viscous force)

(Buoyancy force)/(Surface tension force)

Here, q is the gas generation rate per unit surface area, g is the acceleration due to gravity, PI is the liquid density, Pg is the gas density and (J is the surface tension of surrounding liquid, L is the characteristic dimension and 1'1 is the viscosity of surrounding liquid. From the literature review, it can be concluded that the buoyancy force and inertial force generated by the evolving gases are the main driving forces for the bubble movement and bath motion. Also, the Eotvos number together with Morton number is used to characterize the bubble size and shape moving in a surrounding fluid medium. Hence, in the present study, the Modified Froude number, Eotvos number and Morton number were considered for dynamic similarity analysis and are presented in Table 1:

Tn the physical modeling studies that have been discussed so far, gas bubbles were generated mechanically by injecting air underneath the anode surface. Qian et al. [14, 15] pointed out from their low temperature electrolytic model that electrolytically generated bubbles are smaller compared with the bubbles formed by forcing air through porous plate. 2M NaOH was used as electrolyte in their study and the anode surface was covered with foamy layer of tiny bubbles. It was found that at equal current density or equivalent gas generation ratio, bubble resistivity was 20% higher in case of electrolytically generated bubbles.

Table 1: Fluid properties and dynamic similarity analysis of the present model

There have been a number of studies [16-20] using bench-scale experiments where bubbles were generated electrolytically and the electrochemical reactions are similar to those in an actual cell. However, in those studies the anode surface area was too small (10-20mm) compared with the actual cells except the one of Aaberg et al. [21]. As a result, the measured bubble sizes may be different as it is known from the physical modeling that bubbles coalesce during movement under the horizontal anode. Aaberg et al. [21] carried out bench-scale experiments of real aluminum electrolytic cell using a 100mm graphite anode. The average bubble volume at release, bubble thickness and fraction of anode surface covered by anode were reported. However, it is very difficult to make visual observation of the electrolytic bubble formation, coalescence and growth mechanism due to the opaque electrolytic bath of bench-scale experiments. Therefore, the aim of this work was to enhance the current understanding of the electrolytic bubble formation. movement and detachment characteristics, under the horizontal anode surface, using a low temperature electrolytic model of the Hall-Heroult Cell.

Electrolyte Electrolyte density (kg/m3) Electrolyte surface tension (mN/m) Electrolyte viscosity (kg/ms) Bubble cross section diameter before release, (mm) Anode length, (m) Modified Froude number Eotvos number Morton number

Real Cell Cryolite 2100

Present model 0.2SM CUS04+ 20%H2 S0 4 1195[23]

129

9S.7[24]

0.00251

0.0011 [23]

11- 13[IS, 19]

5 - IS

1.35[1] 1.162x 10' lU

0.35 1.43 x 10-

19.32 - 27 S.6372x 10. 11

2.97 - 3S.4 1.245x 10- 11

Table 1 shows that the Eotvos number and the Morton number of the model and real cell have similar order of magnitude. Hence, it can be said that the generated bubble shape in the model cell is likely to be similar to that of a real cell because these dimensionless numbers characterize the shape of bubbles. The bubble cross-section diameter before release was used as the characteristic dimension in the Eotvos number equation. Table 1 also shows that Modified Froude number of the real cell is one order of magnitude higher than the model cell. This was because the gas generation rate in the present model cell was lower than the desired value. An increase in current density increases the

Design of Electrolytic Cell Tn designing the low temperature electrolytic model, emphasis was on maintaining the geometric and dynamic similarity between the real cell and the low temperature electrolytic model. A 1I4th scale model of the Hall-Heroult cell was built by maintaining complete geometric similarity. In order to maintain the dynamic similarity, five different dimensionless numbers were considered as reported by Zhang et al. [22]. These are

592

bubble release frequency, but the bubble volume at release is not influenced by the current density [21]. The main goal in this experiment was to generate bubbles electrolytically which are comparable with the bubbles of real Hall-Heroult cell and then study the bubble formation, movement and detachment under the horizontal anode. That is why emphasis was given on the Eotvos number and Morton number as these numbers characterize the bubble shapes.

where it was recirculated again into the cell. A DC power supply (0- SOOA) was used to supply current for electrolysis. But current could not be increased over 12SA as the voltage drop was greater than SV which is the maximum voltage reading for this power supply unit. In the electrolytic process, the oxygen bubbles create acid mist at the liquid surface which is hazardous for human health. A 200mm flexible reinforced PVC pipe was placed on top of the cell to extract the generated acid mist as shown in Figure 3.

Experimental rig In the previous low temperature electrolytic study of Qian et a1.[14], 2M NaOH was used as the electrolyte, so bubbles generate both at the cathode and anode. But in the actual cell, no bubbles are generated at the cathode. Therefore, a separator was used to isolate the bubbles that generated at the cathode in order to minimize the effect of cathode bubbles on the anode. In the present study, the electrolyte was selected in such a way that no bubbles should generate at the cathode. After studying the available aqueous solutions, it was found that CUS04 solution is an excellent candidate, and therefore was used in this study. The overall electrolytic reaction is:

Experimental Procedure

The electrolyte (0.28M CUS04 + 20% H2 S0 4 ) was circulated continuously from the heated tank to the experimental rig to ensure that the temperature of the electrolyte remained around 4SSOoC inside the model cell, which is the requirement for the electrolysis of CUS04 solution. The temperature was monitored through a digital stem thermometer (accuracy ±O.IO°C). The circulation was turned off prior to the commencement of electrolysis so that the liquid velocity did not affect the bubble behavior. Then, the power supply was turned on and set to the desired current density to commence the electrolysis. The experiments were run at different current densities and anode inclination angles to investigate the effect of these parameters on the bubble characteristics. The operating conditions used are presented in Table 2:

Figure 3 shows the experimental set-up used in the present study. The inside dimensions of the model cell was SOOmm x IISmm x

Table 2 Operating Conditions. Experiment no 1 2 3 4 S

6 7 8 9 10 11 12 13

Angle of inclination (degree)

Current density (A/cm 2)

Calculated gas generation rate (m3 m"2s"l)x 10"

1 1 1 1 2 2 2 2 3 3 3 3 3

0.074 0.087 0.1 0.112 0.074 0.087 0.1 0.112 0.074 0.087 0.1 0.112 0.31

O.OSI

3

0.06 0.068 0.077 O.OSI 0.06 0.068 0.077 O.OSI 0.06 0.068 0.077 0.213

The average bubble cross-sectional diameter and thickness (l and d in Figure 7 respectively) of the departing bubbles from the edge of the anode surface were measured using the high speed camera at 2S0 frames per second from two different locations: (a) perpendicular to the direction of motion of the bubbles and (b) inclined from the horizontal plane. The captured images were then processed using image processing software "imager'.

Figure 3 Experimental set-up. 400mm. The anode and cathode dimensions were 3S0mm x 11Smm and were placed parallel to each other. These dimensions were decided according to I14th scale geometrical similarity analysis with the real cell[l]. The ACD was fixed at SOmm. Lead and stainless steel plates were used as the anode and cathode respectively. The anode immersion depth was fixed at 200mm. The CUS04 solution was stored in a tank where it was heated and maintained at SOOC. The electrolyte was supplied into the model cell through the inlet at the bottom left corner of the cell. When the electrolyte level reached 2S0mm from the bottom, it passed through the electrolyte overflow line into the heated tank from

Results and Discussions

After the start of electrolytic process, the entire underside of the anode was covered by tiny bubbles as shown in Figure 4. Gradually the immobile bubbles grew in size due to gas diffusion and then through coalescence with the surrounding bubbles. There were no clear areas under the anode as compared to water modeling studies [1, 13, 2S]. The bubbles remained stationary

593

Figure 4 Initial bubble formation under the anode surface at 0.087 A/cm 2 and 0.31 A/cm2 current density and 1 degree angle. at the nucleation point until the component of the buoyancy force, parallel to the anode surface, was large enough to overcome the surface tension force and drag force. This occurred when the bubbles reached a certain volume at a constant angle. After that the bubbles detached from the nucleation site and slide along the anode surface. Bubble size before detachment also depends on the anode inclination angle which will be shown later in this paper. The formation and movement mechanism was similar at both higher and lower current densities. The higher current density only speeds up the process as shown in Figure 4. At t=4 sec, the bubbles started to detach from the nucleation point in case of 0.31 A/cm 2 current density whereas the bubbles were only growing in case of 0.087 A/cm2 current density. These figures also show that although small bubbles generated under the entire anode surface, the bubbles grew bigger only at a limited number of nucleation sites, which may depend on the morphology of the anode surface. A separate study on this particular issue is required to understand the effect of anode properties on bubble characteristics. Figure 5 shows the bubble flow pattern under the anode surface at different anode angles and at a fixed current density of 0.112A/cm 2 . A number of larger bubbles were observed when the anode was 1 degree, and bubble sizes decreased at higher inclination angles. At the beginning of electrolysis, the shape of the bubbles (less than 4mm) were spherical and then slowly

Figure 5 Bubble pattern under the electrode at different anode angle and 0.112A/cm2 CD.

594

converted to ellipsoid as these grew bigger due to coalescence. Figure 6 shows that the cross-section diameter of the detached bubble decreases with an increase in anode inclination angle. This occurred because at higher inclination angle bubbles were travelling faster and didn't get enough time to coalesce. When the anode was nearly horizontal (I degree), the measured mean bubble cross-section diameter was 1O.76mm with a standard deviation of ±4.47mm. This value is lower than the reported value of l8mm by Cassayre et al. [19] at similar current density. The reason for the difference might be that the anode was perfectly horizontal in the previous study[19] whereas in our case the anode angle was I degree. But the present results agreed well with the results of Xue and Oye [18] which ranged from 11-12mm at practical current density of smelting cell.

Figure 7 Bubble contact angle under anode surface Figure 8 shows the change in bubble terminal velocity before detachment from the anode surface with increasing current density. As expected, the bubble terminal velocity was found to increase with the increase in current density. This occurs because, at higher current density, the bubble generation rate increases underneath the anode surface. As the bubble penetration depth inside the electrolyte is limited, more and more bubbles slide under the anode surface and escape through the anode edge at higher velocity if the current density is increased. The figure also shows that bubble terminal velocity increases with the increase of electrode inclination angle. The component of the buoyancy force parallel to the anode plane increases at higher inclination angle which in turn accelerates the gas bubbles underneath the anode surface and the bubbles escape quickly from the edge of the anode.

4.5

If!

E

g

ltl

I

14

11

.!!!

4..0

$5

$.0

E S-

t> 10

2.5 $!

i

to

"

,2

l!!ft

...'"

e ;;

.ill

.

dl;

n'" e

.,

1$ :l3 10

4

0.5

2

aD

""

::l

ro

130

120

Figure 6 Average bubble diameter and thickness under anode before release. Current density is 0.1l2AJcm2 .

g

The thickness of the gas bubble layer under the anode surface reached a maximum of 4.1mm with an average of 3. 71mm in case of I degree anode inclination angle. This is in good agreement with reported maximum thickness of 4mm and 5mm from the previous laboratory scale aluminum electrolysis studies [17, 18, 21] . The thickness of the gas film was also found to decrease with an increase in anode inclination angle as shown in Figure 8, which was also observed by Shekhar and Evans [3] in their physical modeling study. This was expected because at higher inclination angle, the gas bubble velocity increases and the gas film thickness should decrease to satisfy the continuity equation.

.s

f

'ill

110 100

00 00

70 00 00

.ill

.0 JCJ

'"

al

40 20 :11l 0.05

0.10

0.15

0.:11l

0.25

CUf,,,nt Density fA/em')

Wettability is an important parameter for investigating the bubble characteristics underneath the anode surface. The notion of wettability is based on the concept of an equilibrium state between the interfacial surface tension of three phases and the existence of equilibrium contact angle. The contact angle is the angle between the solid surface and the gas-liquid interface as shown in Figure 7. The final contact angle of the bubble with the anode surface before departure was measured using the image analysis software "imager'. The measured bubble contact angle was found to vary from 115 degree to 135 degree which are in good agreement with the results of Xue and Oye [18] that ranged from 110 to 130 degree.

Figure 8 Effect of current density on bubble terminal velocity. Conclusions

A 1I4th scale low temperature electrolytic model of the HallHeroult cell was developed to investigate the bubble characteristics under the anode surface. 0.28MCuS04 + 20%H 2 S0 4 solution was used as electrolyte which deposited Cu at cathode and produced O2 bubbles under the anode during electrolysis. Proper Similarity analysis was carried out to make the model geometrically and dynamically similar with the real cell. The behaviour of the electrolytically generated bubbles were analysed through high speed camera. It was observed that electrolytic bubbles generate uniformly under the anode surface

595

and then grow bigger due to gas diffusion and coalescence. The average bubble size before detachment from the anode edge and thickness was found to be 10.76mm and 3.7mm when the anode was nearly horizontal. These values were close to previous literature predictions. The bubble size decreased and bubble terminal velocity increased with an increase in anode inclination angle. The bubble terminal velocity was also found to increase with an increase in current density. The observed contact angle between anode and the gas bubbles was ranging from 115 to 135 degrees which was also in good agreement with the previous studies. At present, this study is going on to investigate the effect of anode angle and current density on the anode coverage ratio, bubble resistance and bubble volume before detachment. The results will be presented in future publications.

10.

11.

12.

13.

14.

Acknowledgments Authors would like to acknowledge CSIRO Light Metals Flagship, Australia for funding the project.

15.

References 1.

2.

3.

4.

5.

6.

7.

8.

9.

Das, S., Morsi, Y., Brooks, G., Yang, W., and Chen, 1.1.1., The Principle characteristics of the detachment

16.

Fortin, S., Gerhardt, M., and Gesing, A.l, Physical Modelling of bubble behaviour and gas release from aluminium reduction cell. in Light Metals. 1984. TMS: p.721-741. Solheim, A and Thonstad, l., Model Cell Studies of Gas Induced Resistance in Hall-Heroult Cells. in Light Metals. 1986. TMS: p. 397-403. SHEKHAR, R. and EVANS, lW., Physical Modelling Studies of Electrolyte Flow Due to Gas Evolution ans Some Aspects of Bubble Behaviour in Advanced Hall Cells: Part 1. Flow in Cells With a Flat Anode. Mtallurgical and Materials Transactions B, 1994. 25(.Tune): p. 333-340. Xiang-peng, L., lie, L., Yan-qing, L., Heng-qin, Z., and Ye-xiang, L., Physical modelling of gas induced bath flow in drained aluminium reduction cell. Trans. Nonferrous Met. Soc. China, 2004.14(5): p. 1017-1022. Che, D.F., Chen, 1.1.1., and Taylor, M.P., Gas Bubble Formation And Rising Velocity Beneth a Downward Facing Inclined Surface Submerged in a Liquid, in 18th Australiatian Chemical Engineering Conference 1990, CHEMECA: Auckland. p. 384-391. Perron, A, Kiss, L.l., and Poncsak, S., Regimes of the Movement of Bubbles Under the Anode in an Aluminium Electrolysis Cell. in Light Metals. 2005. TMS: p. 565-570. Aussillous, P. and Quere, D., Bubbles creeping in viscous liquid along a slightly inclined plane. Europhysics letters, 2002. 59(3): p. 370-376. Das, S., Morsi, Y., Brooks, G., Yang, W., and Chen, 1.1.1., Experimental investigation of single bubble characteristics in a cold model of a Hall Heroult electrolytic cell. in Light Metals. 2011. TMS: p. 575580. Das, S., Morsi, Y., Brooks, G., Yang, W., and Chen, lll, Principal characteristics of a bubble formation on a horizontal downward facing surface. Colloids and Surfaces A: Physicochemical and Engineering Aspects, 2012.411: p. 94-104.

17. 18.

19.

20.

21.

22. 23.

24. 25.

596

and sliding mechanism of gas bubbles under an inclined anode, in 10th Australian aluminium smelting technology conference9-14 October, 2011: Launceston. Cooksey, M.A and Yang, W., PIV Measurement of Physical Models of Aluminium reduction cells. in Light Metals. 2006. TMS: p. 359-365. Yang, W. and Cooksey, M.A, Effect of slot height and width on liquid flow in physical models of aluminium reduction cells. in Light Metals. 2007. TMS: p. 451-456. Wang, Y., Zhang, L., and Zuo, X., Fluid Flow and Bubble Behaviour in The Aluminium Electrolysis Cell. in Light Metals. 2009. TMS: p. 581-586. Qian, K., Chen, 1.1.1., and Matheou, N., Visual observation of bubbles at horizontal electrodes and resistance measurements on vertical electrodes. Journal ofApplied Electrochemistry, 1997.27: p. 434-440. Qian, K., Chen, Z.D., and Chen, lll, Bubble Coverage and Bubble Resistance using cells with horizontal anode. Journal of Applied Electrochemistry, 1998. 28: p.1141-1145. Dorin, R. and Frazer, E.1., Operational Characteristics of Laboratory Scale Alumina reduction cells with wettable cathodes. Journal ofApplied Electrochemistry, 1993. 23: p. 933-942. Utigard, U., Costa, lH., Popelar, P., Walker, D.T., Cool, G., and Hoang, P., Visualization of the Hall-Heroult Process. in Light Metals. 1994. TMS: p. 233-240. Xue, land Oye, H.A., Bubble Behaivour-Cell Voltage Oscillation during aluminium electrolysis and the effects of sound and ultrasound. in Light Metals. 1995. TMS: p. 265-271. Cassayre, L., Torstein, A., Utigard, U., and Bouvet, S., Visualizing gas evolution on graphite and oxygenevolving anodes. JOM, 2002: p. 41-45. Wang, X. and Tabereaux, T., Anodic PhenomenaObservations of Anode Overvoltage and Gas Bubbling During Aluminium Electrolysis. in Light Metals. 2000. TMS: p. 239-247. Aaberg, R.l, Ranum, V., Williamson, K., and Welch, B.l, The Gas Under Anodes in Aluminium Smelting Cells Part 2: Gas Voumes and Bubble Layer Characteristics. in Light Metals. 1997. TMS: p. 341346. Zhang, W.D., Chen, 1.1.1., and Taylor, M.P., Similarity Analysis of Gas Induced Bath Flow in Hall-Heroult Cells. in CHEMECA. 1990. Auckland: p. 1-8. Price, D.C. and Davenport, W.G., Densities, Electrical Conductivies and Viscosities of CuS04/H2S04 Solutions in the Range of Modern Electrorefining and Electrowinning Electrolysis. Metallurgical and Materials Transactions B, 1980.11: p. 159-163. Electrodeposition: Theory and Djokic, S.S.. Practice201O, New York: Springer. 57. Zoric, 1 and Solheim, A., On gas bubbles in Industrial aluminium cells with prebaked anodes and their incluences on current distribution. Journal of Applied Electrochemistry, 2000. 30: p. 787-794.

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

MATHEMATICAL MODEL VALIDATION OF ALUMINIUM ELECTROLYSIS CELLS AT DUBAL Abdalla Zarouni l , Lalit Mishra l , Marwan Bastaki l , Amal Al Jasmi I, Alexander Arkhipov l , Yinko Potocnik2 IDubai Aluminium (DUBAL), PO Box 3627, Dubai, UAE 2Yinko Potocnik Consultant Inc., 2197 rue de Regina, Jonquiere, Quebec, Canada, G7S 3C7 Keywords: Mathematical models, Model validation, Aluminium electrolysis cells, DX technology, DX+ technology Abstract DUBAL model development started with the concept that full modeling capability can be achieved rapidly if public, generic models, based upon commercial software packages are utilized so that all the processes required for the cell design and optimization are included. These models include automatic data transfer between them if required. The software packages used are: ANSYS, PHOENICS, MHD-Y ALDIS with TECPLOT graphics [9] and MARC. In some cases the same process is modelled with two software packages in order to validate model results against each other, thus increasing the model reliability. For example, busbar currents and temperatures are calculated with ANSYS and MHD-YALDIS; magnetic fields are calculated with MARC and MHD-Y ALDIS; MHD is calculated with ESTER and MHDYALDIS. DUBAL modeling capability comprises thermo-electric models based on ANSYS, busbar design models based on ANSYS and MHD-YALDIS, MHD models based on MHD-YALDIS and a combination of ANSYS - MARC - ESTER, mechanical models based on ANSYS and CFD models of cell gas extraction and potroom ventilation based on PHOENICS.

In recent years DUBAL has developed an in-depth mathematical modeling capability for aluminium electrolysis cells, based on commercial software packages, comprising thermo-electric, MHD and mechanical models of the cells as well as CFD models of gas extraction from cells and of potroom ventilation. In order to validate these models a measurement program was initiated, consisting of a group of DX and DX+ cells instrumented for continuous monitoring of cathode lining and potshell temperatures, bus bar temperatures and busbar currents. Moreover, special measurement campaigns were carried out for cell voltage breakdown, heat fluxes, freeze profiles, current distribution, magnetic fields, metal velocities, potshell deformation and cell gas exhaust flow rate. The modeling results showed excellent agreement with measured data, allowing the models to now be used with confidence for new cell designs and industrial studies of existing potlines. In this paper, detailed measurement and modeling results shall be discussed. Introduction

Modeling is not an exact science that can predict cell parameters from theoretical principles only. Measurements of every calculated measurable parameter are needed in order to validate the models. At DUBAL an extensive measurement program was set up, consisting of a group of DX and DX+ cells instrumented for continuous monitoring of cathode lining and potshell temperatures, bus bar temperatures and busbar currents. Moreover, numerous special measurement campaigns have been carried out for cell voltage breakdown, heat fluxes, freeze profiles, current distribution, magnetic fields, metal velocities, potshell deformation and cell gas exhaust flow rate. The measurement techniques are well known and are described elsewhere [10]. In this paper it will be demonstrated how these measurements were used for model validation on DX or DX+ cells.

Mathematical modeling has become a primary design and optimization tool of aluminium electrolysis cells. In the early years of mathematical modeling, the models were based on inhouse software, specifically designed for different processes in the electrolysis cells. These models required large development effort; however, when developed, their usage was low cost. This is why they are still used in some companies, together with newer models based on commercial software packages. The advantage of commercial software packages is that they are being continuously developed and maintained, whereas this is often not so with inhouse software. General commercial software packages also had to be adapted for specific processes and geometry of aluminium electrolysis cells and potrooms, which in addition required considerable effort and time, often many man years per model. Most of these models are proprietary, but some are generic and available on the market. This is the case with ANSYS based 3] and thermo-electric and mechanical models [1 ESTER/PHOENICS (henceforth called ESTER) based MHD models of the cells (ESTER is a specific adaptation of CFD software package PHOENICS for aluminium electrolysis cells) [4 - 6]. On the other hand, there are also specific commercial software packages developed for a specific domain; this is the case of MHD software package MHD-YALDIS that performs all calculations for busbar design, electric current distribution, magnetic fields, steady state MHD and cell stability [7]. A more limited scope software package MARC (Magnetics in Aluminum Reduction Cells) uses specified busbar currents and cell steel to calculate magnetic fields in the cells, in the potrooms or anywhere else in the smelter [8]. Its magnetic field in the cells is then used in ESTER, together with vertical current density at the bottom of metal pad from ANSYS, to model cell MHD.

Electrical Measurements and Model Validation Electrical measurements comprise cell voltage drops, busbar temperatures and busbar current distribution. Figure 1 shows how cell voltages are decomposed. Cell voltage components consists of the following: anode voltage drop from below the anode clamps to the anode bottom, cathode voltage drop from the metal pad to the end of the collector bars and busbar (or external) voltage drop from the end of the collector bars to below the anode clamps. As per DUBAL practice in DX and DX+ cells, two intermediate reference measurement points on the busbars are used: Cathode Reference Point (CRP) on the downstream cathode busbar just outside the tap end collector bar and Anode Reference Point (ARP) on the anode cross-beam busbar at the duct end of the cell. In practice, the voltage drop between the metal pad and CRP is measured (Veal-CliP), then from CRP to ARP (VCIIF-AIIF) and then from ARP to below each anode clamp (Vext-m ,). Next, the

597

voltage between the end of each collector bar and CRP is measured; this is called "Short Drop" in DUBAL terminology. Finally, cathode voltage drop (Veat) and external voltage drop (Vex /) are calculated as per Equation (1):

Short Drop

=

Vcat-CRP -

=

Short Drop +

VCRP-ARP

80

r

(1)

+

V:xt- an

In this or any other measurement path with reference points, there must be no gaps and no overlap at the reference points. Back EMF (BEMF) and bubble voltage drop are calculated from known equations [II]. From all these voltage components, bath voltage drop and anode-to-cathode distance (ACD) are calculated, but these are not used in model validation. Distance along the cell from duct end

Figure 2. Model validation with measured upstream (US) Short Drops in cells at 420 kA. Numbers on curves indicate individual measurements.

ARP

I-.\NSYS

])IS - 1 - 2

8 0 T · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · ................................................................................................................................................ .

20

+----------------------

Figure I. Measurement of voltage drops in the cell. The cathode and anode model calibrations are made with the carbon block-steel contact resistance chosen so that the overall calculated cathode and anode voltage drops are equal to the measured ones. The busbar voltage drops include two contact resistances: one on the cathode collector bar to flex tabs and the other on the anode beam to anode rods contacts (clamp drop). These were added to ANSYS and MHD-V ALDIS model so that the calculated short drop and anode external drop had good visual tit onto the measured values as shown in Figure 2 - 5. Both models agree very well with each other and with measured data.

Distance along the cell from duct end

Figure 3. Model validation with measured downstream (DS) Short Drops in cells at 420 kA. Numbers on curves indicate individual measurements. -ANSYS

-MTID-VAIDISllS

-1

-2

-3

-..\

-5

I

100 - - - - - - - - - - - - - - - - - - - - - _

got------------------------'

A further important design parameter obtained from the thermoelectric and busbar modeling was the collector bar current distribution. Collector bar currents were obtained from measured voltage drops across the cathode flexes and flex temperatures. Modeling showed an excellent balance between upstream and downstream currents of 50.5 % upstream and 49.5 % downstream. This compared well with the measured data of 51.3 % upstream and 48.7 % downstream. Figure 6 shows the model current distribution in the upstream and downstream collector bars, compared to the measured data.

0'···················································· .........................................................................................................................................................

!

Distance along the cell from tall end

Figure 4. Model validation with measured upstream (US) anode external voltage drops in cells at 420 kA.

Busbar currents and temperatures are monitored continuously in a group ofDX and DX+ cells on upstream and downstream cathode busbars that feed the anode risers. Anode riser currents were measured manually. Individual busbar currents and upstream/downstream balance were compared with the models. The difference between models and measurements in individual busbars was smaller than 5 % on the upstream side and smaller than 10 % on the downstream side. The agreement in anode riser currents was within ± 1 %.

Upstream to downstream current balance was within less than ± I % in both, the model and the measured data. The difference between the modelled and measured bus bar temperatures was 0 10°C in downstream busbars and 1 - 15°C in upstream busbars. The largest difference was in the busbars close to the potshell and in the outside anode risers. It became evident that the problem was that only one ambient temperature was used in the models for all the busbars, whereas in practice the ambient air has quite different temperatures around the cells. For example, busbars adjacent to

598

the cathode potshell receive more heat from the potshell, but the outside anode risers are exposed to a lower ambient temperature than the inside risers. Nevertheless, the agreement between the models and measurements is quite acceptable. Temperature differences between the model and the measurement are the cause of the observed busbar current deviations.

Heat flux probes .. IRGUD A Surface TIC I

I""""'"ANSYS

\.4

g 5

=

\.2 1.0

.Q

'5 0.8 .;; -ANSYSDS

:;

-MTID-VAI.D1SnS

.Q

:E

OJ)

>

g

!

0.6 0.4 0.2

(,(j

0.0 50

o

"

100

150

200

250

300

350

400

450

Sidewall shell temperature (OC)

Figure 7. Average side potshell temperature profile near cell centre of three DUBAL cells: ANSYS model compared to measurements with heat flux probes, infrared pyrometer (IR Gun) and contact surface thermocouples (Surface TIC).

!:I 0"-·- - - - - - - - - - - - - - - - - - - - - - - '

Distance along the cell fl1)m tap end

Figure 5. Model validation with measured downstream CDS) anode external voltage drops in cells at 420 kA. Numbers on curves indicate individual measurements. Measured US

-

Measured DS

-

Model US

-

Model DS

l O T · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · .....................................................................................................................................................

a E

For the measurement of cell heat balance, the slice concept is used. This is a measurement strategy applied to a narrow crosssection of the cell, which represents its state at the time of measurement. It has been proven that a limited number of measurement locations on a slice and a limited number of slices on the cell sides and ends can be generalized to the whole surface area of the cell heat balance boundary. The number of measurement points necessary depends on expected variations of heat flux along a slice. As for the cell heat balance boundary, it is common to use one of the two boundaries delimited by System Boundary 1 (over the hood and superstructure) or 2 (see Figure 10). At DUBAL the System Boundary 2 is used for heat balance and System Boundary I as the boundary condition for heat loss and temperature for potroom ventilation modeling.

,

6

4+----------------------

For heat flux measurements. a set of 12 heat flux probes connected to a multichannel data logger was used, supplied by Hukseflux [12]. These were placed along the height of the potshell and on the cradles of three upstream, three downstream slices and of two slices at each end. The same side slices were used over the top of the deckplate and anode cover, including anode rods up to beneath the clamps. From these heat fluxes, the heat loss was calculated from each area represented by a heat flux probe location. The heat balance surface cuts through the anode rods and through the end of the collector bars. Axial heat loss through these cross-sections was evaluated by measuring the temperature gradient along these elements. The overall heat loss was obtained as: (2)

Distance along the cell from duct end

Figure 6. Busbar model validation with measured collector bar currents in cells at 420 kA. Thermal Measurements and Model Validation Thermo-electric models provide temperature distribution in the cells and heat loss from the cell. A group of DX and DX+ cells is instrumented with thermocouples on the potshell surface and in the potlining, which monitor the temperatures continuously. These temperatures were used for model validation. Additionally, several measurement campaigns were carried out for more detailed analysis. In these campaigns, voltage components were measured for internal heat calculation, heat fluxes from predefined surfaces and additional temperatures on the heat balance surfaces as well as freeze profiles. Heat fluxes were measured with purchased heat flux probes. The temperatures were measured with heat flux probes, infrared pyrometer CIR gun) and by contact surface thermocouples. Figure 7 compares the calculated and measured temperatures on the side surface of the potshell, near the middle of the cells. Measured values have an estimated error bar of ± 30 DC, which is the result of instrument accuracy and standard deviation between the slices. The three measurement methods agree well, but heat flux probes give in general lower temperatures than the other two methods, particularly at high temperatures, because, being inside the heat flux sensor, they are actually at a small distance away from the surface.

Where: Q = overall heat loss (kW), qi = heat flux at location = area assigned to the location i (m2), Qi = local heat loss at location i (kW) (the word "heat" is used for

i (kW/m2), Ai

power and should be interpreted as heat per second, klls

=

kW).

Particular attention was given to anode heat loss measurements and modeling. The anode cover thickness was measured on each anode, including an estimation of hard and loose cover thickness. The crust composition was determined in the laboratory and the thermal conductivity was obtained from [13], with some further adjustment of crust composition and crust hardening temperature if needed for good fit to measured heat flux.

599

Heat flux measurements are prone to errors because of possible problems with heat flux probe calibration and because of great variability of local heat fluxes, particularly on the anode cover. A necessary step in the measurement is to compare the measured heat loss (Q) with the internal heat (Qillt), which is the net heat generated inside the heat balance volume, Equation (3):

Qnt =

..

I

1.2

g S

+ X Vc bum + Y V;,O bum)/

(3) Where: Veeii = Cell voltage, Vex! = External voltage drop as per Equation (1), VAl = Voltage equivalent of enthalpy to make aluminium, x Vc hum = Fraction x of voltage equivalent of enthalpy of excess carbon burn (air and CO 2 ), yVco hum = Fraction y of voltage equivalent of CO burn within the chosen heat loss boundary, J = Cell current. VAl includes base electrochemical reactions to make aluminium and all auxiliary processes such as anode butt removal, cavity cleaning, net bath tapping, fluoride feeding, reactions with alumina impurities, etc. All included with no carbon and CO combustion, VAl = 2.09 V for DX+ cells. This value is 2.05 V if 13 % of excess carbon consumption and 10 % CO is assumed to burn and release heat below the crust; this is the same value as for main and back reactions alone, used commonly in the industry. In practice, it is not well known how much excess carbon and CO burn within the System Boundary 2; this remains to be the main uncertainty for the internal energy calculations. However, even with this uncertainty, the internal heat is more accurately known than the one from measured heat flux. Tn principle, in an exact world, the heat loss is equal to internal heat. It is therefore best to normalise Q to Qilrt and then multiply all heat fluxes and partial heat losses by the normalisation factor F = QiniQ. This was carried out in this work. Figure 8 shows such normalised measured heat flux from the potshell sidewall in comparison with the ANSYS model, which was run with the cubic spline fit to the measured freeze profile, shown in Figure 9. As expected the highest potshell temperature and heat flux are in the lower half of the metal pad. Figure 10 shows model validation with the overall heat balance, normalised to the same internal heat for both the model and the measurements. The largest difference between the model and measurements is on the cathode side wall, where the model heat loss is greater than the measured one. This could be due to somewhat thinner freeze in ANSYS than the measured one as shown in Figure 9. Overall, it can be seen that the agreement between model and measurements of potshell temperatures (Figure 7), heat flux (Figure 8) and heat loss (Figure 10) is good and the thermo-electric model is considered to be validated. -

- VAl

...... ANSVS L4

1.0 O.S

" '" £ Q

0.6

Q

OA 0.2 0.0

Heat Flux (kW/m')

Figure 8. Side potshell heat flux profile near cell centre of a cell: ANSYS model compared to measurements with heat flux probes.

II Measured

"

"

-ANSYS -Sidewall

Distance f.-om pots hell

Figure 9. Measured and model freeze profiles. Along anode rod 3.1 (2.6)%

System boundary I

Total anode 37.7 (38.1)% Total cathode 62.3 (61.9)%

;

System \ ; , ; ' boundary 2 ;' ;'

The normalisation of heat loss to internal heat described above is very important It is through this process, during the first measurement campaign that it was discovered the calibration of the brand new heat flux probes was not correct; this was finally also recognised by the supplier who had to re-calibrate the probes, using an alternative method from that initially used.

5.0 (30.4)%

lector bars

]0.3 (9.8)%

Bottom shell

MHD Model Validation

J 0.7 VcrlicaL Measured

"E'

5

= E

Distance along the pot

Figure 14. DX+ deckplate horizontal and vertical deformation along the pot.

CFD Models of Cell Gas Evacuation The purpose of the CFD modeling of gas evacuation is to design the interior gas collection system so that it provides good gas evacuation for all circumstances in cell operation, such as opening the cell doors and the hoods. The CFD model is based on PHOENTCS. It is fully 3-D in order to represent the opening of any hood, but for normal operation and tapping, half of the model

with longitudinal axis symmetry is used. The model consists of gas channels inside the superstructure, under hood space with anode cover, stubs, yokes and rods, holes in the crust below the crust breakers, gaps between hoods and between rods and the superstructure and also includes some of the gas exhaust duct

601

outside the cell. For tapping, the doors and the tapping hole are open. For anode setting, the corresponding hoods are open. Model input parameters are: mass flow rate in the exhaust duct, temperature of air entering the hooded space, surface temperatures of the anode cover, yokes and anode rods as well as the composition and temperature of gases emanating trough crust holes. Model outputs are pressure, temperature and velocity field. Figure 15 shows pressure distribution on a vertical plane along the cell. The 3-D model structure can also be seen. With macros. the gas mass flow rate is calculated from velocities at any specified channel cross-section. This helps validate the model. Model validation measurements were made through five holes in the superstructure which give access to the gas collection channel. Gas velocity was measured with a Pitot tube, and temperature with a sheathed thermocouple. Mass flow rate was calculated from velocities and gas density. Figure 16 shows the comparison between the model and measurements. The agreement is good. This model is considered to be validated.

Figure 15. Pressure distribution on a vertical slice of the model. 90 80 70

Conclusion

DUBAL has built and validated mathematical models of aluminium electrolysis cells. A very meticulous and thorough measurement methodology was used for model validation and experimental evaluation of the DX and DX+ cell technologies. The validity of the models has been proven and the models have already been used extensively for the design of DX + cells and further optimisation ofDX and DX+ cell technology.

[

60

"

50

.§ 40

i

30 20 10 0

Acknowledgements

0

Throughout the model development and measurement campaigns, there have been many people who are not mentioned in the author list, but who made valuable contributions to the success of these projects. First of all, the DUBAL measurement team showed extraordinary dedication and skill to carry out many measurement campaigns and to assure the best quality of measurements. The model development, measurement campaigns and model validation were supported by Dr. Vinko Potocnik; specific model development was supported by Dr. Marc Dupuis for ANSYS, Dr. Valdis Bojarevics for MHD-V ALDIS and by CHAM for ESTER and PHOENICS models. Special thanks go to Prof. Barry Welch, for his constructive inputs on cell energy balance and process representation. References

3 Measurement point

5

6

Figure 16. Calculated and measured mass flow rate in the gas collection channel (percent of total flow in exhaust duct). The measurement points are indicated with numbered white circles in Figure 15. 7. Valdis Bojarevics, Light Metals, (2010), "Time Dependent MHD Models for Aluminium Reduction Cells", Light Metals, (2010), 199-206. 8. MARC was developed by CERCA (Centre de recherche en calcul applique), Montreal, Canada and is available from one of the authors (Vinko Potocnik). 9. TECPLOT is licenced by Tecplot Inc, 3535 factoria Blvd S.E., Suite 550, Bellvue, WA 98006, U.S.A., http://www.teeplot.comi. 10. Vinko Potocnik, "Measurement Techniques for Pot Analysis", TMS 2012 Industrial Aluminum Electrolysis, September 10- 14, 2012, Chicoutimi, Quebec, Canada. II. Warren Haupin, "Interpreting Cell Voltage Components", Light Metals, (1998), 531-537. 12. Hukseflux Thermal Sensors B.V., Delft, The Netherlands, http://www.huksetlux.eom/. 13. Mark Taylor, "Anode Cover Material- Science, Practice and Future Needs", Ninth Australasian Aluminium Smelter Technology Conference, Terrigal, Australia, 4 - 9 November 2007, paper II. 14. B.F. Bradley, E.W. Dewing, .LN. Rogers, "Metal Pad Velocity Measurements by the Iron Rod Method", Light Metals, (1984),541-552. 15. Valdis Bojarevics and Koulis Pericleous, "Solution of the Metal-Bath Interface for Aluminium Electrolysis Cells", Light Metals, (2009), 569-574.

I. ANSYS is licenced by ANSYS Inc, 275 Technology Drive Canonsburg, PA 15317, U.S.A., 2. M. Dupuis, ANSYS-Based 3-D Thermo-Electric Heat Balance Models, Genisim Brochure, ,-"""",",-,-",-.!.!..!.,-""",-"",-,-,-,-"""=,,,,,,. 3. M. Dupuis, "Mathematical Modeling of Aluminum Reduction Cell Potshell Deformation", Light Metals (201 0),417-422. 4. PHOENICS is licenced by CHAM Ltd, Concentration, Heat and Momentum Limited, Wimbledon Village, London SWl9 5AU, England, http://www.cham.co.uk/ . 5. Vinko Potocnik and Frederic Laroche, "Comparison of Measured and Calculated Metal Pad Velocities for Different Prebake Cell Designs", Light Metals (2001),419-425. 6. Dagoberto S. Severo et aI, "Comparison of Various Methods of Modeling Metal-Bath Interface Deformation", Light Metals (2008),413-418.

602

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

PRODUCTION APPLICATION STUDY ON MAGNETO-HYDRO-DYNAMIC STABILITY OF A LARGE PRE BAKED ANODE ALUMINUM REDUCTION CELL Ruan Shaoyongl, Van Feiya 1, Marc Dupuis2, Valdis Bojarevics3, Zhou Jianfei 1, lCHALIECO GAMI, Guiyang, Guizhou, China 550081 2GeniSim Inc., 3111 Alger St., Jonquiere, QC, Canada, G7S 2M9 3 University of Greenwich, School of Computing and Mathematics 30 Park Row, London, SEIO 9LS, UK Keywords: aluminum reduction cell; magneto-hydro-dynamic stability. This paper introduces the MHD stability theory of a pot, and establishes the 3-dimensional calculation model of MHD stability using pot dimensions and busbar arrangement of an actual pot. It also compares the calculations against observations III one particular smelter operating 340kA pots in China.

Abstract The magneto-hydro-dynamic stability of an aluminum reduction cell has an important influence on aluminum electrolysis production. The paper introduces the research theory of magnetohydro-dynamic stability of a cell and puts forward the concepts of "stationary state" and "transient state" of a reduction cell. A magneto-hydro-dynamic stability software is then used to calculate two different cell conditions. The calculated results prove to be consistent with the actual production, which confirms the model validity.

Magneto-hydro-dynamic stability theory Definition of pot production states The large prebaked anode aluminum reduction pot can be mathematically represented into two states under production conditions. In the "stationary state", the pot is under nondisturbance conditions with stable current and without anode change, tapping, anode effect CAE) and breaking feeding. Such state is a kind of ideal state that never really occurs during actual production, since generally the pots are continuously disturbed by normal process operations. In the other "transient state", the pot is under disturbance conditions with current fluctuation, anode change, tapping and breaking feeding which can not be prevented. In short, "stationary state" is a state without any temporal disturbances, and "transient state" is a state where temporal disturbances are present.

Introduction

An aluminum reduction pot consists of a carbon anode, a bath melt, a metal melt and a carbon cathode. The large DC passing through the aluminum busbar, the anode, the bath, the metal and the cathode etc. during aluminum reduction generates hundreds of Gauss strong magnetic iield. This iield interacts with the current in the metal to generate an electromagnetic force which accelerates the metal circulation in the pot. This results on one hand in an efficient way to dissolve alumina in the bath and then reduce it to aluminium. But on the other hand, it causes metal pad fluctuation, so that the current efficiency (CE) is reduced and the energy consumption must be increased in order to prevent excessive metal pad fluctuation. In severe cases, the metal can splash from the pot and cause accidents [1].

Mathematic definition ofMHD stability of pot For the "shallow water" approximation, the horizontal dimensions Lx and Ly are assumed to be much larger than the typical depth H, and the interface wave typical amplitude A is assumed to be small compared to the depth.

The studies of the conditions that generate metal pad fluctuation caused by the magnetic force in the pot are called magneto-hydrodynamic (MHD) stability studies in the aluminum industry. With the development of large pot designs, the pot capacity gradually increases, so the MHD stability has become the core issue in the large pot design, as well as the important index reflecting the merits of pot design. Generally a pot having good MHD stability is characterized by better busbar arrangement, better magnetic field distribution, low voltage fluctuation noise, and better CE and power consumption index.

With the purpose to derive weakly nonlinear shallow layer approximation Boussinesq equations for the wave motion, the terms in the full three dimensional Navier-Stokes equations of motion need to be estimated. Nondimensional fluid flow equations (continuity, horizontal momentum and vertical momentum transport) are respectively [2,3,4,5]: Ok Uk

Therefore, MHD stability studies have practical significance in analyzing the various input conditions and physical parameters that are affecting the pot stability by performing mathematical simulation of the metal pad fluctuations affected by the pot design and process operation. Researchers have kept exploring and perfecting the calculation method of the MHD stability for years in the aluminum industry, yet it seems from the published literature that there are only few experts and scholars who can successfully solve the issue ofMHD stability calculation.

+8- 1

W

=0

(1)

OtUj +UkOkUj+5-IWOZUj

- ojP+Re- 1

+ 0kVeOkU, )+EI,

-S-lo:zP+Re-l(S-24veo:ow+qVeOkW)+Efz _S-1

603

(2)

(3)

When the depth averaging procedure is applied to the horizontal momentum equations (2) we obtain:

0/1) +ukoii, = - o,p(Ho) - Gale; - flu) +Re- 1 0kV,,0kUO, 1

-

2

(4)

2

+Ef]-26EH;o,ioz +0(& ,6 ,&6) A

The momentum (4) and continuity (I) equations for the two fluid layers can be combined in a single nonlinear wave equation for the interface 1;;(x,y,t).

c\; + c\ ': + E(OjJ;) - c5E(t HOJ;l) -c\; o,;C94 jJ + ' :

= (5)

- \PO;CUkOi'I))

Figure 1. Model of a 340 kA pot including busbar layout Initial "stationary" state

Equation (5) is used for the numerical solution of the interface wave development with coupling to the horizontal circulation obtained from the numerical solution of (4).

Before solving in full non-linear transient mode. it is advantageous to solve first the "stationary" state in order to be able to quickly screen out less promising designs, based on design criteria such as the maximum vertical component of the magnetic field (B z ), the maximum horizontal current component in the metal pad (Jy), the maximum metal pad velocity and the maximum deformation of the metal surface per example.

The equations are by definition transient but, depending on the initial conditions and in the absence of further perturbations, could converge to a "stationary state" where the solution is no longer changing when time passes. That "stationary state" solution can be characterized with high or low "permanent" metal pad deformation and by high or low horizontal circulation flow. Physical definition of pot MHD stability From the above arguments, it can be concluded that the study on the pot MHD stability should have two objectives: 1. quickly study the "stationary state" solution trying to identity characteristics of a stable design; 2. generate a perturbation and carry a much longer fully non-linear "transient state" analysis to really check if the cell design is predicted to return to its "stationary state" after such a perturbation. In summary, the issue of pot MHD stability can be defined as follows: disturbances happen on the pot under normal production conditions, or in a special conditions such as when some anodes are removed. The pot can then return to its original state or transit to the new stable state after a certain time. If it does, the pot is regarded as stable under such normal production state or special conditions, otherwise it is regarded as unstable.

Figure 2. Initial "stationary" state metal flow under ACD = 0.045m

Calculated results under normal conditions

Calculation model of pot and busbar layout Using the MHD-Valdis [2] computer software, the relevant parameters regarding pot and busbar layout are input according to a particular form, thus obtaining a model of 340KA pot and busbar layout in an aluminum smelter. The resulting geometry setup of the model is shown in figure 1. Figure 3. Initial "stationary" state metal surface under ACD =0.045m

604

Voltage fluctuation result chart

Interface oscUlations, (a) 0.005 F,·••.··•· .•·.······•·•···... ·"....··••

In figures 2 and 3, the initial "stationary" state of the 340 kA pot looks quite acceptable, with a low velocity symmetric recirculation flow pattern (maximum of 12 cm/s) and a symmetric and acceptable maximum surface deformation (less than 5 cm). Unfortunately, solving only the initial "stationary" state is not sufficient to know if the pot will be stable in operation; for that, the full non-linear transient solution must be solved as well.

o -0.005

E ....·0.01

:r

'!0.015

The pot voltage fluctuation after an initial perturbation is calculated under different anode-cathode distances (ACD) including 0.055 m, 0.05 m, 0.045 m and 0.040 m. The results are presented in figure 4 for three variables, including the corner anode ACD on the downstream duct end, the corner anode ACD on the upstream tap end, and the total pot voltage. The pot voltage fluctuation resulting from metal fluctuation is observed to see whether it becomes stable as time progresses. ACD=£UIS5m ...........

--

- - DHcomerI .

A.

N

fl

"""I

Graph 9 - Potline DC power consumption in kWh/t (monthly values)

Potline Results

The number of pots in operation decreased in average by 3.9 pots due to - longer waiting period for restarting pots due to the need to be prepared for proper preheating - and increase in voltage by 186 m V per pot leading to the total voltage limit of the potline being exceeded and the stoppage of five pots.

The potline results for three months of operation at 285 kA, from September to November 2011, are considered below compared with three months of operation at 362 kA, from February to April 2011. The main average monthly results during the period at 285 kA for the whole potline are - 1,0 % decrease in current efficiency, - 745 kWh/t increase in DC specific energy.

Except for the five pots stopped due to voltage limitation, no more pots than usual had to be stopped during operation at low amperage and the average age of stopped pots increased slightly showing the stability of the operation and correct thermal balance of the pots.

The positive impact of ACD increase and lower magnetic disturbance with lower amperage was not strong enough to compensate for the negative impact of the decrease in aluminium fluoride excess and the increase in bath temperature [4].

Fluoride specific emissions per ton of aluminium were maintained within the expected range after adjustments of gas exhaust flow rate which had to be increased back to a higher level than that set when amperage decreased.

Conclusion Thanks to the period of operation at 285 kA it was possible to adjust the settings and to show that Aluminium Dunkerque pots could operate at an amperage reduced by more than 20%. corresponding to a 16% decrease in plant power.

638

By implementing a major action plan integrating process adjustments and controls managed by the technical team of the plant with the support of AP TechnologyTM expert team, it was found that the potline could be operated properly and the main conclusions that can be drawn from this experience are as follows: - operation at 79 % of rated amperage with satisfactory operating conditions for the pots was possible and sustainable, - EHS performance was maintained in accordance with company rules and compliance to local regulation, - technical results were maintained at an acceptable level with only I % loss in current efficiency, - very few pots (2 %) had to be stopped due to the Substation voltage limitation and no extra pots were stopped due to operation at low amperage. After this experience, contingency plans were reviewed and the plant is now better prepared for another incident ofthis type. The lessons learnt during this incident have demonstrated the robustness and capability of the technology to operate at lower energy and pave the way for flexible operation of AP TechnologyTM pots. References

[I] C. Vanvoren, 1M. Peyneau, M. Reverdy, 1. Bos, The Dunkirk smelter from 216 to 257 kt/year through 10 years of technology creeping and continuous improvement (Light metals 2003, 185189) [2] C. Vanvoren, P. Homsi, B. Feve, B. Molinier, Y di Giovanni, AP35: The latest high performance industrially available new cell technology (Light metals 2001, 207-212) [3] o. Martin, 1M. 101as, B. Benkahla, O. Rebouillat, C. Richard, C. Ritter The Next Step to the AP3X-HALE Technology: Higher Amperage, Lower Energy and Economical Performances (Light metals 2006, 249-254) [4] B. Langon, 1M. Peyneau Current efficiency in modern point feeding industrial potlines (Light metals 1990,267-274)

639

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

"MAXIMIZING CREEPING VALUE THROUGH RIGOROUS METHODOLOGY" I

Benedicte Champel 1, Nicolas Monnet 1 Rio Tinto Alcan - Smelter Technology, Centr' Alp, BP7, 38341 Voreppe Cedex, France Keywords: "Creeping, methodology, modelling" Due to the continuous development of AP TechnologyTMl pots, all smelters using these technologies have experienced creeping in the past years. Nevertheless, as can be seen in Figure I, increase of current has been general within Rio Tinto Alcan in the past 15 years, whatever the pot technology.

Abstract Creeping of an existing smelter is often considered cost-effective compared to the development of a greenfield smelter. It indeed permits an increase in metal production and/or a reduction in operating costs with lower investment compared to a greenfield case. In order to minimize issues during the execution phase, a rigorous and thorough preparation is required, including the identification of the creeping impacts on all smelter units, and the proposal of possible solutions to mitigate them. Rio Tinto Alcan Technology group has developed an integrated approach to identify the impacts of creeping, starting at the smelter level and then extending to the different units. Such a method is now available as an AP TechnologyTM solution. This paper describes the general approach used, and the main tools developed with this aim.

Importance of preparation for creepings

In order to minimize issues during the execution, and ensure capture of all the gains produced by creeping, a thorough and rigorous preparation is needed at all stages of the project. An insufficient preparation can induce a lower than expected investment return due to: Environment, health or safety issues, Partial achievement of the targeted performances, Delays in implementation of the required changes, Higher costs than estimated.

Introduction

The preparation phase of a smelter project can be subdivided into several steps, each one with a specific purpose. Typical project development steps are: Order of magnitude (OoM): build a business case for the project and detine project options, Pre-feasibility: narrow the number of project options to one and refine the estimate from the OoM, Feasibility: confirm and optimize the recommended option from the previous stage.

Due to the ever growing global competition, all producers of primary aluminium have to face stronger cost pressure. In this context, increasing the production of a smelter and/or reducing its power consumption often represents an opportunity to cut the production cost of the metal. The main levers used to increase the metal production of a smelter are: Current increase in the potline (Refs 1 to 7), Expansion of the potline with additional pots (Refs 2&8). Creeping projects often have to integrate additional constraints, such as the need to adapt to a constrained environmental or energy context (Ref 2).

Over more than 30 years in greentield, browntield and creeping projects, Rio Tinto Alcan Technology has developed the ability to support its clients throughout the entire process of their projects. This article focuses on the application of the integrated approach developed by Rio Tinto Alcan Technology for the specific case of the OoM stud/. This approach is used for all Rio Tinto Alcan smelters projects, and has been put into practice for external customers as well.

Methodology

V.MM

The duration of the OoM study lies between one and three months, that include: A preparation phase and preliminary analysis. An on-site mission (typically two weeks), during which experts from Rio Tinto Alcan Technology, working in close collaboration with the smelter team. conduct a

*Mh

Figure 1 : Amperage increase in Rio Tinto Alcan smelters between 1997 and 2012 (PISS: Alcoa technology, EPT : Alusuisse technology, AP18 / AP30 : AP Technology TM)

I AP TechnologyTM is a trademark of Aluminium Pechiney, used under license by Rio Tinto Alcan Inc 2 The same methodology is proposed during the following steps of the project, going into more details

641

review of the existing operation and performance of the different shops, their potential and bottlenecks. The analysis of the collected information and the production of the deliverables.

The advantages of such a tool are multiple: It allows a consistent picture over the whole smelter of the targeted situation to be obtained: data impacting several shops (such as ladle size or anode change cycle) are defined in a common document. By listing main equipment characteristics, organization of operations and performances of the different shops, it helps identifying the main changes induced by the creeping project3 , It allows a common view of the project to be obtained by all stakeholders, It permits to compare different scenarios on a single document, It is a reference document to be agreed with the client and the technology that defines the case( s) that will be studied in the next steps of the analysis.

The approach can be subdivided into 4 main steps (Figure 2).

Ta ble 1: Example of breakdown for the Basic Data List General & local Data Substation Reduction Buildings Pots Pot Tending Equipment Reduction Services Gas Collection and Treatment Carbon Green Anodes Anode Baking Anode Assembly Recycling Casthouse General Services Utilities Raw Materials Off-site Facilities (port, .. )

Figure 2 : Methodology Process map

Step I : Define the target The first step aims at detining as precisely as possible the initial state and the target to be attained. The definition of the initial state requires collecting all relevant information and data concerning existing operations, performances and organization of the smelter for its various shops (performance reviews, monthly reports, shifts organization ... ). Tn order to define the target, customers' expectations and needs concerning the evolution of the smelter performance have to be analyzed in light of: potential constraints to be taken into account, state of the art technologies used in smelters, available technical solutions for the pot technology as well as for the individual capability of the different shops of the smelter, ongoing research and development and industrialization programs (if required and if the development program fits with the expected implementation of the creeping project), organizational benchmarks.

Material Balance The Material Balance presents the annual flows of products circulating between the various units of the plant, set up on a pattern embodying the entire smelter to ensure the consistency. This document gives a clear and accurate image of the real flows of products linked to the process, and entering or leaving each shop of the smelter. It forms the basic specification for the determination of annual flow rates and equipment sizing.

At the OoM stage, several combinations of technical solutions can potentially meet customers' expectations while integrating identified constraints. Further definition and optimization of the options will be carried out in the following stages of the process in order to select the most profitable one.

Step 2 : Assess the impact of operation at the target The second step consists in assessing the moditications required in each facility of the smelter in order to support the operation at the target.

An open and constructive discussion between all stakeholders in the smelter and the technology experts is necessary in order to develop a common understanding of the needs and the levers that can be used. This shared vision is reflected in the Basic Data List.

The expertise and models owned by Rio Tinto Alcan Technology over the different facilities and equipment of a smelter, permit

Basic Data List Rio Tinto Alcan Technology has developed a specific tool in order to structure this first step. The Basic Data List can be seen as the backbone of a creeping project. This document lists the basic data concerning the whole smelter for a set of milestones (at least present situation and target), on a template structured over all shops and departments (see Table 1).

In the case of amperage increase, the main driver of creeping often arises from modifications performed on the pots: new lining (Refs 1, 3, 4, 5, 6), change of anode format (Refs 1, 5, 6), improved process settings (Ref 9) ... The side-effects on the other shops of the smelter of the changes carried out in the reduction area have to be carefully identitied in order to ensure the capability of the full smelter to support the creeping.

3

642

identification of the potential bottlenecks to reach the identified target for all shops. After bottlenecks have been identified, required modifications and possible contingencies to reach the target can be proposed. The Basic Data List is an essential input to this analysis, as it gives for all shops a view of the major changes to be expected with the creeping.

Step 3 : Define the creeping pathway The objective of the third step is to establish how the target will be attained. All types of constraints linked to the path are integrated into a consistent roadmap for the creeping project (Figure 4). Perfurmam:e {A; kWh/tj

Modeling Modeling tools covering all sectors can be run to better quantify the capability of the different shops to support the expected modifications. They can also predict the impact of different contingencies and then allow discrimination between different options to select the most efficient.

Among the available modeling tools within Rio Tinto Alcan Technology group, the most commonly used for creeping studies are (see Figure 3): Paste plant model (equipment sizing, working modes), Anode baking furnace model (dimensioning of furnace tending equipment), Anode handling and storage model (dimensioning of equipment, working rules), Rodding shop model (dimensioning of machines, trolleys, buffers, storage areas ... ), Pot modeling (thermoelectrical and magneto-hydrodynamic balances ... ) Reduction area models (dimensioning of pot tending equipment, work organization), Metal flow (dimensioning of vehicles, ladles, furnaces, casting units, metal treatment units ... ), Casthouse (dimensioning of equipment), Ventilation models (reduction, anode baking furnace. anode pallet storage, covered roads, ... ) Traffic on site (identification of dangerous zones, improvement proposals, safety recommendations).

Time T zero

Target

Figure 4 : Building of a creeping pathway

Some constraints can be external, such as availability of energy, local regulations, logistics etc. The time required for performing the modifications identified for the different shops and equipment is also taken into account: duration needed to get internal approvals, purchase, install and put into operation the required material or equipment. The planned dates of availability of the technical solutions required to reach the target are considered as well. When some levers are not fully validated, the schedule of validation and industrialization is integrated in the analysis. The robustness and operability of transitory phases during the creeping are checked in order to ensure a proper operation of the whole smelter throughout the creeping. Finally, the sequencing of the implementation of the different levers is carefully planned, so that no bottleneck remains when implementing a new technical solution. Creeping involving a lining modification A particular situation occurs when the creeping project requires modifying the pots' linings. In that case, the creeping pathway has to be carefully drawn up to ensure that pots having two types of linings can be run at a balanced operating point throughout the transition from the existing lining to the new one.

The first check to be performed is the compatibility of the present and the targeted pot designs: the operating windows4 of the two linings need to have common operating amperages (Figure 5). If the targeted operating point is too far from the present situation, this might be impossible, and the use of an intermediate lining might be necessary to accommodate the transition.

Figure 3 : Areas covered by modeling activities Risk assessment At this stage a preliminary risk identification is conducted in order to list critical project risks. Mitigation options can be proposed. The list established at this stage is dynamic. It must be completed and discussed with the client teams prior to a formal risk assessment exercise. This risk register will be updated and detailed during the next stages of the project preparation until the elaboration of the risk management plan.

4 The operating window is an essential tool used by Rio Tinto Alcan Technology to define an acceptable zone for pot operation within prescribed design limitations (see Refs 4&9 for more details)

643

Comparison of scenarios At the OoM stage, several paths to reach the target can be built (Figure 7).

Performance

(A;kWh/t)

Amperage

Figure 5: Compatibility of linings on the basis of their operating windows

If the two linings are compatible, the next step consists of analyzing the pot replacement pattern of the smelter. In recent smelters, the pot relining distribution is peaked, while in older ones, the peak has been flattened at each relining: the relining activity can, therefore, be considered roughly continuous (Figure 6). The ramp-up duration will reflect the relining pattern of the smelter: In the case of a recent smelter, the creeping has to be planned in accordance with the relining peak in order to profit from high pot replacement rate and ramp up current rapidly. In older smelters, the creeping duration will be longer as the transition from one lining to the other one will take five to six years. In some cases, shortening pot lives can be considered as an option to speed up amperage Increase. Number of pots restarted per month

Time

Target T zero Figure 7 : Comparison of two creeping scenarios

The comparison of the different possible pathways, and the selection of the optimal one, will take into account: The cost assessment, An advanced financial analysis, The risks, The availability of the technical levers, The sustainability of the transitory phases involved. Step 4 : Estimate the costs The last step of the analysis consists in building a preliminary cost estimate for the creeping. Drawn up at the early stage of a project with a precision linked to that of the technical assessment, it will be refined throughout the project in order to reflect the progress of the technical analysis.

Cumulative number of pots restarted

:w

Young smelter

Main inputs to this estimate are: The technical assessment on the project, Experience and lessons learned on similar (internal or external) proj ects, Budget proposals for some lots, Information received from procurement departments for some pieces of equipment or material. 8

The cost of a project is made of 3 main parts: Direct costs include construction costs. cost of materials and the cost of various equipment items and equipment assembly. This part can be broken down into the different facilities of the smelter. Indirect costs represent all the costs linked to the project execution, which cannot be allocated to a specific element (such as project management, temporary site installations ... ). Contingencies include all unpredicted costs. They are linked to the project maturity level regarding business ownership, scope, planning, engineering and cost estimate basis.

Years

___"J_

#'

#'

#'

#'

#'

"

#'

Figure 6 : Comparison of relining patterns in a recent smelter (top) and in an old smelter (bottom)

644

The interests of the cost estimate are multiple: It permits to identity the highest budgetary parts, that will have to be optimized during the project, It is a decision-making tool, permitting to discriminate between different options or scenarios (Figure 8), It permits to build a business case for the project.

Aluminium Smelting Technology Conference and Workshop, 2011 ). 5 : M. Bugge, H. Haakonsen, O. Kobbeltvedt, K.A. Paulsen, High Amperage operation of AP18 pots at Karmoy (Light Metals 2011, 415-419)

Performance (A; kWh!t)

6 : D. Woodfield, D. Roberts, M. Wilson, G. Forde, 35 years of improvement at Anglesey Aluminium (Light Metals 2006, 231235)

CAPEX (Sit)

7 : V. Mann, V. Buzunov, O. Burkatsky, A. Krasovitsky, 1. Puzanov, Increase of amperage at Sayanogorsk aluminum Smelter (Light Metals 2008,281-285) 8 : M. Bugge, M. Koniar, K. Skladan, M. Stas, Expansion of the potline in Slovalco (Light Metals 2008, 261-265) 9 : L. Fiot, O. Martin, B. Champel, S. Fardeau, P. Bon, D. Munoz, AP40 : the latest of the AP TechnologyTM solutions (Light Metals 2012,703-707)

Time T zero Target Figure 8 : Comparison of cost estimates for different creeping scenarios

Conclusion

To minimize the number of issues and prevent roadblocks during the execution, and to capture savings and benefits, a thorough and rigorous needed at all stages of a creeping project. The end result for the client is a project that fits needs, and delivers the maximum value.

occurrences of all anticipated preparation is perfectly to its

The knowledge of the different facilities and equipment of a smelter, the expertise acquired over more than 30 years on greenfield, brownfield and creeping projects, the application of a rigorous and proven methodology, and the strong technology transfer capacity, make Rio Tinto Alcan Technology the first choice partner to help smelters prepare their creeping projects. References

1 : C. Vanvoren, 1M. Peyneau, M. Reverdy, 1. Bos, The Dunkirk smelter, from 216 to 257kt/y through 10 years of technology creeping and continuous improvement (Light Metals 2003, 185189) 2 : P. Coursol, 1. Cote, F. Laflamme, P. Thibault, A. Blais, D. Lavoie, S. Gosselin, The transition strategy at Alouette towards higher productivity with a lower energy consumption (Light Metals 2012,591-594) 3 : C. Richard, P. Desrosiers, L. Lefranyois, B. Gaudreault, The Alcan's PI55 smelters now operating at I 95kA, A successful assets optimization strategy (Light Metals 2008, 267-270) 4: E.W. Andrews, T. Martin, J. Parkes, .T. Camire, AP24 Trial and Implementation at Tomago Aluminium (loth Australasian

645

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

THE QUICK SHUT DOWN AND RESTARTING OF 291 kA PRE-BAKED POTLINE AT JSC "RUSAL SAYANOGORSK» FROM MAY TO AUGUST 2011 Victor Buzunov i , Andrey Soldatov 1, Victor Mann 2, Aleksandr Pavin I, Vasily Borisov 1, Sergey Zatepyakin I, Evgeny Shcherbakov 3 , Andrey Guzenkov4 I 2

RUSAL ETC, 37/1 Pogranichnikov St, Krasnoyarsk, 660111, Russia

RUSAL, Technical Direction, 13/1 Nikoloyamskaya St, Moscow, 109240, Russia 3

4

OJSC RUSAL Sayanogorsk, Promploshchadka, Sayanogorsk, 655603, Russia

RUS Engineering LLC, a branch in the city of Sayanogorsk, Promploshchadka, Sayanogorsk, 655603, Russia Key works: potline, start-up, curtailment

management of the Company worked closely with the governor of the region and the region and federal authorities in order to eliminate the consequences of the accident

Abstract Sayanogorsk and Khakas aluminum smelters faced with the serious risk of emergency shut down due to the railway bridge destruction used for the raw materials supplying, after the natural disaster on May 7, 2011. In order to prevent the emergency shutdown UC "RUSAL" decided to reduce temporarily the production volume. As a result, the line amperage was decreased significantly; some amounts of old and sick pots were shunted out in each potline, besides the entire line consisting of 179 pots with a current of 291 kA was disconnected within 2 days. After the resumption of raw materials supply the potline was restarted within 2.5 months. This article describes the sequence of events from the decision to shut down the line to the full production recovery Introduction

On May 9, preparations for the restoration of the bridge - data on the configuration of the terrain were collected and recommendations were given.

In the morning of May 7, two pillars of a railway bridge located near the Kamyshta train station (Khakass region) collapsed (Figure I). The bridge had been used both for delivering necessary raw materials to the aluminum smelter and finished products from the smelter. (No one was injured in the u,"""u"m.

Figure 2. Raw material supply by road Due to the accident, the Khakass region authorities declared a state of emergency. At that time, it was thought that the collapsed bridge would affect the schedule of the recovery of the SayanoShushenskaya hydropower station (equipment for the power station was delivered by rail), the East-Beisk coal mine and other enterprises in the region. An ad hoc committee was established. Data were collected 2417. The sequence of activities aimed at eliminating the consequences of the accident and restoring the bridge was determined by the committee. At the smelter, the amperage was reduced. Individual cells were shutdown in each potline and one potline was stopped. The main reason for the curtailments was the shortage of alumina. Also, there were problems with other raw materials (coke, pitch, etc.)

Figure 1. The collapsed bridge On May 9, the smelter specialists were able to organize the supply of raw materials to SAZ and KhAZ by road (Figure 2). For coordination, representatives of the steering committee ofUC RUSAL, representatives of the Eastern Aluminum Division, academicians, and designers came to the Khakass region. The

647

Potline Shutdown

Restart preparations began right after the potline curtailment (in order to be able to restart the shut-down cells when the bridge is restored.)

Two ways can be used for shutting down cells (followed by curtailing the amperage in full.) The choice depends on the amount of time available.

Restart Preparations

If there is time, some preparations can be made in order to shut

Preparations for the cell restart and cell baking were made according to the approved Program/or the Pottine Restart.

down cells in a more systematic way. Due to the fact that there was no time for such preparations (the bridge collapse), the sequence of operations for shutting down cells was as follows:

Based

•

On the day when the cell shut-down took place, no anode change and no bath composition adjustment were made. The alumina distribution & handling system was disabled (the time for disabling the system in each particular cell was defined individually; the amount of alumina left in the bin was taken into account.)

• • • •

Metal was tapped from those cells which were planned to be shut down. The cells were shutdown with no power reduction by short-circuiting the anode to metal following the established electrical safety procedures. The amount of cells shut down by such a method was limited to the number of wedegs at the smelter (80 sets).

• •

Metal was tapped down from 28 - 30 cm 10 - 15 cm from those cells that were not planned to be shut down.

• • • • • • • • •

Figure 3. The shut-down cell

on

the

program, the

following

was

developed:

schedule for disassembling, maintaining and assembling beams: scheduled for manual/automated cathode cavity cleaning; schedule for removing (cutting) metal pads; schedule for patching cells; schedule for cleaning the bus bar and insulation / checking the insulation (anode superstructure, cathode shell, ducts); schedule for cleaning / checking shunts; schedule for manufacturing, assembling and dissembling shunts: temporary power supply diagram for the first group of cells to be restarted: schedule for restarting the silicon controlled rectifier equipment; schedule for raising the amperage; calculation of the voltage required for the potline (during restarts); schedule for preparing baking equipment; schedule for checking and starting auxiliary equipment: GTC, compressed air, cell control cabinets; standards for the raw materials / materials used for baking / restart; metal/bath balance during restart; and schedule for the supply of raw materials, materials, startup tools, tools.

Based on the schedule for preparations for the cell restart, the readiness of cells for the cell restart, the number of units for flash baking, the number of shunts, the voltage generated by the rectifier, and the crane availability, a schedule for baking and restarting cells was made up. The cells to be restarted were broken up into 4 groups.

Then, the bath was tapped (as maximum as possible) from those cells which were not planned to be shut down (followed by shortcircuiting anodes to metal.) The bath tapped was used for the cells in operation. The bath surplus was poured into special containers. The bath remained was poured into the cathode cavity of the cells shut down for relining (Figure 3).

Preparations

After tapping the bath and short-circuiting the anode to metal, the alumina distribution & handling system was disabled.

Bottom Cleaning

Then, the potline was curtailed. After curtailing the potline, anodes were separated from metal. (The metal remained was tapped.)

After curtailing the potroom, a schedule for cleaning the bottom from the bath and frozen aluminum was made up. First, butts were removed (Figure 4) and bath lumps were collected.

648

Figure 5. The cathode before the cell restart Figure 4. Butts removing

Table 1. The criteria for the evaluation of the condition of the bottom Actions to be taken if the No. Criteria Unit Value criteria are not met

Then, the bottom was cleaned both manually and automatically. Manual cleaning included the use of hammers and concrete breakers. After cleaning, the removed bath / Al was put into containers (by shovels) and sent for weighing and then crushing.

Before cleaninll

During cleaning, attention was paid to fillet (peripheral) seams, block-to-block seams, side and bottom blocks (in order not to break them during cleaning.)

1.

For some cells, it was decided not to remove frozen AI. Such a decision was made in order to be able to do metal preheating. RUSAL's specialists had done metal preheating before that (but of less powerful cells.) Therefore, there were high risks that the restart of such cells would not be successful - the bottom covered with Al could have been damaged; it is also difficult to preheat such a big cell with a thick layer of frozen Al up to the required temperature; such cells are not that MHD stable.

2.

Service life, more than Fe content at the moment of curtailment and/or the presence of any periods of operation with a high Fe content (not related to stub / deck plate melting), more than

Cathode Inspection

month

83

to be relined

%

0,30

to be relined

After cleaninll

After cleaning, the cathode was inspected. A team of experts assessed the possibility of restarting the cell without relining. In those cases when patching was required, the team had to determine the scope of work to be done. After inspecting each cell, a record on the condition of the cell was made up. Such a record contained (if required) information on cathode damages and a list of things to be done.

3.

One or more local corrosion zones on the bottom with a depth of not more than

em

13

to be relined

4.

Damaged bottom blocks to be replaced

each

I

to be relined to be patched

649

5.

AI in seams and cracks within bottom blocks

6.

Wear-out of the hole cap of the fillet seam

-

-

(in order to remove Al and restore the integrity of the bottom. If it is not possible to remove AI. it shall be relined) to be patched

-

-

(in order to restore the fillet seam)

Autopsy & Cathode Cleaning

Before installing side blocks, the cathode shell was inspected. If it was required, burnouts were patched by welding metal plates over the burnout from the internal side of the shell. After welding, the defonnation of the shell was assessed. If local deformations were not more than 2mm, the walls of the shell were leveled by using a mortar based on concrete. Then, a leveling layer was dried for at least 2 hours.

Based on the record received, people from the relining shop cut damaged parts of the bottom up to a depth of 200mm. In those parts where damages or cracks were above the collector bar, the cutting depth was not more than 200mm. After cutting, an additional inspection was carried out. No frozen AI, carbide formations and cracks were allowed.

After drying, SiC side blocks were glued to the cathode shell.

If there were defects (such as frozen AI, cracks, carbide for-

Cell Baking & Preheating

mations), cutting continued. Shrinkage cracks in bottom blocks (not filled with AI) formed due to cooling of the bottom after the process of curtailment were not taken into account.

Relined cells were baked according to the current documentation used for the cell (Figure 7).

Bottom patching

Gas preheating was deemed to be over when the temperature (according to the thermocouples used) was in a range of 580 to 900°C.

After removing the damaged parts of the bottom, seams (up 300mm in length) were rammed by using a manual pneumatic perforator. If damages were more than 200mm in depth and 300mm in

length, fillets (inserts) cut from the bottom block were used. The length and width of the carbon fillet (insert) were 80 to 100mm less than the length and width of the defected part (Figure 6).

Figure 7. Cell baking Bath Preparation & Potline Restart A day before the restart, cells in other potrooms were being prepared to be used as donor cells. Figure 6. Bottom patching

Soup cells were chosen as follows:

Autopsy & Side Block Cleaning

• •

In order to repair the side lining. fillet (peripheral) seams and corroded side blocks were demolished (up to the refectory brick wall.)

• •

After removing the damaged side lining, periphery bricks were inspected. The presence of frozen Al was not allowed. If there was frozen Al in fillet seams near the block that was in a good condition, the seam was continued to be demolished.

the service life should be not less than 12 months; cells which are 12 to 36 months old (not more than 1 time a month); cell with a high temperature (or ifthere was a trend for a temperature raise.); and cells with critical cathode and side lining deformations were not allowed.

The following was done with the soup cells chosen: • •

Side Lining Patching

650

no corrective Al fluoride additions; the voltage was raised;

• • • •

(cell tap-out, insulation problems) or problems with the rectifier (it is to provide the required amperage.)

a hole in the crust was made from which it was planned to take the bath; the bath was taken once a shift during 3 shifts in an amount of300 to 600 kg (depending on the cell); after each bath take, a covering material or alumina was added; and in order to control the process of raw material melting, the level of the bath was measured.

The amperage crept step by step. When an amperage of 100 kA was reached, the team checked the shunts. In 2 to 3 minutes, an amperage of 150 kA was reached; then 200 kA was reached. In the next 2-3 minutes, the required amperage was reached. In order to provide the second group of cells with the amperage, the following operations were performed: amperage curtailment up to 0 kA; shunts were removed from the first group and used for the second group; the restart team and the potline personnel were informed of the readiness of cells for the cell restart: and the amperage begun to creep step by step (as previousIy.)

The criteria for the readiness of the bath were the temperature (not less than 970°C) and the CR (not less than 2.40.) 136 tonnes of the bath had to be prepared, including: • •

88 tonnes for the first 4 cells; and 48 tonnes - additional bath for two backup cells.

The restart of the cells with frozen Al (metal pad) on the bottom was performed according to separate procedures. Nine cells were restarted according to such procedures. One restart was not successful. Due to the fact that the cell were MHD unstable (AI splashes out of the cavity), it had to be shut down.

The delivery time (from the soup cell to the cell to be restarted) had to be not more than 15 minutes. The total time for pouring the bath into 4 cells had to be not more than 45 minutes.

Other cells were started up according to the existing procedures. By August 25,2011 (2.5 months), all the cells were restarted.

Cell Restart

Amperage creep up to the pre-emergency amperage

By June 6, 2011, the bridge was temporally restored (Figure 8). Raw materials started to come to the smelter in full.

After the start-up of the first group of cells, the amperage was set at 255kA. Before the start-up of the following groups of cells (2 to 4), the amperage crept up to 239kA (the pre-emergency amperage) (Diagram 1.) 300.0 . , - - - - - - - - - - - - - - - 290.0

+-----------

280.0

+--------

270.0

+----

260.0 250.0

Figure 8. Temporary bridge

240.0

+-------------------

230.0

+--.,----,--,---,--...,.--..,...--.---.....---...--,

Diagram 1. Amperage, kA

On June 8, 2011, when the rectifier equipment and the bus bar were ready, the first 4 cells (807, 821, 707, and 720) were started to be baked. The preheating (baking) duration was 48 hours.

The possibility to raise the amperage was determined based on the exiting criteria.

The mentioned cells were chosen based on the following: no cathode deformation; 3 to 36 months old; the cells were prepared (patched), their location (it had to be easy to transfer and pour the bath.)

As a result, the pre-emergency amperage was restored in 4.5 month after the restart.

In 12 hours, another 4 cells were started to be baked (801, 803, 701, and 703.) These 4 cells were going to be used as backup cells. (They can be used if there are problems with the first 4 cells

Based on the results of the inspection performed, 40 cells were relined. 139 cells were restarted.

Discussion

651

On August 31, 2012, 43 restarted cells (from 139) were shut down for relining. 96 cells are still in operation (Figure 9). The cells have operated for at least 14.17 months since the restart. They continue to operate (their average service life as of August 31,2012 is 40.67 months.)

fore, the restart (after emergency shut-downs) can be deemed to be technologically and economically successful. Acknowledgment The authors recognize that the results have been achieved due to the work of hundreds of RUSAL's specialists who supported the project in different directions. So, we express our deep gratitude to them.

Figure 9. Potroom

The distribution of the number of shut-downs depending on the service life of the cell after the restart (as of August 31, 2012) can be seen on Diagram 2.

0·1

1·2

2·3

3·4

4·5

5·6

6·7

7·g

g.g 9·1O 10·

11· 12· 13· 14· 1.3 14 15

tart Diagram 2. The distribution of the number of shut-downs depending on the service life of the cell after the restart (as of August 3 I, 2012)

Nine cells (6.5 percent) were shut down right away. Their service life (after the restart) was less than I month. Other cells operated for at least 4 months, i.e. the restart without relining was economically successful. The average service life of the Potline 4 cells before the accident was 61.5 months. After one year of operation, it is safe to say that that the average service life of the restarted cells will be not less than 50 months. Conclusions

The potline was shutdown in 2 days. It took one month to make the potline ready for the restart. It took 2.4 months to restart it and 2 months to reach the targeted indicators. Taking into account the dynamics of shutting down the restarted cells, the service life expected is not less than 50 months. There-

652

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

PRODUCTION GROWTH AND FUTURE CHALLENGES IN ALUMINIUM BAHRAIN (ALBA) Isa AI-Ansaril, Abdulla Habib l, Mittal A.C. l, Nabeel AI-Jallabil 1 Aluminium Bahrain (Alba) Keywords: Production, expansion, current creep, current efficiency, pot life. Abstract

Aluminium Bahrain B.S.C. (Alba) consistently ranks as one of the largest and most modern Aluminium smelters in the world. Known for its technological strength and innovative policies, Alba enforces strict environmental guidelines, and is widely regarded as one of the top ten performers on a global scale. Commissioned in 1971 with a capacity of 120,000 tons per year, Aluminium Bahrain has steadily progressed and today is one of the world's top performing and largest Aluminium producers worldwide. Production output has increased stepwise since start-up through numerous technology improvement projects and several major expansions using latest available technology in the market. The current plant capacity is close to 890,000 tons per year. The main strategy adopted by Aluminium Bahrain since the early days has been to sweat the assets by maximizing the production of the electrolytic cells using latest technology, upgrade of technology, expansions with sustaining cost effectiveness position through different programs, maintammg lowest impact on the environment, developing nationals along the way aiming to zero harm as a core principle of the business. This paper describes the strategy adopted by Aluminium Bahrain highlighting the challenges encountered to achieve the key milestones along with the future plans. Introduction

During the years, Alba had always selected the state of art technologies during the expansions as discussed below. The first expansion was commissioned in 1981 where a new potline-3 was added. The new potline-3 initially had 152 side break end to end pots using the Kaiser which was the latest technology at that time and was compatible and had a lot of synergy with Reduction Lines 1&2. Reduction line 3 pots were initially operated at 115 kA which increased the plant production capacity by 50,000 tons per year. Eight years later, (1989) another 76 pots were added to Reduction line 3. Reduction line 3 was expanded further in 1996 where another 76 pots were added. So presently Reduction Line-3 has total of 304 pots. In 1992, there was a significant shift in Alba approach toward moving ahead with the expansion and increasing the metal production throughput. Hence, Alba decided to take a giant leap by doubling its production. The newly constructed (Reduction line 4) used the state of art Pechiney (AP30) technology. Reduction line 4 consisted of288 AP30 pots initially started at 295 kA which aimed to produce 235,000 tons per year In 2005, reduction line 5 was built using AP30 technology being the longest reduction line at that time with a total of 336 AP30 pots. Reduction line 5 Pots started up in strategic way in the shortest duration of reduction line start up ever that taken only 77 days to bath up all the 336 AP30 Pots. With this expansion, Alba metal production reached new height adding 305,000 ton per year to Alba total metal production to 830,000 ton per year.

With the objective to diversify the economy of Kingdom of Bahrain from oil based along with utilizing natural gas as source of energy and creating employment opportunities for Nationals, Aluminium Bahrain (Alba) was commissioned in 1971. Aluminium Bahrain (Alba) started as a joined venture of Bahrain Government, Saudi Government and Briton at a rated capacity of 120.000 tons per year. Alba started its metal production with commissioning two reduction lines namely Reduction Lines 1&2 using side break un-hooded, end to end Montecatini technology at line amperage of 100 kA. The two reduction lines contained a total of 456 pots distributed on four reduction rooms. Alba's first expansion was in 1982 when Reduction Line 3 was built followed by another expansion in Reduction Line 3 in 1989. The giant step was taken in 1992 when Reduction Line 4 was constructed and finally Reduction Line 5 was built in 2005. Since the beginning, Alba had its own carbon plant where carbon anodes are produced and its own power station to generate the required electrical Power for the reduction lines. On the other hand, Cast House was made to produce different types of finished products starting from standard ingots, T ingots, rolling slabs and extrusion billets. The present Alba production capacity (2012) exceeded 880,000 tons per year. This paper describes the past, present and future of Aluminium Bahrain in improving the smelters productivity, energy efficiency along with remaining as one of the lowest cost competitive curve.

Amperage Creep and Improving Reduction Cells Performance Increasing production through amperage creep and process improvements with objective to increase cells Current Efficiency are considered to be a less costly option as compared with Greenfield expansions. However, this option remains to be more challenging as it involves sweating the present assets and adopting a significant change to the process which might lead to catastrophic failure if the risk is not properly mitigated and precautions not taken. Hence, a systematic approach with extensive planning are required along with a well trained staff to execute the plans. Reduction Lines 1-3 The side worked end to end pre-baked anode technology cells were progressively installed in Alba pot lines 1-2 and pot line 3 during I 971 - 1981'. Pot lines 1-2 and line 3 were initially operating at around 105 kA and 115 kA respectively achieving a current efficiency of around 87%. The performance was gradually improved and by 1991, pot lines 1 & 2 and line 3 were operating at 112 kA and 123 kA respectively, with the current efficiency averaging around 90% in both lines 1-2 and line 3. In the period 1992 -1995, the three pot lines were retrofitted and changed over from side break to point feed cells, which included installation of alumina and aluminium fluoride feeders controlled using

Major Expansions in Reaching a capacity of 890, 000 TPY

653

individual cell controller, gas collection system and changing anode setting pattern. After retrofitting, the line current was gradually increased to 128 kA (lines 1 &2) and to 142.5 kA (line 3) as shown in figure 1. As a result of the change over from side break to point fed technology, the current efficiency improved and maintained at around 92.5% till year 2004. Since 2005 significant improvement in current efficiency has been achieved as shown in figure 2.

Reduce the maximum and increase the minimum additions of Aluminium Fluoride. Reduce the slope of the curves for less aggressive additions of Aluminium Fluoride. Introduce limits on the cumulative deviation from the cell's average consumption of Aluminium Fluoride.

Average line Current Trend 145 . - - - - - - - - - - - - - - - - - - - - - - , 1 140 135

+--------+--------

130 + - - - - - - - - , 1

a.to

125

11S 110

§ § N

N

8 N

§ N

g

rl rl

N

N

The aim for these changes was to stabilize the variation in Aluminium Fluoride addition and enable better cell control since deviations outside the control band are usually due to deviations in other operational routines.

o

Figure 1: Line Current Trend in Reduction Lines 1-3

Introduction of New Alumina Feeding Model Alumina feeding was based on an adaptive feeding model. The characteristic nonlinear relation between the alumina concentration and the pot resistance was modelled where the estimated slope of resistance versus alumina concentration (parameter b 1) gives information about the concentration of alumina. This information was used to control the alumina supply to the pot. In the old model, the average overfeed time was in the range of 60 to 90 mins with a maximum limit of 120 mins. This was effectively increasing the alumina concentration in the bath. The ratio of underfeed to overfeed was extremely irregular which was due to the back feeding and sludge formation. The program didn't take in consideration the changes in Alumina dump weights, the mount of Alumina feed due to self feeding and variation in Alumina solubility. Figure 4 shows considerable variation in under feed times resulting only in 8 feeding cycles in 24 hours.

Average Current Efficiency Trend 955 . - - - - - - - - - - - - - - - - - - - - - , 95.0 94.5 94.0

G'

ii

93.5 93.0 92.5 92.0

+-----------

·.G.... 9L5

ffi

91.0 " 905 g 90.0

(5

D$$ C

Figure 3: Original and modified AIF3 addition strategy

" 120

u

89.5

Figure 2; Reduction Lines 1-3Current Efficiency Trend Improvement in workmanship quality through more dedication, training and sharing the ownership with the employees, along with followings is the major changes factors contributed to the success; Introduction of modified Thermal Control Model. Introduction of new Alumina Feeding Model Pot lining design changes and Voltage review. Modified Thermal Control Model In the old thermal control model there was over emphasis on correcting the high bath temperature by excessive Aluminium Fluoride addition and expecting fast response as shown in Figure 3 below. This approach has following errors. It does not allow for the operating band that is inherent in the process (the oval of the Figure). It assumes constant bath volume. It assumes the difference in bath temperature is due solely to aluminium fluoride concentration. Rarely any ofthese assumptions are valid and there was a need for major change in the philosophy as follows: Allow for the dead band as indicated by the dashed line graph in figure 3. (This illustration is schematic only; the actual algorithm is non-linear but similar.)

Change Over To Fixed Alumina Overfeed Strategy An attempt was made in year 2005, to change over from adaptive alumina feeding control to fixed duration over feeding - alumina feeding strategy. Implementing a new alumina feeding strategy in the existing 20 years old control system hardware, with limitations in the communication system network between the pot controller and the central computer was a challenge. Keeping this in mind a simple fixed over feeding alumina feeding strategy was chosen in

654

preference to the advanced alumina feeding strategies evolved and adopted in modern Aluminium smelters. This proved to be very successful in reducing the anode effect frequency by more than 50% and improve the alumina concentration control in the cell as evident from the increase in the feeding cycles ( change from underfeed to overfeed) per day from around 6-8 cycles to around 13 cycles

protective ledge. The need to address the pot life issue has also been highlighted by the fact that the dimensions of the key components of the cell indicate that further increase in line current should be possible. While significant gains have been made in the current efficiency (reducing the surplus energy by approximately 5.5 kW) the actions taken to achieve this are not the focus of this paper, which is confined to design and heat dissipation issues.

Design Changes & Voltage Review of the Individual Pots The followings changes in 2004 have resulted in a total voltage saving of more than 100 mV and have contributed towards stable side protection ledge and improved cell voltage stability. - Increasing the depth of the anode stub hole from 100 to 120 mm - Introducing single slotted anodes of depth 150 mm. - Increasing the collector bar cross section by 28%. - Composite high thermal conductivity SiC/Carbon sidewall

Strategy for Achieving a Cell Life> 2000 Days The initial focus in 1998 was to extend the cell life to > 2000 days. Short term plan focused on increasing the side ledge thickness through improving operating efficiency. and increasing the heat dissipation through increasing metal level, reducing anode top cover, reducing side and bottom insulation. The trend of Pot life is shown in 5 Average Pot Life Trend 2500 " , - - - - - - - - - - - - - - - - - 2300 2100

+--------------------------------------------------------------------------.

To improve the superheat and to reduce the voltage fluctuation of individual pots over long period, the target resistance was increased by around 0.3 micro ohms (around 40 mV) without significantly affecting the overall voltage of the pots. The overall performance of the pot line before and after implementing the modified lining design, thermal control and alumina feeding strategy is summarized in Table 1 below; Reduction Lines 1-2

III

C1700 1500 1300

1100

Reduction Line 3

1-2

Year 2010 Year 2010 Year 2004 Jan - Oct Year 2004 Jan - Oct PARAMETER Line Current Cell Voltage Current Efficiency Specific Energy Anode Effect Frequency A;erage Age of failed cells

_Line-3

Figure 5, Trend of Pot Life in Reduction Lines 1-3

Before the After the Before the After the change change change change kA 124.4 124.8 137.5 138 V 4.66 4.67 4.66 4.69 % 92.8 94.7 92.6 94.7 kwh/kg AI 14.95 14.70 15.00 14.77 0.33 0.19 0.39 0.18 2381 days 1769 1822 1828 UNIT

Details of some of the significant improvements or changes which contributed to this increase in cell life of between 600 to 900 days over a period of 10 years and their impact on heat balance are given below and summarized in Table 3.

Table 1: Companson of the overall performance of the Pot Ime 12 and Line 3 before and after modifications

Lines 1-2 cells

Line 1-2 in kW Pre 1995 1992 2000 112 120 552 513 483 517

After 2007 125 581 544

Line 3 in kW Pre After 1992 2007 123 138 558 638 522 593

Line Current kA Gross cell \energy Net Electrical energy input to cell Electrical energy 227 248 263 249 used for process 281 273 Electrical energy 256 269 available for heat (+ 13) (+25) dissipation Table-2; Impact of Changes on Cells Heat Balance

Line 3 Cells

Increase in heat loss from the cell ;:::-5 kW - Welding cooling fins on the pot ;:::-4kW shell ;:::-2 kW - Increase in metal level by around 2 ;:::-2kW cm ;:::-9 kW ;:::-6kW - Reduction in the anode top cover ;:::-3kW ;:::-4kW - Reduction in bottom insulation ;:::-3kW ;:::-3 kW - Reduction in side insulation ;::: -23 kW Total increase in heat loss from the ;:::-18kW cell Mismatch m heat balance with +25 kW +30kW Retrofitting & increase in line current ;::: -5.5kW ;:::-6kW Improved control & CE % ;::: 1900 ;::: 2350 Cell life (days) Remarks: Further adjustment through bigger collector bar and high thermal conductivity insert taken up in next phase from 2007 onwards. Table 3; Impact oflmprovement & Changes on Cells Heat Balance

Improvement in Pot life The cell heat balance comparing the situation at line currents of 112 kA. 120 kA & 128 kA for line 1-2 and 123 kA and 142.5 kA for line 3, as expected, highlighted the increase in the amount of heat for dissipation from the cells, a consequence of various process changes as summarized in Table 2.

Period

+---T---t---'---T---r---r----r---r---y----r---t---'----T----y----r----r---r----r--T---r---l----T---T---li

290 303 (+30)

Strategy for Cells Life to > 2500 days Along With Increase in Line Current As the cells were not operating at their full productivity potential a detailed heat balance study and modelling was conducted in 2005 to enable further line current increase without compromising

Not only does this explain why there is a predominance of sidewall failures, but it also highlights the need to improve process efficiency (thus increasing the process energy utilization) and sidewall heat dissipation to increase the thickness of the

655

on cell life. In the long term plan of improving the cell life further included: • Increasing the diameter ofthe anode stub, • Increasing the cross section of the collector bars • Using Silicon Carbide sidewall with high thermal conductivity carbon inserts. • Installing vertical cooling fins outside on shell.

The followings were carried out to enable the cells to cope with the extra heat inputs and improve the current efficiency in Line-4; • The major current increase was started on 1996 where second generation cells with graphitised cathode blocks introduced and the cell voltage was reduced by around 30 mvolt due to lower cathode resistance. • The target bath level of the first generation pots was reduced by I cm (from 19 to 18 as first step then to 17 cm) with the objective to reduce the risk of tap out and maintain good pots metal purity. • The anode height was increased from 600 mm to 618 mm starting from week 48/1996 in order to compensate for the decrease in butt thickness with increase in line current. This is in tum to eliminate the negative impact of the amperage increase on the pots metal purity. • Anode length increased from 1450 mm to 1500 mm during 2001 which gave a saving of 100 mvolt • Alumina feeding intervals were modified to increase the alumina feeding to the pots as the pots are acquiring alumina due to higher metal productivity Double Anode slots of 150 mm per block were introduced in • 2003 which gave a saving of around 50 mvolt. Then it was increased to 200 mm in 2009. The cells were squeezed more than their original design in • order to cope with the current increase. A precise operation has been required to maintain good operation and maintain high efficiency.

Since 2007, cells are being installed with the required increased cross section of collector bar and composite high thermal conductivity sidewall. Reduction Line 4 Performance of Reduction Line-4 is considered to be one of the best among AP30 smelters and considered to be a bench mark in term of the highest current efficiency, lowest energy consumption and one of the best metal purity produced. This has been achieved through the years by outstanding workmanship quality along with the followings, modifications which will be discussed in details; - Amperage Increaser along with changes associated - Performance and Modification of cells lining through the years - Modification in Alumina feeding control system Amperage Increase in Reduction Line-4 Line 4 Current was increased from 295 kA in 1992 to 344 kA in 2012 with maintaining high current efficiency as shown in figure 6 &7 below.

Performance and modification of cells lining through the years in Reduction Line-4.

line Current increase & net votage line-4 350

•

-

Cuurent

The start up of the first generation pots in the line commenced on 17-05-1992. The start up of the 288 pots in the line was completed smoothly on 2-12-1992 without any significant abnormalities. The start up & normalisation of the first generation pots was an example of the harmonious blend between Alba experience & the state of art AP30 technology. Despite all difficulties encountered through the years, pot life in Line-4 continued to increase as shown in fi ure 8 below; Average Pot Life Trend 2200 2000 1800 1600 3,400

-Voltage

.,..---------------------,!

Figure 6: Trend of Line Current Increase vs Net pot voltage 800 600 400 200

Average Current Efficiency Trend 95.5 95.0 945 94.0 93.5 93.0 925 '" 92.0 91.5 w 91.0 1;1 90.5 90.0 U 89.5

.,..--------------------,!

o Avg. PotUfe Achieved

Figure 8, Trend of Pot Life in Reduction Line-4 The upgrade of lining designs to cope with the current increase along with potlife achieved from first generation till now is discussed in table 4 below;

!l

Gen

# 1

Figure 7: Current Efficiency Trend in Line-4

656

Avg. pot age 1649

Comments

Shells lined with semi graphitic cathodes. Carbon side wall and rammed with hot ramming paste. Collector bar window sealing

phase to continue but with very low alumina concentration leading to anode effect.

only from inside. 126 pots (43%) cut out prematurely due to side wall tap out. Carbon side wall found oxidized due to air ingress from collector bar windows. Pots restarted after replacing sidewalls. stuffing box was fixed and filled with KUB concrete to seal the area/stop air ingres. 2 1646 Initially shells were lined with semi graphitic and graphitic cathode blocks, SiC side walls and hot ramming paste from . U se of stuffing box started to stop air ingress from collector bar windows. In Sept 1996 graphitized cathode blocks using warm ramming paste. In 2 nd generation pots ware failed mainly due to high iron caused by cathode erosion About 80% of pots lined with graphitized and 3 1589 improved quality of graphitized cathode blocks. Main reason for achieving lower pot age is that during the year 2005 significantly high numbers of pots (124 pots) in line 4 failed at lower age due to power outages taken to put in service a swing rectiformer between Line 4 and 5 and deficiency in operation control All pots of 4th generation were lined with 1755 4 graphitized or with improved quality of graphitized cathodes. 12 pots had to cut out at average age of 800 days mainly due operational problem after 4 hrs of D.C. power outage in Aug 20 10. Test group of cathodes with 490mm ht was installed and achieved avg. pot age of 1929 days. From 2006 onwards composite side wall were introduced in line 4. 565 The following changes were carried out in 5 (6% Second generation cells as preparation for failed) current increase. - Cathode height increased from 450mm to 490mm - SiC side wall was replaced by composite side wall (SiC + C) Collector bar cross section change from 150 x 100mm to 122 x 122mm Table 4: Changes In lining desIgn In ReductIon Llne-4

Phase-2; The resistance during phase-2 doesn't become lower than the target in spite of the higher overfeeding rate which ends with increase in resistance during the next underfeed and causing anode effect. New Alumina Feeding Control System The followings are the main feature of the new control system; - Calculation of the slopes and the changes in the resistance is done all the time both in underfeed and overfeed. - The decision to switch to overfeed depends not only on the slope value but depend on the difference in resistance and also on the average resistance at the end of phase-I. This makes the system response faster before anode effect occur. - The resistance is filtered in order to determine the actual alumina feeding concentration from just the noise because ofthe current signal or others. - The feeding reduces by itself during the underfeed when the resistance stays in under feed for a long period. Performance of the new Alumina feeding Control System - It could be seen from the figure 9 below reduction to anode effect frequency about 0.10 AE/pots/day when the new software

Figure 9; Trend of Anode Effect Frequency in Line-4

Modification in Alumina Feeding Control System in Line-4

Reduction Line 5 The followings are the main features in Reduction Line-5 SInce the start up till dates; Current Increase with minimum major expenditures Improvements in Pot life along with upgrade of Lining design

Due to the higher anode effect frequency in Reduction Line 4 (0.25 A/E/pot/day), a study was conducted to identify the shortcoming the feeding model and come up with necessary corrective actions. The shortcomings in the old alumina feeding program are summarized below; End of Phase-I: Sudden increase in the resistance at the end of underfeed phase without the ability of the system to detect such increase earlier. This leads the pot to go on anode effect

Current Increase in Line-5 Maintaining high current efficiency while increasing the current has been the main challenge in Reduction Line-5. Reduction Lines is considered to be one of the best AP30 plants in term of highest current efficiency and lowest energy consumption as shown in figure 10:

Interruption of Phae-I: During the underfeeding period, the target resistance could change by modifying the components of the target resistance. This leads to disturbance in Phase-l and lead to lose the time to follow the evolution of the resistance. Slope below Critical: Due to unrepresentative of the slope calculated at the end of Phase-I , the slopes resulted from the calculation is lower than critical which allow underfeed

657

Conclusion & Future Challenges

line Current increase & net votage line-S 350 350

1 40 330 320 1..0

_

r--.

00

Cuurent

m

9

zi

Since the commissioning of Reduction Lines I & 2 in 1971, Alba adopted the continual improvement and growth strategy. It is obvious that the adopted approach was successfully implemented which had taken Alba production from around 120,000 tpy in 1971 to about 890,000 tons per year in 2012 (as shown in Figure12 below) and planning ahead to cross 1.2 Millions tones per year in the comi few

4.30 4.28 4.26 4.24 4.22:: 4.20 g 4.18 4.16 4.14 4.12

-Voltage

Figure 10: Trend of Line Current vs Net Voltage in Line-S The following were done to cope with this increase in current; - Increasing anode length from 1500 to 1530 mm (saving of 35 mvolt) and Increasing anode height from 620 mm to 650 mm in order to maintain high metal purity without decreasing the anode cycle - Increasing the anode slots from 150 mm to 270 mm (saving of 50 mvolts) - Reducing the ACD along with putting more control and tight operation to keep cells stable and maintain Current Efficiency.

Figure 12: Hot Metal Production Trend in Reduction Lines 1-5 Alba Future is promising further and followings are being planned; Operate Lines 1-2 above 140 kA after increasing the anodes length and anodes stub diameter along with modifying anode beam and lining designs Operate Line-3 above ISS kA after increasing anode length and anodes subs diameter along with modifying the lining designs. Introduce Alumina feeding to the cells via Hyper Dense Phase System instead of using Alumina point feed vehicles. Operate Lines 4&5 above 400 kA after enlarging anode size and modifying lining designs with minimum specific energy requirements As both of Level-I and Level-2 cells control systems are obsolete, a program has been initiated to upgrade and retrofit the existing systems with objective to improve cells performance There are many spare capacities within the plant made available for the provision ofLine-6 since the start up of Line-S in 2005. Alba is in the stage of selecting the best and latest technology, which maximizes the plant production beyond 1.2 million tons per year with attractive return of Investment.

Performance of Lining in Reduction Line-S The following Figure-II shows the performance of the cells lining since the start up. It could be seen that the average of the cut out pots has increased since the start up. Average Pot Life Trend

:-E Bath depth 0.34 0.46 -0 -0 Added liquid bath 0.73 0.33 « Daily doses of Al z0 3 -0.18 -0.40 0.14 Number of point feeder's action 0.66 Current intensity -0.71 -0.59 > 0.61 0.68 tl.O Cell's resistivity .... Cell's voltage 0.57 0.59 QJ C UJ Bath temperature -0.04 0.32 Silicon content in molten Al -0.43 -0.66 Noise -0.26 -0.20 II) .... Total number of anodic incidents -0.25 -0.42 QJ ..c ...... Iron content in molten Al -0.43 -0.66 0 Total number of anode effect -0.08 0.25 *Values III bold are wlthlll a 95% confidence marglll accordlllg to the number of points considered [27]. Values are listed between -I and I. Extremums correspond to a perfect correlation and 0 is equivalent to no correlation. II)

(4) Thonstad et al.[7] indicated that the alumina concentration had no important relation with the sodium and Haupin [9] quantified it to be very small, but negative. Our results indicate that more doses of alumina negatively correlate with the sodium level in aluminum. Finally, when the point feeders are activated, they push the alumina more rapidly into the cell than a normal dosage by gravity. When dosed too fast, part of the alumina can accumulate at the bottom of the metal pad as sludge. The Na20 content ofthis sludge can react with the aluminum to form alumina and dissolved sodium in the metal pad according to eq. 5. This could explain the strong correlation observed between the number of point feeder's action and the sodium content of the metal.

Additives Dissolved in Molten Bath An increase in the fluoride content showed no interesting results from a statistical point of view even though theory indicated that a relation existed with the CE% and the Na(Al). The divergent results are explained by the daily variations of the cells. Only 20% of the pots are analyzed daily for the bath ratio. This low number can hardly correlate with 100% of the CE% value as it does not consider the daily variation of the remaining pots. Results according to the total number of daily doses are recorded daily and a weak correlation can be observed. A negative correlation with CE% and sodium dissolved in aluminum is

(5) Energy Transfer in the Electrolysis Cell Increasing the line current intensity has a direct impact on the cathode current density. Theory[12] indicates that a positive correlation should be observed by increasing the current density. However, a strong and negative correlation has been observed from the smelter's results. As mentioned previously, the ACD is very small at AAl. Therefore, a small increase in intensity

743

obtained by diminishing the ACD will have a more important impact on the bath-metal interface stability than it would have at a higher ACD. This difference will cause the negative correlation to overcome the positive increase we could observe in both cases. The cell voltage and cell resistance are highly linked to each other in an electrolysis cell because the intensity is maintained on a target. A strong positive correlation can be observed for the current efficiency and the sodium content of the aluminum. If these parameters are at higher values, it is plausible that the ACD is larger, causing a reverse effect that one observes when increasing current intensity. On the other hand, with this increase, the chances for eq. 6 to happen, as described by Welch and Tabereaux[28], are higher and would favor the sodium dissolution.

This region has the highest probabilities to be in direct contact with the aluminum. Tron content in metal usually comes from anode stubs which are exposed to molten cryolite.[31, 32] No correlation was expected nor observed with the sodium. However, as described by Sterten and A1.[23] from laboratory experiments, the iron had a negative correlation with the CE% (0.23±0.04 % per 100 ppm Fe(bath))' Results from AAI in production cells show a decrease of 0.17% per 100 ppm FeCAl) with regard to the CE% It is explained by Sterten and al. that the different states of Fe ions produce a loss in the efficiency by changing constantly. No particular correlation is applicable between the total number of anode effects and the current efficiency. An overview of the data indicates that a small number of cells have contributed to a large number of anode effects. These few cells caused an important increase in the number of anode effects, but the real impact on CE% is diluted in the total number of cells considered. A small positive correlation is observable between the anode effects and the sodium content. When an anode effect occurs, the cell voltage will considerably increase. It is plausible that the reaction from the eq.6 occurs during this erratically high voltage. However, anode effects generally last less than a minute and the impact ofthis reaction should be less than what is observed.

(6) In opposite to results from Tarcy and Sorensen [17], the bath temperature had no impact on the current efficiency. It is supposed that temperature variations between the cells caused the average temperature to be non-representative. A weak correlation with the sodium content can be observed but the statistical analysis is not sufficient to explain this correlation. A negative correlation has been observed for the silicon content in aluminum. The CE% correlation is easily explained as silicon is directly linked to the cell power. The cell power can easily be unbalanced when a lot of reoxydation occurs. This phenomenon generates heat in the cell and increases the superheat causing the sidewalls to melt. Hence, silicon is a consequence of a low CE% and not a cause. It is supposed that the strong correlation between silicon and sodium is indirectly related to the high instability at the bath-metal interface that would cause low current efficiency in the first place.

Cell to Cell Performances - Multivariate Analysis Data Collection and Analysis Procedure Most algorithms used in multivariate analysis require an important number of values to correctly represent the system. In order to achieve a high number of values for the analysis, every electrolysis cell was considered, from January 2007 to November 2011. The data considered for the analysis were divided into periods of four months. This period length was chosen because it almost corresponds to 100 tap cycles. For this time step, and according to the results from Fredrickson [3], the accuracy on the calculation of the current efficiency is known to be a little higher than ±0.66 %, for the calculated value. The remaining variables included results from the average value of the measurements attributed to this time period.

Other Indicators The pot noise is an indicator of the variations in the cell resistance. An increase in the noise is generally due to strong fluid movement in the cells [29] or caused by an incorrect anode. Both of these phenomena should lower the CE% according to theory. However, our results did show a negative correlation but it is not as strong as one would expect. Dissolved sodium is also related to noise with a similar coefficient as the CE%. The increase in movements at the bath-metal interface and the impacts observed correlate with the hypothesis of this paper less than expected. To add precision, it would be necessary to identify the origin of the instabilities (metal pad or anode incidents) and analyze the results in two groups separately.

Calculations were effectuated using the "STATlSTICA" software. A total of 40 predictors and over 8000 values were considered. The data used were filtered using Henry's chart to eliminate extremely out-of-range values and validate the normal distribution of the values. Tn some cases, data were transformed using a logarithmic function to approach a normal distribution. The analysis was divided in three studies: I. Defining the most reliable predictors for CE% (excluding sodium dissolved in aluminum). 2. Defining the most reliable predictors for the sodium dissolved in aluminum 3. Developing a model based on the predictors that correlates with the observed value of current efficiency (including sodium dissolved in aluminum) The values were computed using a boosting trees algorithm. This algorithm generates an important number of simple and poorly accurate decision trees. Then it uses the results from one tree as input for another. Different weight is assigned to every tree during iterations to increase the accuracy of the model. Elith and al. [33] described the importance of setting the input parameters and their

The total number of anodic incidents shows a stronger influence than noise on the current efficiency. Data collected at the V oerde smelter [20] indicate that spikes can contribute to lower CE% for as much as 1.6%. The relationship according to anodic incidents was present but unclear with the present analysis. Data from the analysis also included air burned anodes. These cases do not lower the current efficiency, therefore weakening the correlation. The correlation with the sodium is weak and negative. It is highly plausible that spikes absorb sodium from the aluminum when the carbon is dipped into the metal pad. Chemical analysis[30] from anode incidents taken at AAI showed an important increase in the sodium content of the anode. This high sodium concentration is specific to the first centimeters of carbon.

744

impact on the obtained using Study #3 will model and the production.

resulting model. Results from analysis 1 and 2 this algorithm, will be discussed simultaneously. be discussed according to the efficiency of the possible implementation as a tool for aluminum

agreement with the observed results. It could be the case of the sodium oxide concentration in primary alumina and could be the reason why the alumina dosage speed (time in overfeed vs underfeed) have this much of an importance on the Na(Al) predictions.

Reliable Predictors for CE% and Na(A!l From the computed results, it was possible to determine which parameters have the most of influence on the dependent variables considered in this article. The 10 most important predictors are listed in Table III in decreasing order of importance.

The developed models that are based on the previous indicators had a correlation coefficient of 0.53 and 0.9 respectively to the CE% and the Na(AI) between the predicted and observed values. Developing a Model that Correlates the Calculated Current Efficiency of a Cell to its Design and Performance Parameters.

Table TIT: Most important predictors for current efficiency and sodium content of the aluminum resulting from multivariate analysis Current efficiency Silicon in Al Line current Total bath transfer Daily dose of A1F3 Target for cell resistivity Metal Height Phosphorus in Al Cell's days in operation Cell's resistivity Vanadium in Al

Using a boosting tree analysis, the most efficient predictors provided by study # 1 were considered. Moreover, the sodium dissolved in molten aluminum was included in the model predictors to identify its impact on the calculated CE% value. In this case, sodium is the second strongest predictor, closely following the cell's intensity. The obtained correlation between the predicted and observed values is illustrated in Figure 3. The correlation coefficient for this model is the same as study # I : 0.53. The results indicate that Na(AI) and CE% are evolving similarly with a change of parameters. However, no increase of the correlation has been observed when including Na(AI). This indicates that the sodium content variation is strongly related to the same parameters that were used in the study # 1. These results are in agreement with the premise of this paper.

Sodium Al Year of measurements Line current Alumina dosage speed Silicon in Al Target for cell resistivity CaF z content in bath Metal height Number of point feeder action Iron in aluminum Cell's voltage

Four of the ten strongest indicators were found to have an important impact on both the CE% and the line current, silicon in molten aluminum, metal height and the target for cell resistivity. Two of these have been discussed in the previous section of this paper. According to a model from Biedler and Banta [34], an increase in the metal depth tends to lower the bath ratio. This is because the surface area adjacent to the metal pad region increases, while the amount of heat available for dissipation remains the same. Therefore, a significant amount of bath is consumed to thicken the ledge in the metal pad region. The change in cryolitic ratio can explain the change observed towards the CE% and the Na(AI).

Predicted Vii. Observed values fOl' current efficiency using multivadate amd,-sls

The target for cell resistivity was not considered in the first section because of the strong correlation it had with the resistivity itself. Out of the remaining indicators, many were already discussed in the first section. However, it is interesting to notice that in some case, (e.g. iron in aluminum, total bath transfer) the correlations were not observed in univariate analysis. This can be attributed to non-univariate relations that were not considered in the first part of this paper. On the other hand, some relations observed previously are confirmed by the multivariate analysis, i.e. CaF z relation with sodium, number of point feeder's action.

0.90 0-9$ Ol"",y.d vol....

1.00

Fig. 3: Predicted vs. observed values of the current efficiency, obtained with a multivariate model. (Full line represents the ideal model; the area between dotted lines represents the error attributed to the calculation ofCE%) The model developed is a very good start and indicates that it is possible to use multivariate analysis to approximate the current efficiency of a particular cell. The present results need to be optimized for the model to be used efficiently. To be optimal, the model should be able to adequately represent a time step of one month operation at the most. Moreover, the correlation coefficient needs to be increased so that the deviation is lower than the potential difference between two cells. The following adjustments of the model will be examined in a future analysis to possibly improve its efficiency:

Two of the remaining indicators (P & V dissolved in AI) have been previously studied by Sterten et al. [23] and many others [35-37]. The observation from this paper correlates with the literature. The year of the measurements had a major impact on the sodium concentration in aluminum. An important part of this correlation is attributed to the increase in the potline current with the years.. However, it is supposed that some parameters, that were not included in the analysis, were more important than expected. Therefore, if important variations of these hidden parameters occurred during the different years, it would be in

745

•

•

•

•

References

To increase the impact of the sodium content of the aluminum, indicators from the study #2 need to be considered in the analysis. The sodium level could also be classitied according to the level (low, medium, high). The use of categorical values instead of continuous ones could create an important difference in the results. The daily tluctuations in the sodium could be a predictor much more accurate about the day that passed than the current level. By grouping the electrolysis cell in small groups with similar current efficiency, it would be possible to decrease the error in the calculation of the current efficiency.

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

Conclusion This paper describes the relation between the current efficiency and several performance parameters of an electrolysis cell. The correlation between these indicators and the sodium dissolved in aluminum is also investigated trough the performance of pot lines for five years of operation.

12. 13. 14.

Results trom the literature were investigated and many indicators related with the current efficiency were in strong agreement with the published results. However, bath temperature and AIF3 content did not correlate as expected. Most results according to sodium were not in close agreement with the literature, e.g. temperature and noise. Strong correlations were found with the dosage rate of alumina and with the number of point feeder's action. Both the sodium and the current efficiency were strongly correlated with the silicon level in metal, the current intensity and cell's resistivity. These last two indicates that the anode-cathode distance probably has a strong intluence on the results.

15. 16. 17. 18. 19. 20. 21.

22.

Finally, it was possible to develop a model with multivariate analysis that can approximately calculate the current efficiency of an electrolysis cell based on the performance results of four operating months. The use of the sodium indicates that this predictor is strongly correlated with the other variables used in the model, illustrating the premise of this paper. This model is still in development and needs to be optimized to be efficiently used as a production tool for follow-up of the cells.

23. 24. 25. 26. 27. 28.

Acknowledgments

29.

This work is supported by the courtesy of Aluminerie Alouette Inc. Permission to publish the results is gratefully acknowledged. Financial support for this study also came from the "Fond de Recherche du Quebec - Nature et Technologies" through a BMP grant. The authors wish to thank Thomas Bastien and Arnaud Destaville from StatSoft-France for their help and support on the work related to multivariate analysis.

30. 31. 32. 33. 34. 35. 36. 37.

746

Kvande, H.. Chapter 0 : Course on industrial aluminium electrolysis, 2012: Chicoutimi, Qc, Canada. Fredrickson, G.L., Light Metals, 2003: p. 299-306. Fredrickson, G.L., Light Metals, 2003: p. 307-314. Tabereaux, A., in The international Harald A. Oye Symposium, 1995: Trondheim, Norway. p. 115-127. Tabereaux, A.T., Light Metals, W.R. Hale, Editor 1996. p. 319-326. Polyakov, P.V., et aI., Tsvetnye metally, 1993. 34(3): p. 29-31. Thonstad, l, et aI., Light Metals, 2001. Solheim, A., Light Metals, 2002. p. 225-230. Haupin, W.E., Light metals, 1997. p. 319-323. Othman,1. and M. Ali, Light Metals, 1997. p. 411-415. Danielik, V., P. Fellner, and J. Thonstad, Journal of Applied Electrochemistry, 1998.28: p. 1265-1268. Fellner, P., et aI., Electrochimica Acta, 2004. 49(9-10): p. 1505-1511. Keller, R., lW. Burgman, and P.j. Sides, in Light Metals 1988. p. 629-631. Sterten, A., P.A. Solli, and A. Solheim, in AlSymposium 1995: Donovaly, Slovakia. p. 209-219. Kent, J .H., Journal of Metals, 1970. 22( II): p. 30-36. Tingle, W.H., J. Petit, and W.B. Frank, Aluminium, 1981. 57: p. 286-288. Tarcy, G.P. and l Sorensen, Light metals, 1991. p. 453459. Simoes, T., et aI., Light metals, 2008. p. 361-368. Liu, Z., et aI., Light metals, 2012. p. 935-938. Rolofs, B. and N. Wai-Poi, Light metals, 2000, p. 189193. Kurenkov, A., et aI., Magnetohydrodynamics, 2004. 40(2): p. 203-212. Saevarsdottir, G., et aI., 10th australasian aluminium smelting technology conference, 2011: Launceston. Sterten, A., P.A. Solli, and E. Skybakmoen, Journal of Applied Electrochemistry, 1998. 28: p. 781-789. STATSOFT, Statistica 9.0,2009. Dion, L., F. Laflamme, and D. Dube, 2012, Aluminerie Alouette inc. 17 pages. Rapport interne Coursol, P., et aI., Light metals, 2012. p. 591-595. Achelis, S., L'analyse technique de A az. ed. Valor. Welch, B. and A. Tabereaux, Fourth australasian aluminium smelter technology workshop, 1992. Shin, D. and A.D. Sneyd, Light metals, 2000. p. 279283. Chabot, J. and B. Beaulieu, Profil de megots, 2011, Aluminerie Alouette Inc. Rapport interne Lindsay, S.. Chapter 16: Course on industrial aluminium electrolysis 2012: Chicoutimi, Qc, Canada. Dion, L., L. Kiss, and P. Coursol, 8th international conference on mechanical engineering, 2012, p. 72-82. Elith, l, lR. Leathwick, and T. Hastie, Journal on Animal Ecology, 2008. 77(4): p. 802-13. Biedler, P. and L. Banta, Light Metals 2003. p. 441-447. Thonstad, J., et aI., METSOC - Light metals and matrix composites, 2004: Hamilton. p. 595-602. Thisted, E.W., 2003, Institutt for materialteknologi. p. XVI, 248 s. ill. Schmidt-Hatting, W., R. Perruchaud, and lE. Durgnat, Light metals 1986: New orleans. p. 623-625.

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

GAS-SOLID FLOW APPLICATIONS FOR POWDER HANDLING IN ALUMINUM SMELTERS PROCESSES Paulo Douglas S. de Vasconcelos l , Andre L. Amarante Mesquita2 IAlbras - Aluminio Brasileiro S/A, Barcarena-PA, Brazil 2Pederal University of Para, Belem-PA, Brazil. Keywords: Gas-solid flow, Dense phase, Dilute phase, Minimum pressure drop, Minimum fluidization velocity

Abstract The old aluminum smelters feed their electrolytic cells with the overhead cranes as can be seen in figures 2. This task is very hard to the operators and causes spillage of alumina to the pot room workplace. This nuisance problem is being solved by the development of a special multi-outlets nonmetallic fluidized pipeline [2].

Gas-solid flow occurs in many industrial furnace processes. The majority of chemical engineering unit operations, such as drying, separation, adsorption, pneumatic conveying, fluidization and filtration involve gas-solid flow. Poor powder handling in an industrial furnace operation may result in a bad furnace performance, causing errors in the mass balance, erosion caused by particles impacts in the pipelines, attrition and elutriation of fines, overloading the bag houses. The lack of a good gas-solid flow rate measurement can cause economic and environmental problems due to airborne dust. The paper is focused on the applications of powder handling in relation with furnaces of the aluminum smelter processes such as anode baking furnace and electrolytic furnace (pot cell) to produce primary aluminum.

Introduction The anode baking furnace illustrated in figure 1 is composed by sections made up of six cells separated by partitions flue walls through which the furnace is fired to bake the anodes. The cell is about four meters deep and accommodates four layers of three anode blocks, around which petroleum coke is packed to avoid air oxidation and facilitate the heat transfer. During the baking process, the gases released are exhausted to the fume treatment center (PTC - [1]), where the gases are adsorbed in a dilute pneumatic conveyor and in an alumina fluidized bed. The handling of alumina is made via a dense phase conveyor.

Fig 2. Aluminum smelter pot room being fed of alumina by the multipurpose overhead crane. The fundamentals of powder pneumatic conveying and fluidization will be discussed in this paper, such as the definition of a pneumatic conveying in dilute and dense phase, the fluidized bed regime map as illustrated in figure 3 and finally the air fluidized conveyor.

- - - - - - - - - IMft!!ingGfts Vd«ily

)

Fig 3. Flow regime map for various powders. Pig 1. Anode baking furnace building overview.

Firstly, petro coke and alumina used as raw materials in the primary aluminum process is characterized using sieve analyses (granulometry size distribution). Then, bulk and real density are determined in the laboratory analyses; with these powder physical

The baked anode is the positive pole of the electrolytic furnace (cell) which uses 18 of them by cell. The pot room and the overhead multipurpose crane are illustrated in figure 2.

747

properties, they can be classified in four types using the Geldart's diagram as illustrated in figure 4 [3]. The majority of powders used in the aluminum smelters belong to groups A and B considering Geldart's criteria. This figure 3 summarized the fluidized bed hydrodynamics related with powders classified according to Geldart's criteria. Once the velocities associated with each mode of operation are determined, the pressure drop of the regime is calculated so that the gas-solid flow is predicted using the modeling and software adequate to optimize the industrial installation. Finally one case study applied in the baking furnace of pneumatic powder conveying in dilute phase are shown as a result of a master degree dissertation. Another case study is the development of an equation to predict the mass solid flow rate of the airfluidized conveyor as a result of a thesis of doctorate [2]. The equation has design proposal and it was used in the design of a fluidized bed to treat the gases from the bake furnace and for continuously alumina pot feeding the electrolyte furnaces to produce primary aluminum.

Mode oftransport of solids Dilute phase Dense phase

Solids - to - air ratio (JL) 0- 15 > 15

.. Table 1 - Systems' classIfIcatIOn concermng solJds-to-alr ratIO source: [4].

Fig. 5 - typical pneumatic conveyor layout - source: [4]. Figure 6 illustrates a variety of solids modes of transportation and the states diagrams showing the log of the pipeline pressure drop versus the log of the air velocity inside the pipeline. From figure 6 it is concluded that in dilute phase the pneumatic conveyor has high air velocity, low mass load ratio, and low pressure drop in the pipeline. Tn dense phase mode the conveyor operates with high mass load rate, low air velocity but high pressure drop in the pipeline. The engineer responsible for the project has to analyze which is the best solution for each case study.

aeratable

dl' Fig 4. Powder classification diagram for fluidization by airsource: [3]. Fundamentals of pneumatic conveying of solids Pneumatic conveying of solids is an engineering unit operation that involves the movement of millions of particles suspend by draft in dilute phase or in a block of bulk solids in dense phase inside a pipeline. Figure 5 illustrates a pneumatic conveying of solids with the essential components, like the air mover, feeding device, pipeline and bag house. A good criterion showed in table I to decide if the transport of solids in air will be in dilute phase or in dense phase is the mass load ratio ( JL ) calculated by the equation 1.

"",

--.:....:;,.;----------

(1) G

Solid mass flow rate (kg/s)

v

Gas (air) volume flow rate (m 3 / S) Fig. 6 - Conveying conditions showing the changes in solids loading; Vp: State diagram horizontal flow, down: State diagram vertical flow - source: [4].

Gas (air) density (kg / m 3 )

748

Pressure drop calculation in the pipeline

43.6 < K < 2360

The equations given here are based on the hypothesis that the gassolid flow is in dilute phase. Some assumptions such as: transients in the flow (Basset forces) are not considered nor the pressure gradient around the particles (this is considered negligible in relation to the drag, gravitational and friction forces). The pressure drop due to particle acceleration is not considered. The flow is considering incompressible, Omni dimensional and the concentration of solids particles is uniform. The physical properties of the two phases are temperature dependent. The mass flow rate for each phase can be expressed by the following equations:

(9)

K is a factor that determines the range of validation for the drag coefficient expressions, when the particle Reynolds number is unknown, and given by:

(10)

Acceleration due to gravity (m I s2) Particle diameter (m )

(2) mg

Gas mass flow rate (kg/s)

Vmc

Minimum air velocity in dilute phase (m/s)

cmc

Volume occupied by the gas inside the pipeline (-)

Ap

Pipe cross section (m 2 )

Jig

Gas dynamic viscosity (Pa.s)

II

Solid friction factor (-)

Ar

Archimedes' number (-)

The total pressure drop, ;:"'Pj' for gas-solid flow is calculated with the contribution of the static pressure ;:"'PE

(3)

'

and friction loss ;:"'PF

for both phases:

(II)

(4)

(12) AI' cn}c' = -A = I -

C.

.\

P

Ps

(5) (13)

Solid density (kg I

m3 ) The contribution due to the friction factor given by the Darcy equation:

Solid velocity (m/s) Porosity or volume occupied by the solid inside the pipeline

8,

(-)

(14)

Area occupied by the solid inside the pipeline (m 2 ) Area occupied by the gas inside the pipeline (m 2 )

(15)

Pipe diameter (m ) Velocity of the particle VI and the particle terminal velocity

The friction factor for the gas is calculated by the Colebrook equation.

calculation: In this paper it is considered the models of Yang [5] for the pressure drop calculation.

(16)

+ _1_;;=.7=]

-["";1"F-I- = 1.74 -V 4.fg

Re- 4.fg

(6)

Re - PI') VI = ---'-----'-'-,

K
N CIT

11

'

Il

(m) 5.46E-04 3.38E-04 1.66E-04 1.15E-04

Electromagnetic priming has been demonstrated to prime CFFs with less than the industry standard metal heads. The required metal head for electromagnetic priming of high PPI CFFs (50 and 80) was 1I3 rd of the standard height as published previously for industry. Furthermore, using electromagnetic priming, stacks of up to 3 filters have been primed using the same low metal heads.

tTf

,(1: mm:!in Pf'l C'rr

5.08E-08 3.lOE-08 1.57E-08 6.52E-09

Eq.5 Forchheimer k2

Conclusions

- 1 t»1} mn!::.m P?!. err "" ,.

Eq.5 Forchheimer k1 (m 2)

The maximum allowable metal height in filter boxes currently dictates the type of filters and filtration rates that can be applied in industry. Based on the analysis of the filter permeability, it is the priming height and not the pressure drop during casting, which should typically limit the type of tilter applied. Electromagnetic priming should allow higher PPI filters to be applied in existing filter bowls, than could otherwise be applied.

17

Figure II. Metal head required to sustain flow as a function of superficial velocity, u., shown in [mm/s] and filter thickness, L shown in [mm] from Equation [5], for 30 and 50 PPI filters.

Furthermore the more efficient removal of gas from the CFFs should allow for lower metal height during filtration for a given throughput or a higher throughput for a given metal height. This may also make it practical to use for example 50 PPI filters with more consistent tiltration performance to replace 30 PPI filters.

A comparison of Figure 3 and Figure 11, indicates that much higher metal heads are required to prime than to sustain tlow after priming, even assuming much thicker, e.g. 150 mm tilters, than applied in industry today. An opportunity exists to apply electromagnetic priming to stacks of filters in existing filter box installations to prime 2 or 3 standard thicknesses of 50 mm filters and achieve higher filtration efficiencies, as indicated by the estimates shown previously in Figure 10.

Improved melt quality can be achieved either by the use of higher PPI filters, lower velocities (and hence larger filtration areas) or thicker filters. With the use of electromagnetic priming both thicker filters and higher PPI filters become practicable for existing tilter boxes.

Alternatively electromagnetic priming can be used to prime and enhance the productivity of higher grade, e.g. 50 PPI filters, which achieve improved filtration performances over 30 PPI as shown previously in Figure 9. Figure 11, indicates the required pressure drop to maintain flow for such a filter, verifying that priming head and not pressure drop during filtration is likely the factor determining usage.

Future Work Experimental trials will be conducted to electromagnetically prime a stack of 3, 30 PPI grade CFFs. Lower grade CFFs may also be tested, e.g. 10 or 20 PPJ. Gravity tiltration experiments will be conducted to determine the filtration efficiency and verify the estimates given in Figures 9-10. Furthermore the application of an electromagnetic field to prime standard sized tilters will be demonstrated in slightly modified industry standard filter boxes.

978

Acknowledgements

10. M. W. Kennedy, S. Akhtar, J. A. Bakken and R. E. Aune, "Improved Short Coil Correction Factor for Induction Heating of Billets," 3rd International Symposium on High-Temperature Metallurgical Processing, (2012), 373-382. 11. S. Ray, B. Milligan and N. Keegan, "Measurement of Filtration Performance, Filtration Theory and Practical Applications of Ceramic Foam Filters," Aluminium Cast House Technology, (2005), 1-12. 12. J. E. Dore and C. Bickert, "A Practical Guide on How to Optimize Ceramic Foam Filter Performance," Light Metals, (1990), 791-796. 13. N. Keegan, W. Schneider and H. P. Krug, "Evaluation of the Efficiency of Fine Pore Ceramic Foam Filters," Light Metals-Warrendale, (1999),1-10. 14. R. Fritzsch, "Filtration of Aluminium Melts using Ceramic Foam Filters (CCF) and Electromagnetic Field," Trondheim: NTNU, Norway, (2011), 1-86. 15. T. Iwasaki, J. Slade and W. E. Stanley, "Some Notes on Sand Filtration [with Discussion]," Journal of American Water Works Association, vol. 29, (1937), 1591-1602. 16. D. Apelian and R. Mutharasan, "Filtration: A Melt Refining Method," Journal of Metals, vol. 9, (1980), 14-19. 17. C. Conti and P. Netter, "Deep Filtration of Liquid Metals: Application of a Simplified Model Based on the Limiting Trajectory Method," Separations Technology, vol. 2, (1992), 46-56. 18. H. Duval, C. Riviere, E. Lae, P. Le Brun and J. B. Guillot, "Pilot-Scale Investigation of Liquid Aluminum Filtration through Ceramic Foam Filters: Comparison between Coulter Counter Measurements and Metallographic Analysis of Spent Filters," Metallurgical and Materials Transactions B, vol. 40, (2009), 233-246.

The present study was carried out as part of the RIRA (Remelting and Inclusion Retining of Aluminium) project funded by the Norwegian Research Council (NRC) - BIP Project No. 179947/140. The industrial partners involved in the project are: Hydro Aluminium AS, SAPA Heat Transfer AB, Alcoa Norway ANS, Norwegian University of Science and Technology (NTNU) and SINTEF Materials and Chemistry. The funding granted by the industrial partners and the NRC is gratefully acknowledged. The authors also wish to express their gratitude to Egil Torsetnes at NTNU for helping with the design and construction of the experimental apparatus. Sincere gratitude is also due to Kurt Sandaunet and Arne Nordmark of SINTEF for their support and help, as well as for the use of the SINTEF casting laboratory.

References 1. D. E. Groteke. "The Reduction of Inclusions in Aluminum by Filtration," Modern Casting, vol. 73, (1983),25-27. 2. K. Butcher and D. Rogers, "Update on the Filtration of Aluminum Alloys with Fine Pore Ceramic Foam," Light Metals, (1990),797-803. 3. M. W. Kennedy, S. Akhtar, J. A. Bakken and R. E. Aune, "Electromagnetically Enhanced Filtration of Aluminum Melts," Light Metals, (2011), 763-768. 4. R. Fritzsch, M. W. Kennedy, S. Akhtar, 1. A. Bakken and R. E. Aune, "Electromagnetically Modified Filtration of Liquid Aluminium with a Ceramic Foam Filter," Accepted for Journal of Iron and Steel Research International, (2012), 1-4. 5. M. W. Kennedy, R. Fritzsch, S. Akhtar, J. A. Bakken and R. E. Aune, "Electromagnetically Modified Filtration of Aluminum Melts Part II: Filtration Theory and Experimental Filtration Efficiency with and without Electromagnetic Priming for 30, 50 and 80 PPI Ceramic Foam Filters," To be submitted to Metallurgical Transactions B, (2012), 1-69. 6. M. W. Kennedy, R. Fritzsch, S. Akhtar, J. A. Bakken and R. E. Aune, "Apparatus and Method for Priming a Molten Metal Filter," U.S. Patent Application, (2012), 1-26. 7. M. W. Kennedy, K. Zhang, R. Fritzsch, S. Akhtar, J. A. Bakken and R. E. Aune, "Characterization of Ceramic Foam Filters used for Liquid Metal Filtration," To be submitted to Metallurgical Transactions B, (2012), 1-46. 8. M. W. Kennedy, S. Akhtar, J. A. Bakken and R. E. Aune, "Analytical and Experimental Validation of Electromagnetic Simulations Using COMSOL®, re Inductance, Induction Heating and Magnetic Fields," COMSOL Conference 2011, Stuttgart, Germany, (2011), 1-9. 9. M. W. Kennedy, S. Akhtar, J. A. Bakken and R. E. Aune, "Analytical and FEM Modeling of Aluminum Billet Induction Heating with Experimental Verification," Light Metals, (2012), 269-275.

979

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

PLANT SCALE INVESTIGATION OF LIQUID ALUMINIUM FILTRATION BY Ah03 AND SIC CERAMIC FOAM FILTERS Sarina Bao l , Martin Syvertsen l , Arne Nordmark l , Anne Kvithyld l , Thorvald Engh2, Merete Tangstad2 lSINTEF Materials and Chemistry, Trondheim, N-7465, Norway 2Norwegian University of Science and Technology, Trondheim, NO-7491, Norway Keywords: Aluminium, Filtration, Al 20 3 filter, SiC filter, Ceramic foam Sunndals0fa, Norway: one type high in AI 20 3 and one high in SiC. These two types of filters were produced in the same line by the same supplier, giving similar porosity and wall thickness.

Abstract

Plant scale filtration experiments of 10" x 10" x2", 30PPi Al 20 3 and SiC industrial tilters were carried out. Wetting experiments show that the SiC tilter wets better with molten aluminium than A1 2 0 3 . The assessments by LiMCA IT and laser were employed to study the behaviour of the two tilters. The Al 20 3 filter shows improved time dependent behaviour, increasing filtration efficiency, during one hour filtration. This is not the case for the SiC filter. It decays faster than the Al 20 3 one. The SiC filter requires less pressure drop to infiltrate the metal. The result suggests that the SiC can be a new filter choice in the aluminium industry.

A top view of the tiltration loop is shown in Figure I. The melting furnace contains 15 tons of aluminium alloy, melted by a burner inside. A porous plug in the furnace bottom injecting argon gas is used to stir the metal. Mechanical stirring from gate I is employed, especially after standing a long night. The stirring is to increase the inclusion number in the metal from the melting furnace. The pump is used to control the mass flow of the metal.

Gate 2

Melting Furllace

Gate 1

Introduction

Pump

The presence of non-metallic inclusions is considered to be one of the most widespread causes of defects encountered in aluminium product: poor machinability, poor surface finish, reduced extrusion die life, cracks, reduced strength, ductility, and fatigue resistance, and pinholes [I]. Liquid metal filtration during the casting process is now a common technology to remove inclusions [2]. Ceramic foam tilters (CFF) have an open pore reticulated structure with very high porosity and very high surface area to trap inclusions. The open foam structures are composed of ceramic material, such as alumina, mullite or silica. Alumina is the most common filter material. Ceramic foam filters operate in a deep bed filtration mode where inclusions smaller than the pore openings are retained throughout the cross-section of the filter [3]. Ceramic foam filters are produced by impregnating reticulated polyurethane foam with a ceramic slip, removing the excess slip by squeezing the foam, and then drying and tiring the body. The result is a ceramic replica of the original foam [4].

Figure I The schematic top view of the filtration loop Initially metal runs in the launder from the outlet of the melting furnace bypassing the filter bowl, and goes back to the melting furnace. Dam 2 is opened when the LiMCA IT reading before the filter shows a relatively stable inclusion level, and metal is led through the preheated tilter, tills the lower space of the filter bowl, and goes out. Dam 3 is opened when the right side groove of the filter bowl is full of metal. Then dam 1 is closed to ensure a constant mass flow through the filter. After each experiment, dam 1 is opened and dams 2 and 3 are closed. Dam 4 is opened to drain the metal in the tilter bowl after each experiment. Dam 5 is opened to drain the metal in launder at the end of the day. The tllter in a filter bowl is preheated by a gas burner in the lid to avoid thermal shock and freezing of the metal when filtration starts.

Earlier tests have shown that pure SiC is significantly better wetted by molten aluminium than Al 20 3 [5]. Improved wetting of aluminium on ceramics probably is an advantage in getting molten metal to infiltrate the filter during priming. Also better wetting should increase the filtration efficiency of inclusions during tiltration due to better contact between filter and metal. Thus, SiC can be a filter choice and the use of SiC as an alternative tilter material is investigated. Therefore, plant experiments comparing SiC industrial filters and Al 2 0 3 industrial filters have been carried out.

Analysers (LiMCA) II [6] which Two J)quid Metal give on-line information for inclusion level in k/kg were positioned before and after the filter.

Experimental Procedure

Two lasers positioned before and after the tilter bowl give the metal height in the launder in mm. Finally a thermocouple

Four tiltration experiments were performed with two types of 10"xI0"x2", 30ppi tilters in the reference center of Hydro

981

positioned in the launder measures the temperature after the filter in °C.

1200°C for the AI 20 3 industrial filter, and 39° and 28° at 1100°C and 1200°C for the SiC industrial filter.

The AI 2 0 3 industrial filter (for aluminium filtration) contains A1 20 3 , P20S, Si0 2, and 1% K20+Na20, while the SiC industrial filter contains 5-9% A1 20 3 , 58-64% SiC, and 29-33% Si0 2. This SiC filter is normally used for DC casting and continuous casting of copper and copper alloy. The data are given by the supplier. The densities of filter materials were tested by AccuPyc 1330 I at SINTEF. AI 20 3 and SiC industrial filters have average porosity of 88.2% and 85.0%, respectively. The porosity is calculated on the basis of mass and volume measurements using the relation:

Due to settling, the inclusion concentration declines with time. Since the rate of reduction in concentration is proportional to the concentration, the concentration will follow an exponential function. It is necessary to take the time dependency of a narrow size range into account when the filtration efficiency is calculated.

porosity = 1 _

the bulk density the material density

Using a narrow size range is more accurate when taking settling into account. Figure 7 to Figure 10 show the number density of inclusions in the size range of 25-30 /lm before and after the filter, as well as filtration efficiencies in the 4 experiments.

(1)

We can calculate the 68% confidence intervals (dashed lines) for the fitted curves and use those curves to give errors for propagation. The confidence lines give the range where the true value for a given measurement is likely to be given that the fitted curve is of correct form. Approx. 68% of values drawn from a normal (or Gaussian) distribution are within I standard deviation away from the average. In this way it is possible to calculate both the filtration efficiency as a function of time for the various size ranges, and also the uncertainty for the filtration efficiency. For more details of this statistical treatment, please refer to [8]. Note that the filtration efficiency is defined as the inclusion number left in the filter divided by the initial inclusion number.

Using the same procedure, each filtration experiments lasted for I hour. The filters were AI 20 3 in Exps.l and 3 and SiC in Exps.2 and 4. 3 groups of disk samples for spectrographic analysis from both before and after the filter were taken in each experiment at approximately 0 min, 30 min, and 60 min. The wettability of the filter materials had been tested in a furnace as described in [7] in a high vacuum of 10.8 bar using the same procedure. The received flat filter material was cut into a small tablet and used as the substrate. Results

The filtration efficiency for N25-30 in Exp.1 are given from 15% to 32% according to the confidence lines, and from 33% to 11% (decreasing), from 10% to 38%, and from 35% to 51 % for Exp.2, Exp.3, and Exp.4, respectively. The s_e increasing trend for the AI 20 3 filter and decreasing trend for the SiC filter are observed for inclusions until 40 /lm, except 25-30 /lm inclusions in Exp.4 (Figure 10). See Table 1. Only less than 13.2% of the inclusions are larger than 40 /lm in all four experiments, as shown in the last row of Table I, which results in a huge uncertainty (more than 100%) for filtration efficiencies in that range. Thus the results larger than 40 /lm are unreliable.

The alloy contained approximately 1.00 wt% of Mg, 0.14 wt% of Fe, 0.07 wt% ofSi. Other elements are all less than 0.05 wt%. All four experiments have relatively stable inlet and outlet metal compositions, except for a sudden increase at inlet level at 60min in Exp.1 and outlet level at 0 min in Exp.3 for Ca. As an example, Figure 2 shows the composition of the alloying elements in Exp.2 from spectrographic analysis.

Exp.2

_ _

I11III _

Before the filter,Omin Before the filter,30min Before the filter,60min After the filter,Omin After the filter,30min After the filter 60min

Table I Filtration efficiencies (%) with inclusions until 40/lm and % of inclusions larger than 40 /lm Exp. N20-25 N25-30 N30-35 N35-40

Si FeCuMnMgZn Ti Cr Ni PbSnNaCa B Bi Zr V BeCoSbGa P Ag Sr

N40-100

Chemical Element

2 3 21-28 33-9 20-32 15-32 33-11 10-38 16-51 4-47 40-11 1-76 76-0 0-66 % of inclusions> 40/lm 8.2 11.1 5.3

4 46-33 35-51 52-21 70-46 13.2

Note data such as 2/-28 means it increases from /2% to 28%; and 33-9 means it decreasesfioom 33% to 9%.

Figure 2 The content of alloying elements before and after the filter in Exp.2 using a SiC filter

Figure I I shows the pressure drop with the metal temperature. At the same temperature around nO-730°C, Exp.4 (SiC) has a lower pressure drop than Exp.2 (SiC) due to the warmer bowl. Exp.3 (AI20 3 ) with a higher metal temperature experiences a lower pressure drop than Exp.l CAI20 3 ). The warmer the metal or the bowl, the less pressure drop is required.

Figure 3 to Figure 6 give the time dependent behaviour of the contact angle of pure aluminium on filter materials at higher temperatures. The first degree exponential decay fittings show the contact angles approaching 84° and 44° at lIOO°C and

AccuPyc 1330 is a density analyser from Micromeritics, 4356 Communications Dr. Norcross, GA 30093-2901, U.S.A.

982

120 160 100

Equation

140

o OJ

Reduced Chi Adj. R-Squar

•

§ 120 jg

0

OJ

Value

?$OP'F=l ?$OP'F=l

8357039630:

yO

Al

Standard E

§ 80

771588 613034

jg

§

0

v 60

V 100

•

Experimental data

-Fitted line

o

10

20

40

30

50

60

40

70

0

10

5

Time/min

Figure 3 Contact angle at II OO°C for the AI 2 0 3 industrial filter 160

SiC, 1100

120

°c

SiC,1200

100

120

80

••

§

jg 80

jg

0

0

u 60

U

40 20

• Experimental data - - Fitted line

0

5

10

20

25

1.80386

0.99251

0.3103

40 • Experimental data - - Fitted line

2

4

Time/min

.

Exp.1 N25-30

Figure 6 Contact angle at I 200°C for the SiC industrial filter

50

•

Exp.2 N25-30

In

Out - - Exponential fit of in - - Exponential fit of out - - Filtration efficiency

bJl

40

2

In

Out - - Exponential fit of in ............ Exponential fit of out - - Filtration efficiency

@

"'" ""'""

10

8

6

Time/min

Figure 5 Contact angle at 1200°C the SiC industrial filter 3

l)+yO

Adj. R-Square

175933

0

30

y=Al'exp(-xfj

Reduced Chl-S

60

20

15

-»u

20

u::v

"

.g

.s

r.c.

0

160

170

180

190

30 u '"

"Gv

..D

150

°c

ExpOecl Equation

ia

25

Figure 4 Contact angle at I 200°C for the AI 2 0 3 industrial filter

140

""§l100

20

15

Time/min

10

10

30

40

50

60

70

80

90

""G

Eu

"c

100

Time/min

Time/min

Figure 7 The inclusion number (density) and filtration efficiency for inclusions 25-30 11m in Exp.1 using an AI 2 0 3 filter

Figure 8 The inclusion number (density) and filtration efficiency for inclusions 25-30 11m in Exp.2 using a SiC filter

983

Exp.3 N25-30

•

ExpA N25-30

In Out

$

• ®

- - Exponential fit of in - - Exponential fit of out - - Filtration efficiency

In Out

- - Exponential fit of in - - Exponential fit of out - - Filtration efficiency

60

"

u

..=

30

40

50

60

70

80

90

30

40

100

50

60

70

80

90

Time/min

Time/min

Figure 9 The inclusion number (density) and filtration efficiency for inclusions 25-30 11m in Exp.3 using an AI2 0 3 filter

Figure 10 The inclusion number (density) and filtration efficiency for inclusions 25-30 11m in Exp.4 using a SiC filter

34 32 30 S 28 26

g.

'"

24 22 OJ P:: 20 18 16 690

•

Exp.1, AI 2 0 3

o

Exp.2, SiC Exp.3, AI 2 0 3

ExpA, SiC

700

__

710

720

730

740

Figure 13 Cross sectional view of the SiC spent filter in Exp.4 Dark parts are filter materials

750

Temperaturei'C

Discussion

Figure 11 The pressure drop vs. metal temperature

The contact angle is believed to decrease exponentially with time during isothermal holding, approaching a stable angle at the end [5]. Thus, first degree exponential decay fittings in Figure 3 to Figure 6 should closely describe filter behaviour. The final contact angles indicate that the SiC industrial filter has better wetting with aluminium then the Al 2 0 3 industrial filter. The Ah03 industrial filter does not wet aluminium, the contact angle >90°, at the casting temperature of 700°C; while the SiC industrial filter might wet aluminum at the same temperature. The wetting becomes poor at lower temperatures. As indicated in Figure 14, we assume that the metal travels closer to the wall when the filter- Al wettability increases. The inclusions carried in the metal will get more chances to collide with the wall and be captured by it. The liquid prefers rough surface (it gives better wetting) then a flat one. From these points of view, the SiC industrial filter gives a better filtration efficiency.

Figure 12 Cross sectional view of the AI2 0 3 spent filter in Exp.3 White parts are filter materials As examples, the cross sectional views of the spent filters in Exp.3 and Exp.4 were shown in Figure 12 and Figure 13. Filters were partly infiltrated by the metal. Part of the metal may have drained away during solidification due to gravity and the cohesion work of the metal, especially for non-wetting filters. We also obverses that reddish materials cover the exposed Al 2 0 3 spent filter. No obvious foreign objects in the SiC filter were found.

According to the thermodynamics Mg will react with AI 2 0 3 in the filter to MgAI 2 0 4 spinel at the interface: 3Mg(l) 3Mg(l)

984

+ Al z0 3 = 3MgO(s) + 2Al(l) + 4Al z0 3 = 3MgAlz04(s) + 2Al(l)

(2) (3)

explanation is that SiC industrial tilters need less pressure drop than Al 20 3 industrial tilters to let the metal run through the filter.

The reddish materials in the spent Al 20 3 filter in Figure 12 may be the products from reactions (2) and (3). Chemical analysis is requested in future work.

Priming is an issue in the aluminium industry, especially when metal freezes with a too low filter or metal temperature. The SiC filter requires less pressure drop to infiltrate the filter.

The phase diagram of the Al-Si-C system [9] shows that the following ternary quasi-peritectic reaction [10] occurs isothermally at 650°C (923K):

3SiC

Conclusions

+ 4Al(l) = Al4 C3 + 3[Si]

(4) I.

However, no apparent change of Si and Mg content was detected. The total amount of silicon available in the tilter is probably too little to give a significant change in the alloy composition.

· ·· ·

2. 3.

1 4.

>}

The SiC tilter has better wetting with aluminium then the Al 2 0 3 filter. The latter one does not wet aluminium at the casting temperature of 700 0 • Both filters did not influence the metal composition. The Al 20 3 filter filter shows better time dependent behaviour, with increasing filtration efficiency, during one hour tiltration. But the SiC filter does not. It decays instead. This SiC tilter is now used in the copper industry. Changes of composition and properties are required for use in aluminium refining. The SiC filter requires less pressure drop to infiltrate the metal. This gives improved priming. Acknowledgment

This research was carried out as part of the Norwegian Research Council (NRC)-funded BTP Project (No. 179947/140) Remelting and Inclusion Refining of Aluminium (RIRA). It includes the following partners: Hydro Aluminium AS, SAPA Heat Transfer AB, NTNU and SINTEF. Fundings by the industrial partners and NRC are acknowledged gratefully. Thanks are also owed to Bj0fn Rasch at hydro, for support in the industrial experiments.

Figure 14 The schematics of AI-filter wettability in a filter cell The SiC and Ah03 industrial filters did not change the metal composition. Moreover, there is no indication that carbide was formed or entered the metal from the filters. An aluminium alloy with less than 10 at% Si [II] at 700°C could allow Al 4 C3 to be produced according to reaction (4).

References 1.

However, reaction (4) is probably slow. In the current one hour filtration with 0.10 wt% of Si alloy, no significant increase of Al4 C3 was measured2 in Exps.2 and 4 (SiC filters). The Si0 2 and Al 2 0 3 components in the SiC industrial filters may slow down the kinetics of reaction (4).

2. 3.

LiMCA results are intluenced signiticantly by micro bubbles. However, no gas bubbling retining unit was involved. As shown in Table 1, Exps.I-3 have similar filtration efficiencies. However, Exp.4 has a higher value. The reason may be the improved wetting for SiC at metal higher temperature [2]. The filtration efficiency tends to increase with time for Al 2 0 3 filters in Exps.1 and 3, and to decrease for SiC tilters in Exp.2 and 4. The same trends are found for larger inclusions (up to 40 flm). It is more obvious for larger inclusion load. For example, Figure 9 (started with 4k/kg inclusions in Exp.3) has a more obvious increasing trend than Figure 7 (started with 1.5k/kg inclusions in Exp.l) for Al 2 0 3 filters. Figure 11 indicates that Al2 0 3 requires higher temperatures than SiC to achieve the similar pressure drops in overall. For example, Exp.3 (AI 2 0 3 ) and Exp.4 (SiC) show similar pressure drops at around 735-745°C and nO-730°C, respectively. The

4. 5.

6.

7.

8.

It has been measured by PoDF A. PoDFA (Porous Q.isc Eiltration Apparatus) is an off-line method to detect inclusion types and concentration area in mm2 /kg on a polished surface.

985

S.K. As. Fatigue Life Prediction Of An Aluminium Alloy Automotive Component Using Finite Element Analysis Of Surface Topography. 2006, Norges teknisk-naturvitenskapelige universitet: Trondheim. S. Bao, Filtration of aluminium-experiments, wetting and modelling. PhD Thesis. 2011, Norges teknisknaturvitenskapelige universitet: Trondheim. p. 204 s. Z. Taslicukur, Production of ceramic foam filters for molten metal filtration using expanded polystyrene. Journal ofthe European Ceramic Society, 2007(27): p. 637. S.F. Ray, S. Pyrotek. Recent Improvements in The Measurement and Control of Ceramic Foam Filter Quality. 2001. S. Bao, A. Kvithyld, T.A. Engh, M. Tangstad, Wettability of Aluminium with SiC and Graphite in Aluminium Filtration. Light Metals, 20 11: p. 775-782. L. Liu, F.R. Samuel, Assessment of melt cleanliness in A356.2 aluminium casting alloy using the porous disc filtration apparatus technique: Part I Inclusion measurements Journal of Materials Science, 1997. 32: p. 5907-5925. S. Bao, K. Tang, A. Kvithyld, M. Tangstad, T.A. Engh, Wettability of Aluminium on Alumina. Metallurgical and Materials Transactions B, 2011. 42(6): p. 13581366. S. Bao, M. Syvertsen, B. Rasch, A. Kvithyld, T.A. Engh, M. Tangstad, Performance Evaluation of Al20 3

9.

and SiC Ceramic Foam with the Use of LiMCA 11 Data. Metallurgical and Materials Transactions B, 2012. Submitted. le. Viala, P. Fortier, 1 Bouix, Stable and Metastable Phase Equilibria in The Chemical Interaction between Aluminium and Silicon Carbide. Journal of Materials Science, 1990.25(3): p. 1842-1850.

10. II.

986

V. Laurent, D. Chatain, N. Eustathopoulos, Wettability of SiC by aluminium and Al-Si alloys Journal of Materials Science, 1987. 22( I): p. 244-250. D.l Lloyd, The solidification microstructure of particulate reinforced aluminium/SiC composites. Composites Science and Technology, 1989. 35: p. 159-179.

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

CASTING PRACTICES INFLUENCING INCLUSION DISTRIBUTIONS IN BILLETS Ghadir Razaz and Torbjiirn Carlberg Mid Sweden University, 851 70 Sundsvall, Sweden Keywords: Inclusion distribution, holding time, remaining melt, settling velocity. Abstract

Experimental procedures

A macro-etching method has been used to analyze the distribution and amount of inclusions along billets and on cross sections. Main parameters that have been varied are holding time before casting and amount of liquid remaining after casting. The result show that short holding times, in the order of 10 minutes, give increased amount of inclusions in the beginning of the billets, but holding times in the range from 30 to 60 minutes do not show any significant differences. If the melt remaining in the furnace after casting is less than about 3000 kg, the inclusion density increases towards the end of the ingots. The distribution of inclusions over the cross section of billets show that most inclusions are found in the centre of the billets, however, at increased total amount of inclusions, they tend to appear evenly over the whole cross sections. The results are discussed based on convection in furnace and settling rates and convection at solidification front.

In the present work samples were collected from two cast houses. At cast house 1, after producing primary aluminum in electrolytic cells, aluminum melt is transferred to holding furnaces, while at cast house 2, first a melting furnace is charged by scrap and then, aluminum liquid is transferred through a launder to a holding furnace before start of casting. At cast house 2, two castings are done from the same charge. Inclusion distributions were investigated in aluminum DC-cast billets. These billets were collected from castings with various holding times and different amounts of liquid remaining after casting. Table I shows the details of the samples. The billet slices were cut transversely in different position along the billets from the bottom to the top. To analyze the inclusion distribution, the cross-section of each slice was divided into three zones with equal areas; the center (C), mid-radius (M) and surface (S). Figure I shows schematically the details.

Introduction

Table I Details of collected samples Holding time

The presence of nonmetallic inclusions in Aluminum DC-cast billets is a common defect, which almost can be observed in every ingot production, and it has negative effects on mechanical properties of casting products and can cause problems in extrusion processes by damaging the dies [I, 2]. Moreover, presence of these inclusions in final products are not desired when the surface appearance of products are important. Therefore, different treatments such as sedimentation, degassing units and ceramic foam filtration are being used to reduce the amount of inclusions in molten aluminum [3]. Main nonmetallic inclusions can be classified in groups of oxides, carbides, borides and nitrides in shape of particles and films [4]. The most common inclusion in aluminum melt is oxides, which originate from the thin layer of oxide that always protects the liquid aluminum in casting processes. The oxide skin is fragile, can break and be introduced into the molten metal [5]. One way to improve the cleanliness of aluminum melt is by sedimentation. which means that the melt is held in a furnace for a period of time (holding time) to allow the inclusions to settle to the bottom of the melt [6]. This work aims to examine the effect of holding time on the distribution of inclusions, as well as the effects of different amount of liquid remaining after casting. The nonmetallic inclusions along the billets were analyzed by a macro-etching method, which recently has been developed and been described in [2,7]. The method gives the actual position of inclusions in a billet. and therefore more information than what can be obtained from, e.g. LlMCA measurements, which has earlier been used to study similar effects [1, 6].

[min]

30 60 25 10 85 80 30 40 35

Remaining melt in furnace after casti ng [kg)

AI alloy

5270 7550 9470 10000 6030 2670 1700 15000 3000

6063 6060 6060 6060 6063 6060 6060 6060 6060

Billet diameter

Cast house

[mm]

152 152 152 178 178 178 178 228 228

1 1 1 1 1 1 1 2 2

A deep etching method was used to assess the amount and distribution of non-metallic inclusions over cross-sections of the samples. First, a turning machine was used to remove scratches and to obtain flat surfaces. and then the slices were immersed in a solution of 15% sodium hydroxide and etched for about 15 minutes at 338 K (65 QC). After etching, samples were washed with water and cleaned with brush to make the billet surfaces clear. Subsequently, some holes, visible by naked eyes, appear after dissolving out the inclusions during deep-etching and these etch pits (holes) represent the places of the inclusions. These holes are about 10 times larger than original inclusions [7]. To evaluate the inclusion distribution on the cross-sections of the slices, pits on the etched surface were marked, and then copies of the surfaces were printed. The inclusions appearing, as black dots, could then easily be counted. According to the mean diameter of the etch pits, inclusions has been classified into two groups: large and small size, the large with diameters greater than 0.5 mm and small with diameters less than 0.5 mm.

987

cast after longer holding time and a high amount of inclusions is observed at start. The results also show that most of the inclusions are located at the center of the billet in each position. It can generally be concluded that the number of inclusions reduces towards the middle of the casting from a high amount at the start and then increases again towards the end. It can also be stated that it shows a high level of inclusions along the whole billet. Furthermore, it is noteworthy to mention that large size inclusions were found in all slices.

(a) (b) Figure 1. (a) Division of cross-sections [2] (b) Inclusion distribution in a deep etched sample. Results As was mentioned in the introduction the aim is to analyze and couple the amount, size and distribution of inclusions in the billets to different holding times and amount of melt remaining in the furnace after casting. From economical point of view and productivity, it is very important to evaluate the effect of different holding times on the settlement behavior of inclusions and find the optimum time [6]. Therefore, slices from certain various distances along the ingots from bottom (start of solidification) to top (end of casting) were cut and investigated. It should be noted that results about samples from the two companies have been presented and discussed separately due to differences in furnace design.

rn'ln tlu' bottolu (an}

(.o!litton

Figure 2. Inclusion distribution for 30 minutes holding time. 1.40

Samples from Cast House I Different Holding Times. It can be seen from figures 2 and 3. from 30 and 60 minutes holding time, that the highest number of inclusions was observed at the bottom of the billets (start of casting). The number of inclusions considerably decreases at 180 cm and remains in almost the same range at all other positions in the 30 minutes holding time casting, whereas in the 60 minutes holding time casting the number of inclusions declines towards the middle length of the billet (360 cm) and then increases again towards the end. Generally, there are very small differences between the two casting regarding level of inclusions overall in the billets. In addition, in casting with 30 minutes holding time, large inclusions (light blue bars) mainly were detected at the bottom of billet (at start of casting), while after 60 minutes holding time; they were found at both ends. Furthermore, the inclusions were concentrated to the center at all positions along the billets in both castings. In a billet from 25 minutes holding time, two slices, from the bottom and top, were investigated, and the results are given in table II. The number of inclusions is relatively high at both bottom and top, compared to the billet from 30 and 60 minutes holding time. The number of inclusions at the bottom is significantly higher than at the top. At the bottom, the inclusion population is more shifted to the periphery. At the top, it is more concentrated to the center zone. In addition, the number of large inclusions is relatively high in both slices and especially at the bottom, but it should be noted that the positions, at which the slices were taken, were not the same as for the billets analyzed in figures 2 and 3. Data from the casting with 85 minutes holding time is also given in table II. As it can be observed, the number of inclusions at the top of this billet is rather similar to the amount of inclusions at bottom, and also, there are noticeable numbers of large inclusions at both positions. Figure 4 displays the results of short holding time (10 minutes). Firstly, the total amount of inclusions is higher than in the billets

Figure 3. Inclusion distribution for 60 minutes holding time.

510, ...1"" IHHr,ittOll

tltt bW""t ft'(lIU titt b(lttOIU {nu }

Figure 4. Inclusion distribution for 10 minutes holding time. Table II Inclusion distribution for 25 and 85 minutes holding time Sample position (em) Bottom Top Bottom Top

# (e)

#(M)

91 76 63 79

89 50 47 56

#(S)

#Total

119

299 181 180 193

SS

70 58

Large size

33 7 11 10

Holding time (min)

25 25 85 85

Different amount of melt remammg after casting. In normal casting the amount of remaining liquid after casting is about 5000 kg, and it is assumed that the liquid is enriched with more inclusions at the end of a casting due to settling [1]. Samples from two castings with small amount of remaining melt were investigated and the results are shown in table Ill.

988

It can be seen that in the billet with 2670 kg of melt remaining after casting, the amount of inclusions at the end (top of billet) is substantially higher than at the bottom (start of solidification). Furthermore, relatively high numbers of large size inclusions were found in both slices. In this casting with high level of inclusions, it was found that the inclusions were distributed evenly over the whole cross sections of the billets and not in a certain zone. It is clear from table III that in the billet with 1700 kg remaining melt, the number of inclusions in top of the billet is approximately two times higher than at the bottom and that some large inclusions were found in both ends of the billet. What also can be noted is that the total number of inclusions in the casting with 2670 kg melt remaining is higher than in the casting with 1700 kg remaining melt, but it should be noted that the slices were not taken at exactly the same positions in the two cases. A clear difference between the castings with the lower amount of melt remaining at the end, table III, compared to the normal castings, table II and figures 3, and 4, is that a low melt level at the end gives a higher amount of inclusions at the end slices, and not in the bottom slices in contrast to the other cases.

The trend of inclusion distributions from both castings is illustrated in figure 8, and it shows that the number of inclusions is totally different in each position. It can be stated that close to the bottom of the billet in the first casting, a higher number of inclusions were observed. while in the second casting a strong increase was observed towards the top.

$:

to u(\ smnplf' IHllliliol1llllong; thtbillf'f .frollitilt bottOill (('In)

Figure 5. Inclusion distribution in first casting.

Table III. Inclusion distribution in casting with 2670 and 1700 kg remaining melt Sample position (em) Bottom Top Bottom Top

#(C)

#(M)

#(5)

71

72

71

115 35 65

98 19 20

100 23 23

#Total

214 313 49 108

Large size

17 17 4 3

Remaining melt (kg)

2670 2670 1700 1700

Samples from Cast House 2 In cast house 2, as it is described in the introduction, two castings normally are produced from one charge. First casting is done with 40 minutes holding time and about 15000 kg remaining melt after casting and second casting with 35 minutes holding time and about 3000 kg remaining melt. That means that the second casting can be considered as a casting with low melt level at the end of casting. The results in inclusion distributions are shown in figures 5-6. Figure 5, from the first casting, shows that at the start the number of inclusions is high, then it decreases dramatically (220 cm). A weak increase can than be seen towards the top of billet (end of casting). In addition, it can be stated that inclusions tend to appear in the central zone of the cross-sections. Furthermore, in these samples the number of large inclusions is negligible, except at the slice collected at 440 cm from the bottom. It can be seen in figure 6, from the second casting, that the number of inclusions at the very bottom of the billet (10 cm) is somewhat increased, then after a reduction at two positions, it is followed by a clear trend of increasing number of inclusion, which continues towards the top of the billet. It can be observed that at 440 cm from the bottom of the billet a sharp increase in number of inclusions was detected. Furthermore, these inclusions at 440 cm have large and shallow characteristics and were not found with the same frequency in other samples. Moreover, these shallow etching pits looked like agglomerates and were identified as aluminum oxide films by optical microscopy. Figure 7 shows the image of these large etch-pits and corresponding microscopic picture. Besides the mentioned sample (440 cm), there were also a noticeable number of large size inclusions at the bottom slice (start of solidification).

Figure 6. Inclusion distribution in second casting.

(a) (b) Figure 7. (a) Macro-photo image of shallow voids in sample from 440 cm from the bottom of the billet, (b) Micrograph of oxide film.

J.($

1.,1,0

z:t()

];:1H)

440

fi.(.()

Slllnple po,;,itlon ;llong tblt" btHet h'otn tillt" botnn (

FactSage™ 6.1 and the FactLite database was used for thermodynamic calculations in the ternary AI-Si-V system and the results were compared to existing literature on multicomponent phase diagrams [6-8].

...;u

,.

200

Liq. + Al + Si 2 V) that due to compositional changes during solidification is not likely to be crossed. In contrast, with nominal V concentrations ranging between 0.2 and 0.8 wt% Si 2V solidifies first. Upon further cooling and while the solute concentrations in the liquid changes. the general increase in V concentration displaces in particular the reaction of Liq. --> Liq. + Si 2V and consequently also the theoretical line of the reaction Liq. + Si2V --> Liq. + Al + Si2V to higher temperatures as shown by the sequence of isopleths in Fig. 10 (c) and (d) and in the liquidus projection. The resulting change of the formation temperatures for a-AI were calculated, summarized and compared to the nucleation temperatures of a-AI obtained from thermal analysis in Table IlL For the idealized case of a ternary Al-Si-V system assuming equilibrium conditions. the same trend of increasing temperatures was found. The FactSage™ calculations predicted a increase when the V concentration was increased to 0.8 wt%. It should be noted that the predicted formation temperature of a-AI is up to 7 K higher compared to the experimental data.

Figure 8 SEM-EDS spectrum from a spot measurement of the globular particle in Fig. 6 confirms the V enrichment of the AIsFeSi phase.

The results in Fig. 9 document a gradual V enrichment in the AlsFeSi phase with increasing nominal V additions to the melt. The data plotted here was obtained by measuring a statistically significant number of particles by EDS. When 0.06 wt% was added, only a few phase particles showed weak counts for V, mostly in Fe layers on eutectic Si crystals. This changed as the V concentration was increased to 0.8 wt%. Peak values of up to particles. wt% V were measured in the globular

Table III Comparison of temperature of formation of a-AI obtained by thermodynamic calculations and experiments. V concentration in wt% 0.5 0.8 Temperature Ref. 0.06 0.2

{;

(0C) 5

FactSage™ Experiment

4.71 4

617.8 612.6

618.7 613.8

620.7 614.4

622.6 615.9

Discussion

.S 11 :>

According to Mondolfo [9] V has a mild grain refinement effect. but compared to conventional grain refiners such as Ti and TiE2 vanadium is rather inefficient. The phase that comes into question to add to grain refinement effects is Al3 V which has an Al3 Ti-type crystal structure. Thus V could theoretically substitute for Ti and act as the coating layer on TiE2 [10]. There is. however. considerable disagreement in the literature to what extent V plays a role as grain refiner. Edwards et al. [II] report a decrease of grain size in a near eutectic AI-Si piston alloy with a combined addition of Ti, Zr and V. Maitland [12] reports a slight grain refinement with V addition. This was most prominent for a combined addition of Mn and V in wrought A1MgSi alloys. Others deny any effect of V [13]. In the present study we cannot define a trend consistent with V additions. The grain size was rather stable with some scatter in a band between 200 and 250 /lm. We argue that high Si concentration of7 wt% as in the A356 alloy does not allow for the formation of AI3 V. Even if the Si concentration was lower « 5 wt%) and the V content sufficiently high (> 0.5 wt%) to allow primary crystallization of Al3 V. it seems unlikely stable Al3 V particles would form, because solidification proceeds through a cascade of univariant reactions

3: 1 0

617.0 612.2

O.l 05 Nominal V addition in wt'M.

Figure 9 Relationship of overall V concentration and V content of the phase. The standard deviation (J is given as error bars.

Thermodynamic calculations The liquidus projection of the AI-rich corner in the ternary Al-SiV as calculated with FactSage™ between 540 and 800°C in the composition range of the present investigations is shown in Fig. 10(a). An enlargement of the liquidus projection in the region of the ternary eutectic point (Liq. --> Al(s) + SiCs) + Si 2 Yes)) is shown in Fig. lOeb). Isotherms are plotted as dotted lines, whereas straight lines show the eutectic valleys. The primary crystallization fields of AI3V, AI23V4' AI7V. AllOV (in order of formation). AI. Si and Si 2V are outlined. The main liquidus line intersects a number of peritectic points in the sequence of the AIV compounds until the ternary eutectic point at 573.4 °C is

[7].

A polyhedral phase precipitated at V concentrations of> 0.2 wt%. This phase was indentified by its at% ratio as Si 2V. The particles

1026

Wdgll, r_l1. 51

(b)

(a) 800

740

!laO

" i=

t.i

;: 620

560

O(J2

(J,()4

(l1

0.06

0.12

••

Figure 10 Liquidus projection of the AI-rich corner of the ternary AI-Si-V phase diagram (a); enlargement of the eutectic valley at the ternary eutectic point (b); and vertical sections at 0.06 wt% V (c) and 0.8 wt% V (d) calculated with FactSage™.

were located close to dendrite tips or within the AI-Si eutectic close to the dendrite-eutectic interface. This indicates that Si 2V nucleated and grew during the early stages of solidification and was pushed into the interdendritic spaces by the growing a-AI dendrites. Factsage™ calculations confirm this. They predict primary crystallization of Si2 V for V concentration exceeding 0.06 wt% and primary crystallization of a-AI, given that solidification proceeds sufficiently slow to meet equilibrium condition as predicted in the FactSage ™ calculation. Furthermore, it was shown that a temperature window exists for simultaneous crystallization of a-AI and Si 2V, when the eutectic valley which slopes to the ternary eutectic point is reached. Phase diagram evaluations in the literature [6, 7] are in contradiction to the Factsage ™ calculations. They predict primary crystallization of a-AI for the Si content typical for an A356 alloy and V additions of up to 0.8 wt%. An example of a liquidus projection in the AI-rich corner of the ternary AI-Si-V diagram from the literature shown in Fig. II (note the phase composition is given in at%). However, experimentally obtained temperatures for the a-AI reaction are consistent with the present thermodynamic calculations. Indicating that for the cooling rate of 1.1 K/s the primary crystallization field of Si2 V extends further into the reported phase field of a-AI.

Figure 11 AI-rich part of the liquidus surface projection of the AI-Si-V system with primary crystallization fields and magnification of the AI-rich corner [7].

1027

V additions were found to slightly change plate-like to more globular structure. Similar findings were reported by Rao [14] for an A319 alloy. It was shown that coarse plate-like Fe-rich phases were modified and formed complex intermetallics with the tendency of possessing equiaxed morphology. Furthermore, from investigations of an AIMgSi alloy it is concluded that V promotes the formation of the (l(A1FeSi) phase [12]. The correlation of Fe and V can be observed by inspection of the binary Fe-V, ternary Al-Fe-Si and quaternary Al-Fe-Si-V system. It becomes apparent that a number of binary and complex Fe and V containing intermetallics can form. These phases are however, not likely to appear in the composition range of an A356 alloy. Others report the substitution of Fe by V in e.g. Fe3AI [15]. Moreover, almost every FexAlx phase dissolves some V. Vanadium concentrations in these phases can reach 10 at% even at room temperature, indicating that the increase of V concentration in the particles is due to incorporation of V into the lattice during the growth. However, the exact mechanism for the V enrichment in and the morphology change was out of the scope of the present investigations and remains unclear. Here, further studies are needed to understand the effect that V exerts on Ferich intermetallic phases.

References I. T. Furu et aI., "Trace Elements In Aluminium Alloys: Their Origin And Impact On Processability And Product Properties" (Paper presented at the 12th International Conference on Aluminium Alloys, Yokohama, Japan, 2010). 2. G. Jha, F. Cannova and B. Sadler, "Increasing Coke Impurities - Is this really a problem for metal quality?" Light Metals 20i2, 2012,1303-1306. 3. J. Grandfield and J.A. Taylor, "The Impact of Rising Ni and V Impurity Levels in Smelter Grade Aluminium and Potential Control Strategies," Materials Science Forum, 630 (2009), 129-136. 4. Lennart Backerud, Guocai Chai, and Jarmo Tamminen, Solidification Characteristics of Aluminum Alloys. Vol. 2. Foundry Alloys. (Des Plaines, IL, AFS, 1990) 127-150. 5. J. Tamminen, TAW32 - Thermal analysis for Windows. 2002. 6. V. Raghavan, "AI-Si-V (Aluminum-Silicon-Vanadium)," Journal of Phase Equilibria and Diffusion, 32 (01) (2010), 68-71. 7. B. Huber, H.S. Effenberger and K.W. Richter, "Phase equilibria in the Al-Si-V system," intermetallics, 18 (4) (2010),606-615. 8. E. Gebhardt, G. Joseph, "Uber das Dreistoffsystem Zeitschriji fur Aluminium-Silizium-Vanadium," Metallkunde, 52 (1961), 310-317. 9. Lucio F. Mondolfo, Aluminum alloys: structure and properties (London, Butterworths, 1976),392-394. 10. P. Schumacher et aI., "New studies of nucleation mechanisms in aluminium alloys: implications for grain refinement practice," Materials Science and Technology, 14 (5) (1998), 394-404. II. W.M. Edwards et aI., "Development of Near-Eutectic AI-Si Casting Alloys for Piston Applications," Materials Science Forum, 625 (2002), 396-402. 12. A.H. Maitland and D. Rodriguez, "Vanadium in Aluminium," Proceedings of the 8th international Light Metals Congress, 1987, 423-425. 13. P.S. Cooper, M.A. Kearns, and R. Cook, "Effects of residual Transition Metal Impurities on Electrical Conductivity and Grain Refinement of EC Grade Aluminium," Light Metals 1997, 1997, 809-814. 14. A.K. Prasada Rao. "Influence of Vanadium on the Microstructure of A319 Alloy," Transactions of the indian institute ofMetals, 64 (4-5) (2011), 447-451. 15. G. Effenberg et aI., eds., Landolt-Bornstein, Ternary Alloy Systems Phase Diagrams, Crystallographic and Thermodynamic Data, vol. IIA3 (Springer-Verlag Berlin, Heidelberg, 2005), 1-10.

Conclusion The influence of V on the microstructure and solidification path of an A356 foundry alloy was investigated by thermal analysis, microscopy and thermodynamic calculations. The concluding remarks of the present study are as follows: 1. Increasing the V concentration from 0.06 wt% to 0.8 wt% results in a shift of nucleation, minimum and growth temperature of a-AI compared to a reference alloy by up to 4.5 K. The undercooling for the a-AI reaction remains unaffected. 2. The addition of V has no apparent influence on the grain size of a-AI. 3. V additions exceeding 0.06 wt% result in the formation of the Si 2 V phase with a distinctive polyhedral morphology. 4. Experimentally obtained data for the prediction of the solidification path of the A356 alloy correspond well with thermodynamic calculations using FactSage™ software. The sparsely available phase diagram data in the literature for the Al-Si-V system appears to be incomplete and might be misleading for alloy compositions in the V-lean region. 5. The phase becomes enriched in V with increasing nominal V additions. It appears as if this is accompanied with changes in the morphology of Acknowledgement Funding by Hydro Aluminium AS (Norway) is gratefully acknowledged. Thanks are also due to Jan Ove Havik from S0fNorge Aluminium AS (Norway) for the generous supply of AlV master alloy.

1028

Aluminum Cast Shop IV

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

Influence of die and casting temperatures and titanium and strontium contents on the technological properties of die-cast A356 in the as-cast and T6 condition Sebastian F. Fischer l, Veronika F. Grotenl, Johannes Brachmann l, Carolin Fixl, Thomas Vossel l, Andreas Blihrig-Polaczek l lFoundry-institute, RWTH Aachen University; IntzestraJ3e 5; 52072 Aachen, Germany Keywords: Taguchi, interaction effect, AlSi7Mg temperature of 700 DC to 720 DC seems to be ideal since this temperature supports the feeding of the melt and the stability of nuclei [19]. The die temperature should be high enough to ensure complete filling of the cavity and should also be low enough for an increased cooling rate. The latter causes a lower dendrite arm spacing (DAS) and grain size leading to increased mechanical properties [20-23]. Mostly however, the single effects of the particular processing steps on the mechanical properties of A356 are investigated and therefore only limited statements exist about the interaction and the intensity of the effect of these parameters. To optimize A356's technological properties, the influence of the interaction of its processing parameters is vitally important. In the present study, the die and casting temperatures, and the titanium and strontium contents were varied according to an orthogonal LlS-Taguchi array to show their single and interaction influences on the mechanical properties of an A356 in the as-cast and T6 states. These results allow the investigated parameters to be optimally set to both increase mechanical properties as well as to minimize energy and material usage leading to minimized production costs.

Abstract Automobile manufacturers demand downsized components manufactured from materials possessing increased technological properties. To fulfill this, the processing of existing materials must be optimized. In the present study, the individual effects and interactions of the melt's treatment and process parameters on the mechanical and metallographic properties were investigated in the as-cast and T6-states. This was carried out with the help of the Taguchi method to optimize the technological properties of this much used die-cast A356 alloy. The interaction effects of the parameters considered in this study are hardly addressed in the literature. However, the results of the present investigation show that this aspect is fundamental for optimizing the technical properties of a die-cast A356 alloy. For example, the titanium content of 0.20 wt.%, conventionally used for grain refinement, can be reduced by up to 75 % when the die temperature is relatively low (here: 190 DC). This leads to increased as-cast technological properties and cost savings. Introduction

Materials and Methods In times of increased energy prices and stricter environment laws, downsizing of castings is one key to reduce the fuel consumption and to decrease the emission of automobiles [1-3]. Downsizing of castings represented by lower wall thicknesses demands a casting material with increased technological properties to withstand the service loads. Besides adapting the chemical composition, this objective can be achieved by optimizing the casting material's processing [4]. Aluminum alloys are gaining popularity in the field of automobile manufacturing since these materials combine a relatively low density with good casting and mechanical properties, e. g. A356 alloy [5-9]. The processing of A356 includes, besides the typical degassing procedure, the addition of a grain refining agent. This agent mainly contains titanium to decrease the size of the primary aluminum solid solution, and the eutectic silicon is modified by adding strontium. Both process steps have a large influence on the casting and the mechanical properties of aluminum-silicon casting alloys; which has already been investigated in numerous studies [7-lS]. Owing to the introduction of the grain refining agent and the associated change of the solidification morphology from exogenous to endogenous solidification, the melt's flowability, mold filling and feeding ability is increased, and the tensile strength and ductility of the casting is also improved due to an increased number of grain boundaries and smaller grains [13, 14]. Usually, to achieve the best grain refinement of A356, alloy producers and the literature suggest titanium contents of 0.20 wt.% [15, 16]. The modification with strontium (usually 200 ppm) causes a smaller, compact and fibrous eutectic silicon phase which leads to improved tensile strength and ductility of A356 castings [17, IS]. Besides these melt treatments, the casting process parameters also exhibit an influence on the microstructure and the technological properties of A356. To process A356, a casting

Production and testing of the specimens The A356 base material, with very low titanium and strontium contents (table 1), supplied by RHEINFELDEN ALLOYS GmbH & Co. KG (Rheinfelden, Germany), was melted using a 5.5 kW 7 kg resistance furnace (Nabertherm GmbH, Lilienthal, Germany) at a furnace bulk temperature of SOO DC in a SiC crucible (Aug. Gundlach KG, Grossalmerode, Germany). Table 1. Chemical composition of the A356 base material in weight percent. Si

Mg

eu

Mn

Ti

Sr

Fe

Al

7.32

0.3S

0.002

0.003

0.003

0.0004

0.09

rest

After the aluminum reached a temperature of 740 DC. the preheated modification agent was added as AlSrlO rods (KBM Affilips B.V., ass, Netherlands). After 20 minutes, the preheated grain refining agent AlTi5B 1 master alloy was introduced to the melt in form of rods (KBM Affilips B.V., ass, Netherlands). Following grain refining, the hydrogen content of the melt was measured with the aid of a partial pressure-density testing device (mk Industrievertretungen GmbH, Stahlhofen a. W., Germany) and, on reaching a density index greater than 1.5 %, the melt was degassed by introducing argon 4.6 to the base of the crucible using a graphite lance (HASCO Hasenclever GmbH + Co KG, Llidenscheid, Germany). When the hydrogen content and the particular casting temperature was stable, the melt was stirred and then cast with the aid of a pouring spoon into a Diez die, which was preheated to the specified mold temperature with the aid of an

1031

oil heating system. The die was coated with graphite in the area of the specimen and with an insulating coat in the area of the feeders according to norm P 372 of the Association of German Foundrymen (VDG). 45 seconds after casting, the specimen was removed. The quality of the melt was controlled by thermal analyses and spectrometer measurements. 10 specimens per test run were cast, whereof 5 specimens were heat treated according to a T6 treatment (530 °q6 h, water quenching, 165°qS h). From 4 cast rods, do = S mm tensile specimens were machined according to DIN EN 50125 and tested with the aid of a S033 Instron tensile testing machine using a cross head speed of 0.35 mmls according to DIN EN 10002. Metallographic sections were prepared from 3 different positions of one as-cast rod per test run; representing 3 different cooling rates (position A is the highest, position C is the lowest rate), by embedding the specimens in Araldit combined with the hardener Ren HY 956 (both from Huntsman, Germany). The specimens were grinded using abrasive paper (320 to 1000 grades) and polished using a VibroMet (Buehler, DUsseldorf: Germany) machine. Optical micrographs were taken at different magnifications using the Axio Imager AIm (Carl Zeiss, Oberkochen, Germany) light microscope. These images were analysed with respect to the number and the area of eutectic silicon particles using the image analysis software Axiovision (Zeiss, Oberkochen, Germany). The DAS was manually determined from the 50x magnified micrographs according to the Association of German Foundrymen's (VDG) Norm P 220; the average grain size was manually measured from the 50x magnified micrographs taken from metallographic sections which were etched according to Barker (figure 1).

gas porosity in the castings [25]. Besides the mean value of 0.20 wt.%. a titanium content of 0.05 wt.% and 0.35 wt.% was chosen to show the influence of an assumed too low and too high titanium content, respectively. The strontium content was varied from 50 ppm to 200 ppm and 350 ppm. Design of experiments A full factorial design prescribes a minimum of 54 test runs for the analysis of the single effects. To minimize the number of test runs and specimens as well as to visualize the interaction between the parameters, an orthogonal LIS Taguchi array was chosen. By using IS test runs, this array enables both the analyses of the single effects of the chosen parameters as well as that of the interaction between the die temperature with both the titanium content and with the strontium content [26]. Table 2 summarizes the adopted LIS array. Table 2. Orthogonal LIS Taguchi array with the settings of the individual test runs. interaction

interaction AID (E)

casting temp. (G) [0C]

50

1

690

200

2

720

die temp. (A) [0C]

titanium content (B)

[wt.-%]

AlB (C)

190

0.05

1

2

190

0.05

2

3

190

0.05

3

350

3

750

4

190

0.20

50

2

750

5

190

0.20

2

200

3

690

6

190

0.20

3

350

7

190

0.35

8

190

0.35

2

350

2

750

9

190

0.35

3

50

3

690

10

290

0.05

350

3

720

11

290

0.05

2

50

12

290

0.05

3

200

2

690

200

3

350

test run

13

290

0.20

290

0.20

2 3

15

290

0.20

290

0.35

17

290

0.35

18

290

0.35

[ppm]

720

200

14 16

strontium content (D)

720

750 750 690

50

2

720

350

2

690

2

50

3

720

3

200

750

The results of the tensile tests and the microstructure measurements were analyzed and evaluated with the aid of the statistical methods analyses of means (AN OM) and analyses of variance (ANOVA). The ANOM indicates the direction of optimization of the factors, in which the mean deviation from the total average caused by every factor level shows the main effect of every single factor level. Using the AN OVA, the statistical significance of an effect of a parameter on the command variable can be evaluated. For this, the total result is divided into single variances. The variance expresses the squared deviation of the particular average [27].

Figure 1. Grain size analysis of a metallographic section of test run 11 (table 2). Setting of the parameters To show the single and interaction effects of the specified parameters on the mechanical properties of A356, two settings for the die temperature and three settings of the remaining parameters were defined. The lower die temperature was set to 190°C to ensure a complete mold filling without cold runs. The upper die temperature was set to 290 °C in order to produce an appropriate difference to the lower setting. The lower and mean casting temperatures of 690°C and 720°C represented a typical overheating for the used alloy [19, 24]. The upper casting temperature was set to the maximum of 750°C to avoid marked hydrogen absorption by the molten aluminum and thus prevent

Experimental Results

In its as-cast state, the die temperature mainly affects the material's tensile strength and the elongation at fracture (table 3). By decreasing the die temperature from 290°C to 190 °C, the A356's tensile strength and elongation at fracture are significantly

1032

increased (figure 2). The image analyses show that this effect can be attributed to the influence of the reduced die temperature on the DAS and the size of the silicon particles (table 4 and figures 3 and 4a). With decreased die temperature and the associated increased cooling rate, the DAS and the silicon particle sizes are decreased which, according to the literature, lead to increased mechanical properties [20-23]. Except for very high cooling rates (position A), the grain size is not significantly influenced by a decrease in the die temperature (figure 4b).

phases (table 4, particle size) which disturb the microstructure and thus could lead to a decrease in the tensile strength [28].

....

!

J::

ie

110

330

200

325

no to

190

S! 180

315

"

310

1ii

f-value

tensile strength F

tensile strength T6

elongation at fracture F

die temp. (A)

2.83 l.l7

casting temp. (G)

190

290

10

::s

.. tl

J;l

i" ""

S! .;; c

l!!

die temperature ['C]

fa

"

.2

"

130

290

die temperature ['CJ

b)

Figure 2. Influence of the die temperature on a) the tensile strength and b) the elongation at fracture of A356.

is

0.15

strontium content (D) interaction AID (E)

170

a)

elongation at fracture T6

titanium content (B) interaction AIB(C)

$

...""

.£:

t:

Table 3. ANOVA results of the tensile tests. Tabulated critical fvalue is 4.08 for the die temperature and 3.23 for the remaining parameters (95 % confidence level). Significant factors are highlighted.

l

!

0.31

0.47

0.88

2.41 1.33

i

0.34

1.76 2.95

!

0.75

0.76 0.42

Z3 21

19 17 15

2.10

b)

1.04

190

290

die ttmpt!0- 2.4% (3)

Table I shows the low sulfur cokes have a relatively low AR and the higher S cokes show a higher AR. The R2 correlation for sulfur and AR is 0.66 and for vanadium and AR is 0.49. The equation developed by Hume [3] which includes S, V and Na does not show an improved R2 value for the fast method, but it does for the slow heating method (0. 5°C/min), with an R2 similar to what was reported forthe equation.

Figure 4 shows the trend in real density for the same cokes. The low sulfur coke shows a steady increase in real density with temperature whereas most of the higher sulfur cokes show a decrease in real density with the onset of thermal desulfurization. The highly isotropic coke with a sulfur level of 5.6% shows quite a different temperature-real density relationship as a result of its significantly different texture or microstructure. 2.15

Figure 1 shows that the air reactivity decreases as the temperature increases as has been previously reported [10,11,12], and then increases as porosity opens due to desulfurization, confirming what has been reported previously [3]. The data in Table I and Figure I highlight the wide range in coke air reactivities for cokes that are all currently used routinely in anode blends. Many of these cokes could not be used today if historical coke air reactivity specifications were in place.

'" 2.00 Qj cr:: 1.95 1150 1200 1250 1300 1350 1400 1450 1500

Temperature I'C) Figure 4: Real Density (-200 microns fraction)

Figure 5 shows the change in the average crystallite size or Lc as a function of temperature. The Lc is not affected by coke desulfurization because it is a measure of the degree of ordering of the carbon structure. At higher temperatures, the carbon structure becomes more ordered and the average crystallite size increases. It is not affected by changes in the micro-porosity of the coke like real density.

-

1.50 1.00

45

0.50

40

VI

9

2.05

Qj

Cl

2.00 E

2.10

l:

Temperature vs air reactivity curves using the slow method (0. 5°C/min) are shown in Figure 2. The general shapes of the curves are similar to the fast method curves for all of the cokes shown. The air reactivity results for the 6.5% sulfur coke using the slow method are not shown in Figure 2 because all the results were out of the equipment measurement range (i.e. > I.S%/min).

:?

3.8%

"U u

3.8% " 5.1% [email protected]% -1iIr- 5.6% _6.5%

35

0.00

.... 30 u

1150 1200 1250 1300 1350 1400 1450 1500

Temperature ('C) Figure 2: ISO AR - 0.5°C/min (slow method)

25 20 1150 1200 1250 1300 1350 1400 1450 1500

The slow method was developed in response to earlier work [13] to provide an air reactivity result that was more indicative of the air reactivity of an anode when considering diverse impurities in the coke, and although there was some improvement, the correlation was not good in practice. Several papers have been published using data that was developed using the slow method [ 12,14,16,1S].

Temperature I'C)

Figure 5: Crystallinity (Lc) of Calcined Coke The specific electrical resistivity of coke is also unaffected by thermal desulfurization and it shows an almost linear decrease with temperature for all cokes, Figure 6.

Figure 3 shows the sulfur level as a function of temperature for the same six cokes shown in Figures 1 & 2. The effect of thermal desulfurization for the high sulfur cokes at high temperatures is obvious. The temperature range in the lab experiments is higher than those used in a rotary kiln and the very high levels of thermal desulfurization would not be expected during normal production. The low sulfur coke shows almost no desulfurization which is typical of low sulfur cokes.

0.060 c:

E cr::

6.0

g

'"'''''*'''''''''

1150 1200 1250 1300 1350 1400 1450 1500

5.1 %

4.0

[email protected]% -1iIr- 5.6%

3.0

""ft"" 6.5 %

:::I

VI

0.020

_3.8%

5.0

Temperature I'C) Figure 6: SER of Coke as a Function of Temperature The data in Figures 3-6 highlight the importance for calciners to be aware of the calcination behavior of different green cokes in blends. The data in Figures 4-6 in particular, highlight the importance of measuring both the real density and Lc or the real density and electrical resistivity when calcining cokes with widely varying sulfur levels. If a calciner is only measuring real density

2.0 1150 1200 1250 1300 1350 1400 1450 1500

Temperature I'C)

Figure 3: Desulfurization in High Sulfur Cokes

1052

to control the calcination level, it is very easy to start desulfurizing a blend without being aware of it.

The air reactivity shows a strong correlation with the temperature and decreases to the relatively low level of 0.13%/min (fast heating rate) at the highest temperature. The VBD on the other hand, shows a significant decrease at the final temperature illustrating the negative etfects of desulfurization on coke porosity. Tf the calciner was trying to meet a coke air reactivity specification of 0.20%/min, this could easily be achieved by increasing the calcining temperature but it would negatively affect coke porosity and would not allow the VBD specification to be met.

Tfthe calciner is trying to achieve a high real density target of 2.07 or 2.08 g/cm 3 for example, it may not be physically possible to achieve such a target with a blend containing high sulfur cokes due to thermal desulfurization and the presence of isotropic cokes with lower average real densities. An operator's natural inclination to increase the calcination temperature will typically make the problem worse and drive additional desulfurization and a further decrease in real density. The Lc test is useful for avoiding this problem.

This example illustrates why care must be taken when calcining cokes with different sulfur levels and structures, and when real density is used as the control measure. Most coke specitications are based on real density and it has become much more difficult to achieve high real density values in a rotary kiln with the range of green cokes used today.

Kiln Trial Data The above physical changes can be illustrated with data from a full scale kiln trial. Tn this case, a blend of green cokes with sulfur levels ranging from 1.4-6.5% was used to give a blend with an average sulfur level of 4.0%. Figure 7 shows the change in real density and Lc of the calcined product as the calcination temperature was increased.

In an ideal world, it would be beneficial to calcine cokes separately at different levels to avoid desulfurization of high sulfur cokes. This is not practical however given the higher number of cokes used in blends and the need to store each separately to allow blending after calcination. Calciner S02 permit limits also typically prevent high sulfur cokes being run separately. High sulfur green coke typically loses 12-15% sulfur during calcination and this would significantly increase S02 emissions relative to a blend of high and low sulfur cokes .

37.0 2.045 u

32.0

c

u .....

2.035 27.0

"iii

The data in Table 2 show the quality of some high sulfur cokes that were calcined separately in a rotary kiln, and are currently being used in anode blends. These blend cokes are typically used in lower percentages due to the high sulfur and vanadium levels, but their high air reactivity values would have disqualified them from use a blend coke in the past. Note the difference in real density results for cokes C and D even though they were calcined to a similar level.

&! 2.030 2.025

22.0 1600

1400

1500 Temperature (C)

1300

Figure 7: Real density and Lc for Kiln Trial The kiln temperature was changed over a relatively short period during this trial and the refractory brick temperature did not reach a steady state level at each temperature, which is the primary reason the coke real density and Lc did not change significantly over the temperature range of 1330 - 1430°C. The real density and Lc showed a more significant increase for the next data point and then a much larger change at the tinal temperature of I 560°C. Thermal desulfurization of the coke at the tinal temperature is very apparent from the decrease in real density. The Lc on the other hand, shows a steady increase consistent with the insensitivity ofthis test to desulfurization.

Table 2: High Sulfur Cokes Calcined in Rotary Kiln

·E

.......

e

0.850 0.845

0.25

0.840 ]'

0.20

'"

D

29.7

29.5

Real Density, -200flm (g/cmJ)

2.09

2.04

1.99

2.05

VBD, -28+48 (%)

0.91

0.85

l.01

0.88

Air Reactivity - Fast (%/min)

0.53

0.43

0.47

0.25

3.6

5.9

4.0

3.2

Calcium (%)

0.003

0.011

0.016

0.002

Tron (%)

0.008

0.017

0.033

0.017

Nickel (%)

0.038

0.020

0.026

0.017

Silicon (%)

0.017

0.025

0.024

0.005 0.011

Sodium (%)

0.008

0.010

0.011

Sulfur (%)

4.00

3.72

4.66

5.77

Vanadium (%)

0.074

0.049

0.062

0.050

Air Reactivity - A Brief History



calcination, especially when the calcination is performed under higher-than-standard temperatures [20]. High-temperature calcination might therefore be used to shift sulfur dioxide (S02) emissions from the pot rooms to the coke calciner. This would allow the capture of a larger portion of the coke S at the calciner. As the S02 concentration in the calciner flue gas is much higher than in the pot room gases, the scrubbing costs would be lower.

0.95 Coke C Coke A CokeD

0.75 1300

1400

Temperature [0C]

1500

Figure 7 Calcined coke VBD (Alcan method) after pilot calcination at different temperatures, heating rate: 50°C/min, residence time: 5 min.

However, thermal desulfurization is associated with creation of coke porosity. Anodes made with desulfurized coke therefore have undesired low bulk densities. It can be assumed that a portion of the porosity created during thermal desulfurization is lost during grinding of coarse coke particles. This might limit the negative impact of porosity creation on the anode properties. In order to verify this hypothesis it was decided to calcine a high-S coke at elevated temperatures and to manufacture laboratory anodes with the corresponding fines.

In addition to these expected results, some interesting observations were made. Due to its highly isotropic texture, coke C had a very high VBD (0.94g/cm 3, after calcination at 1300 0C). Calcination at 1500 °C reduced the S content by about one third. Even higher desulfurization was reached by increasing the residence time from 5 to 10 minutes. In spite of the associated porosity creation, the VBD was still acceptable (0.89g/cm 3, residence time of 5 minutes). In fact, coke C calcined at 1500°C had a similar VBD and S content as Coke A calcined 1300°C (typical for commercial kilns). Apparently, high-temperature calcination of a high-S green coke yielded a product close to medium-S coke calcined under standard conditions. Given these promising results it was decided to manufacture laboratory anodes with coke C, partly desulfurized at 1500 0C.

Pilot Calcination Runs Pilot calcination runs were performed with three different green cokes: cokes A, C, and D (Table I). The pilot kiln used was described in a prior publication [21]. An important advantage of this kiln is that it well replicates calcination on commercial scale. The calcination temperatures were varied between 1300 and

1060

Considering other impurities, the concentration of Ni and V increased during calcination at 1500°C (for example by 120 and 150ppm, respectively, for coke C, Tables 1 and 5). These nonvolatile impurities are concentrated as volatile organic compounds are lost during calcination. High-temperature calcination therefore cannot be used to reduce Ni and V. Laboratory Anodes with Desulfurized Coke Laboratory anodes were made with the reference coke and coke C, calcined at 1500°C. As compared to the reference coke, coke C had a higher bulk density and impurity concentrations (Table 5). In the reference recipe, the reference coke was used in all fractions, whereas in the test recipe the fines fraction «0.3 mm, Table 6) of the reference coke were replaced with coke C. Green anodes were manufactured at five different pitch levels. Properties of the test and the reference anodes are compared at their respective optimized pitch level, indicated by maximum baked anode density (BAD). Table 5 Cokes used for the manufacture of the laboratory anodes Coke C, calcined at Coke Reference 1500°C VBD [g/cm 3] 0.83 0.89 S [%] 1.1 3.5 V [ppm] 30 650 Ni [ppm] 141 300

Figure 8 Micrograph of a particle of coke C, calcined at 1500 °C, in a baked anode. Porosity appearing black. filled by pitch during paste manufacture. Thus, porosity created during desulfurization remains in the green and baked anodes. Micrographs of baked anodes showed that indeed voids in desulfurized coke particles were present (Figure 8). Other factors had a limited impact on the BAD, for example shrinkage and the in-situ pitch coking value were very similar for both anode series (Table 7).

Table 6 Coke fractions used for laboratory anode manufacture Mass [%] Fraction [mm] Recipe Max. Min. Reference Test 13.20 9.50 9.0 9.50 4.75 15.5 4.75 2.36 13.5 Reference 71.8 coke 2.36 1.18 14.0 10.0 1.18 0.60 Reference coke 0.60 0.30 9.8 0.30 0.15 5.2 Coke C, 0.15 0.075 3.4 28.2 calcined at 0.045 2.6 0.075 1500°C 0.045 17.0 0.000

Furthermore. use of desulfurized coke increased the electrical resistivity and decreased the Young's modulus considerably. Presently, no explanation for this observation can be given. Finally, the oxidation behavior of the anodes can be explained by the different impurity content of the cokes. As compared to the reference coke, the de sulfurized coke contained much more Ni and V which favored air oxidation. However, even after partial desulfurization, the S content was still high enough to inhibit CO 2 oxidation. It can be concluded that introduction of desulfurized coke in the fines fractions considerably deteriorated important anode properties, particularly the BAD.

Upon the use of de sulfurized coke C in the fines, the green and baked anode densities decreased significantly (-0.04g/cm3, Table 7). This appears surprising as coke C has a higher VBD than the reference coke (Table 5). However, it is known the thermal desulfurization preferentially creates microporosity [22] that is not

Summary and Conclusions The concentration of critical impurities (S, Ni, and V) in anode coke is expected to increase. These elements are evenly dispersed in the coke matrix, which makes physical separation techniques inapplicable. Chemical methods, performed at mild conditions as required for commercial processes, did not reduce impurities substantially enough to develop a viable process. A significant proportion of the S can be removed via high-temperature calcination. The trade-off is an important bulk density loss. Even under favorable circumstances (high bulk density coke, use of desulfurized coke in the tines), the quality of the corresponding anodes is poor. Impurity removal from cokes at the commercial scale is therefore unlikely to be viable in the near future. Thus, other techniques should be studied or revisited.

Table 7 Change of anode properties upon replacement of reference coke fines «0.3 mm) with coke C, calcined at 1500 °C Property Change* -0.04 Green anode density [g/cm 3] In-situ coking value [%] +0.7 Shrinkage [%] 0.0 Baked anode density [g/cm 3] -0.04 Electrical resistivity [/lnom] +9.1 -30 Air residue [%] Young's modulus [GPa] -1.0 CO 2 residue [%] +7.5 * At respective optimized pitch contents

1061

References 21. 1. 2. 3. 4.

5. 6.

7.

8.

9.

10. 11.

12.

13.

14. 15.

16.

17.

18.

19.

20.

L. C. Edwards et aI., "Evolution of Anode Grade Coke Qual ity", Light Metals, 2012, 1207-1212. "Pace Petroleum Coke Quarterly", Document of Jacobs Consultancy, Houston, TX, USA, 2012. T. W. Dixon., "US Refining Economics - A Model Based Approach", Light Metals, 2009, 941-944. A. Marafi, A. Hauser, A. Stanislaus., "Atmospheric Residue Desulfurization Process for Residual Oil Upgrading: An Investigation of the Effect of Catalyst Type and Operating Severity on Product Oil Quality", Energy & Fuels, Apr. 4, 2006, 20, 3: 1145-1149. Nongbri, G., Alpert, S. B., and Wolk, R. H.: "Production of Coker Feedstocks", US Patent 3,773,653,1973. E. Sturm, G. Wedde., "Removing impurities from the aluminium electrolysis process", Light Metals, 1998,235240. S. Broek, B. Rogers., "The Origin and Abatement of S02 Emissions from Primary Aluminum Smelters", Light Metals, 2009, 999-1005. Stewart, M. "Petcoke Industry Overview: A Growing Market in a Shrinking World", 11th Annual Petcoke Conference, Orlando, Fl, USA, 2012. U. Mannweiler, R. E. Frankenfeldt., "Techno-Economical Comparison of Various Sulfur Removal Processes: Crude Oil - Petroleum Coke and Alumina Reduction Process", Light Metals, 1985, 1395-1410. J. Xiao et aI., "Study on Removal of Sulfur from Petroleum Coke by Mixed Acid", Min. Metall. Eng., 2010, 30,62-65. H. Shlewit, M. Alibrahim., "Extraction of sulfur and vanadium from petroleum coke by means of salt-roasting treatment", Fuel, 2006, 85, 878-880. 1. Alvarado et aI., "Extraction of vanadium from petroleum coke samples by means of microwave wet acid digestion", Fuel, 1990,69,1: 128-130. H. Al-Haj-Ibrahim, B. 1. Morsi., "Desulfurization of Petroleum Coke: A Review", Industrial and Engineering Chemistry Research, 1992,31,1835-1840. H. H. Brandt, R. S. Kapner., "Desulfurization of Petroleum Coke", Light Metals, 1984, 883-887. P. Agarwal, D. K. Sharma., "Studies on the Desulfurization of Petroleum Coke by Organorefining and Other Chemical and Biochemical Techniques Under Milder Ambient Pressure Conditions", Petroleum Science and Technology, 2011,29, 1482-1493. L. Zeng, C. Y. Cheng., "A literature review of the recovery of molybdenum and vanadium from spent hydrodesulphurisation catalysts: Part I: Metallurgical processes", Hydrometallurgy, 2009, 98, 1-9. Hay, S . .I., "The Formation and Fate of Carbonyl Sulfide (COS) Gas in Aluminium Smelting", Ph.D. Thesis, The University of Auckland, Auckland, New Zealand, 2002. R. W. Bryers., "Utilization of petroleum cokes for steam raising", Prepr. Pap. -Am. Chem. Soc. , Div. Fuel Chem., 1993. M. Constanti, 1. Giralt, A. Bordons., "Desulphurization of dibenzothiophene by bacteria", World Journal of Microbiology and Biotechnology, Sept. 1, 1994, 10, 5: 510-516. L. C. Edwards, K. J. Neyrey, L. F. Lossius., "A Review of Coke and Anode Desulfurization", Light Metals, 2007,

22.

1062

895-900. M. 1. Dion et al.. "Prediction of calcined coke bulk density", Light Metals, 2011, 931-936. P. J. Rhedey., "Structural changes in petroleum coke during calcination", Transactions of the Metallurgical Society ofAIME, 1967,239,1084-1091.

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

Calcined Coke Round Robin 19 and the Precision of Bulk Density Tests Marvin Lubin [, Les Edwards [, Lorentz Petter Lossius 2 [Rain CIT Carbon, 2627 Chestnut Ridge Rd, Kingwood, TX, 77345, USA 2NorskHydro ASA, P.O. Box 303, 0vre Ardal NO-6882, Norway Keywords: Calcined Coke, Anode, Bulk Density Abstract Round robins (RR's) are useful for laboratories to benchmark performance against other labs. RR 19 was a collaboration between Rain Cll, Hydro Aluminium and R&D Carbon and was organized after the special session on coke bulk density arranged by TMS and ASTM at the 2011 TMS Annual Meeting. Five calcined coke samples representing a range of chemical and physical properties were prepared and sent to 28 laboratories around the world. A key objective was to compare the repeatability and reproducibility of different bulk density and apparent density methods. The paper discusses the organization of RRI9 and presents a statistical analysis of the following quality parameters: S, V, Ni, Fe, Ca, Si, Na, P, real density, Lc, VBD, TBD and Hg apparent density. In a companion paper, the properties of bench scale and pilot scale anodes produced with the cokes are presented along with correlations to coke properties.

Table 1: List of Participating Labs

Introduction This worldwide ASTM round robin (RR) was a collaborative effort between Rain Cll Carbon, Hydro Aluminium, and R&D Carbon organized after the 2011 TMS special session on coke bulk density testing. The primary focus of the RR was to examine the repeatability and reproducibility of different bulk and apparent density tests currently used in the industry. Real density, Lc' sulfur, and trace metals were added to the RR to generate additional precision data on these properties. It was the 19th round robin organized by Rain CIT and is hereafter referred to as RR 19. A companion paper published in these proceedings [1] reports on the relationship between calcined coke properties and pilot scale anodes produced with the five cokes from RRI9. A particular focus is on bulk density correlations and the paper provides some additional background information and references from the 20 II TMS bulk density session [2]. A key problem addressed by this work is the lack of a universally accepted test method for measuring coke bulk density which provides both good repeatability and good reproducibility, and some level of predictability with respect to anode quality. As a result, at least three different coke bulk density tests and specifications are in common use making it very difficult to compare results between laboratories and anode plants.

AJ Edmond - Long Beach

USA

AJ Edmond - Mead

USA

Alcoa Aluminerie Deschambault

Canada

Alcoa- Europe Spain

Spain

Alcoa Lake Charles

USA

Aluminerie Alouette

Canada

BHP Billiton Hillside

South Africa

BHP Billiton Mozal

Mozambique

Boyne Smelters Limited (RTA)

Australia

BP Cherry Point Refinery

USA

BP Europa SE

Germany

BP Wilmington

USA

Dubai Alurniniurn Company

UAE

Emirates Aluminum Company

UAE

Hydro Aluminiurn, Ardal

Norway

Hydro Aluminiurn, Porsgflurn

Norway

Hydro Aluminium, Sunndal

Norway

New Zealand Aluminium Smelters Ltd

New Zealand

Petrocoque

Brasil

R&D Carbon

Switzerland

Rain ClI Carbon Lake Charles

USA

Rain ClI Carbon Moundsville

USA

Rain CII Carbon Vizag

India

RTA Arvida Reseacrh and Development Center

Canada

RT A Centere-Analytique Vandreuil

Canada

RTA LRF

France

Statoil

Norway

Tomago Aluminium Company

Australia

Round Robin Plan A wide range of laboratories were invited to participate in RR 19 and Table 1 shows the list of 28 labs that participated. There was no cost for participants and RRI9 ran from October 20 II to January 2012. A total of five calcined petroleum coke samples with a wide range of properties were sent to each lab. The coke samples originated from 300 kg lots that were homogenized and divided using rotary sample dividers as shown in Figure 1. A large 150 kg lot was sent to R&D Carbon for bench scale and pilot anode preparation and testing.

Figure 1: Rotary splitter used for sample preparation Each lab was asked to split their coke sample into two and perform the preparation and sample analyses for each split in duplicate. Laboratories were asked to undertake all analyses that

1063

they were capable of running using industry accepted standards from ASTM and ISO. Sample Preparation Rain Cll supplied 3 coke samples, all produced with a rotary kiln: • Coke A - 100% straight run, low sulfur coke. • Coke B-1 00% straight run high sulfur coke with a highly isotropic structure. • Coke C - blended coke. Hydro supplied 2 calcined coke samples: • Coke S - 100% straight run, low sulfur coke. • Coke HB - blended coke.

Figure 2: GeoPyc Bulk Density Measurement

The two blended coke samples were prepared differently. Coke C contained five different green cokes blended prior to calcination. The blend is used routinely by several smelters in North America. The HB blend was generated by Hydro by blending a low sulfur, straight run calcined coke with a US Gulf Coast coke blend. It is typical of the quality used at two Hydro smelters in Norway.

Sulfur and Trace Metal Impurities Sulfur and impurity elements including Na, Si, P, Ca, V, Ti, Mn, Fe, and Ni were analyzed by several methods as follows: • XRF - ASTM D6376-06 (2010) and ISO 12980 (2000) • ICP - ASTM DS600-04 (2009) and ISO 14435 (2005) • AA - ASTM DS056-02 (2007) and ISO 8658 (1997)

Analytical methods The tests performed for the RR are summarized below.

Lc: Average Crystallite Height by XRD • ASTM D5187-91 (2010) and ISO 20203 (2005)

Vibrated Bulk Density (VBD)

Real Density

Traditional VBD equipment requires the use of a vibrating feeder, graduated cylinder and a vibrating table and usually measures a particular size fraction. Multiple crushing and screening steps are required to prepare samples for the VBD tests examined in RR19.

Real density was analyzed by xylene, helium pycnometry, or calculated from Lc using the following standards/procedures: • Helium - ASTM D2638-1O and ISO 21687 (2006) • Xylene -ISO 8004 (2010) • Lc - real density calculated from the Lc result using an algorithm based on many comparative analyses.

• ASTM D4292-10 - "Standard Test Method for Determination of Vibrated Bulk Density of Calcined Petroleum Coke". Samples were prepared to 28x48 mesh (0.3-0.6mm). • ASTM D7454-08 - "Standard Test Method for Determination of Vibrated Bulk Density of the 1.17 - 4.7 mm Calcined Petroleum Coke Fraction Crushed to 0.42 - 0.83 mm, using a Semi-Automated Apparatus".

Round Robin Results RR19 was a proficiency RR where all participants' results are shown as returned. A set of statistical tools designed for evaluating consistency was used to determine possible outliers. The outliers were removed from the between-lab averages and the standard deviations to make the averages and the standard deviations representative of normal analysis levels and ranges. All precision calculations are according to ASTM E691.

Tapped Bulk Density (TBD) This test is similar in principal to a VBD test but uses tapping equipment instead of a vibrating table. The method is based on ISO 10236 (1995) utilizing naturally occurring fractions. There is no sample preparation involved other than screening at sizes of: •

This paper is a summary of the much larger and complete RR reports issued to the participating labs involved [5,6] The RR reports will be submitted as official ASTM reports so anyone will be able to request a copy in the future.

0.25-0.5 mm, 0.5-1 mm, 1-2 mm, 2-4 mm and 4-8 mm. The last four were tested in RRl9

GeoPyc - Trans Axial Pressure Mode (TAP)

Vibrated Bulk Density - Standard and GeoPyc Comparison The determination of the bulk density of calcined petroleum coke is an important property because it is an indirect measure of coke porosity which influences anode pitch demand and density.

The GeoPyc method measures bulk density by controlling the force and measuring the displacement of a teflon plunger used to compact the bed of coke, Figure 2. The method has been described previously [3] and an ASTM standard is currently under development. It is important to note that no standard was available for RR 19 so each lab selected their own instrument measurement parameters. The GeoPyc equipment can be used for measurement of bulk density using any of the preparation methods in common use. Seven labs with GeoPyc equipment participated in RR19 and reported results for the following preparation methods:

ASTMD4292 The ASTM D4292 method requires the sample to be crushed and prepared to 28x48 Tyler mesh (0.6-0.3mm) using a specific and rather time consuming procedure. The prepared material is transferred using a vibrating spoon into a graduated cylinder which sits on a vibrating table. There is no industry standard sample to calibrate or check the equipment but the VBD set-up must be in accordance with the D4292 procedure.

• ASTM D4292 (28x48 mesh), ASTM D7454 (20x35 mesh) and ISO 10236 (0.5-1 mm, 1-2mm, 2-4mm and 4-8mm) Mercury Apparent Density (Hg AD) Based on the Pechiney Hg AD method [4].

1064

A total of 12 labs measured ASTM 04292 YBO's and it is the most widely used test in the industry for coke bulk specifications.

Although the repeatability is better using the GeoPyc equipment, just as it was for 04292 test, the results for the GeoPyc were slightly higher than the ST AS YBO results.

Seven labs that routinely measured YBO's by the 04292 method also had the GeoPyc equipment. All these labs agreed to analyze their "prepared" 28x48 mesh YBO samples on both the standard vibrating table equipment and the GeoPyc equipment. Results are shown in Table II. "Count" = number of labs that provided results but the number varies because outlier results are excluded.

Table III: D7454 Averages for VBD and GeoPyc (g/cm3)

A

Table II: D4292 Averages for YBD and GeoPyc (g/cm 3) C A C HB

S B

04292

12

0.834

0.027

0.080

04292 GeoPyc

7

0.831

0.029

0.090

04292

12

0.860

0.025

0.082

04292 GeoPyc

7

0.858

0.018

0.053

04292

11

0.878

0.035

0.108

0.020

HB

S B

6

0.787

07454 GeoPyc

3

0.794

0.003

0.006

07454 07454 GeoPyc

6 3

0.818 0.834

0.015 0.011

0.037 0.022

0.056

Methods

Count

Mean

Range 0.044

D7454

6

0.838

0.016

0.040

07454 GeoPyc

2

0.848

0.003

0.004

07454 07454 GeoPyc

6 2

0.856 0.867

0.020 0.002

0.058 0.002

07454

6

0.949

0.006

0.016

07454 GeoPyc

3

0.955

0.003

0.006

04292 GeoPyc

6

0.875

04292

11

0.903

0.044

0.145

04292 GeoPyc

6

0.893

0.024

0.067

ISO 10236

04292

12

0.993

0.024

0.089

04292 GeoPyc

7

0.981

0.022

0.064

The ISO 10236 TBO method uses a very simple preparation method which requires no crushing and separates the coke into different naturally occurring size fractions by screening. The bulk density of each size fraction is measured by feeding a graduated cylinder connected to a tapping device. Size fractions of 8x4mm, 4x2mm, 2xlmm, and IxO.5mm were measured in RRI9.

Results are shown in Figure 3 for an easy comparison. On average, the GeoPyc results are lower and more repeatable than the 04292 YBO results. 1.05 1.00

D7454

Std. Oev. 0.017

Coke

+:===0.993

A total of seven labs used this method and six labs had the GeoPyc equipment. The TBO results reported in Table IV are the average of all five RR samples for each fraction. The agreement between the TBD and GeoPyc averages was generally excellent.

iIIII D4292-Standard

Table IV: ISO 10236 TBD for Tap and GeoPyc (g/cm3)

0.95

Measurement Type

0.90

8x4mm TBO 0.85

8x4 mm GeoPyc 4x2mm TBO

0.80

4x2mm GeoPyc 2xl mm TBO

0.75

2xl mm GeoPyc lxO.5mm TBO

Figure 3: Lab average [g/cm 3] 04292 on standard equipment vs. GeoPyc shown with 3 decimals

IxO.5 mm GeoPyc

ASTMD7454

Count

Mean

Std. Oev. 0.015

Average Range 0.038

7

0.721

4-6

0.718

0.009

0.023

7

0.767

0.015

0.038

4-6

0.758

0.011

0.026

7

0.828

0.018

0.048

4-6

0.823

0.007

0.016

7

0.873

0.019

0.055

4-6

0.871

0.014

0.037

Apparent Density Using Mercury (Hg AD}

The ASTM 07454 method requires samples to be prepared and crushed to 20x35 Tyler mesh (0.85-0.60mm) and measured using the STAS semi-automated YBO equipment. The sample preparation method is time consuming just like the 04292 test. The STAS equipment uses a photo-electric sensor to detect the coke bed once it reaches the 50-mL mark in a graduated cylinder. This eliminates operator parallax errors reading heights in the graduated cylinder.

The Pechiney Hg AO test requires samples to be prepared to lOx20 mesh (1.7-0.85mm). Samples are then placed in a pycnometer and subjected to vacuum before Hg is added. This test has now been largely phased out for occupational health and safety reasons and few labs are able to run the test today. Hg AO results for the five RR samples are shown in Table Y. It is worth noting that the difference in Hg AO results for the five cokes was only vs 20-35% for the other YBO/TBO methods.

Six labs measured YBO by the 07454 method and three of these had the GeoPyc equipment for comparison. All results are shown in Table III. The ST AS equipment must be calibrated using coke standards whereas the GeoPyc requires no coke calibration standards.

1065

Apparent Density by Mercury (Hg AD) [g/cm3]

Table V: RR Lab Averages for Hg AD Ig/cm 3 1 Coke A B C HB S

Mean 1.718 1.763 1.724 1.723 1.738

Std. Dev. 0.021 0.012 0.021 0.023 0.009

Range 0.070 0.035 0.075 0.065 0.025

Repeatability (r) Reproducibility (R)

Improving the overall precision of bulk density measurements has been a focal point over the last few years. Precision data calculated for the various tests as a result of RRI9 are shown below. Repeatability refers to precision within the same lab and reproducibility refers to the precision or agreement between labs.

Repeatability (r) Reproducibility (R)

GeoPyc Precision

0.014 0.046

0.015 0.087

0.014 0.070

Table VI: Various Bulk Density Results for RR19 Samples

The stated reproducibility of ASTM D4292 is poor at 0.046 g/cc, and it was found to be even worse in RRI9 (0.087). Anode producers need better certainty of bulk density than ±0.05 g/ cm 3 when comparing potential coke supplies and evaluating conformance to coke VBD specifications. It was anticipated that there would be improved reproducibility for the D4292 test as a result of revisions made in 2010 but this was not observed. The test is poor for comparing results between different laboratories.

Method [g/cm 3] D4292 D7454 ISO 8x4mm IS04x2mm IS02xlmm ISO IxO.5mm

D7454 - ASTM [g/cm 3 ]

Un

Repeatability (r) Reproducibility (R)

Documented Precision 0.0036 n/a

RR-VBD Precision 0.013 0.043

RR-GeoPyc Precision 0.014 0.019

RR-TBD Precision 0.015

RR-GeoPyc Precision 0.019

0.052

0.026

An

A 0.831 0.787 0.629 0.655 0.713 0.770

C 0.860 0.818 0.700 0.725 0.791 0.830

HB 0.878 0.838 0.702 0.762 0.829 0.867

S 0.903 0.856 0.720 0.780 0.850 0.898

B 0.993 0.949 0.851 0.911 0.960 0.998

1.718

1.724

1.724

1.738

1.763

The Hg AD test was correlated with the various bulk density tests. Correlations were generally good and a correlation matrix for all tests is shown in Table VIT. The best correlation between the Hg AD and VBD tests was the ASTMD4292 test with an R2 of 0.98.

Alcan (now RTA) developed this method with a more automated measurement method compared to D4292. The within-lab repeatability is quite good, similar to the D4292 test. The reproducibility is better than the D4292 method but still quite poor overall. The GeoPyc results are much better but the results represent only three labs and are not valid according to E69l. ISO 10236 (1-2mm) [g/cm 3 ] Documented Precision Repeatability (r) 0.01 0.02 Reproducibility (R) 0.03 (2012)

0.023 0.051

Correlation between VBD/TBD and Hg AD The five samples used in RR 19 represent a wide range of low/high density cokes. Results for all the bulk density methods are summarized in Table VI. Coke C and HB both have the same Hg AD results but a significantly different bulk density using the D4292 and D7454 test methods. Coke A has a similar Hg AD to Coke C and HB but the VBD/TBD methods all show a significantly lower density.

D4292 - ASTM [g/cm3] RR-VBD Precision

RR Precision

Similar to the D4292 VBD and D7454 VBD, Hg AD is an analysis that is useful within the same lab but has a larger uncertainty between labs. This makes it difficult to undertake reliable comparisons of cokes when using data from different labs and coke supplies. The spread in Hg AD for these cokes was low making it difficult to distinguish between the five cokes in RR19.

VBD, TBD, Hg AD: Repeatability and Reproducibility

Documented Precision

Stated Precision 0.006 0.011

Table VII: Correlation Matrix of Various Density Methods D4292

07454

1.00

-

IS04x2mm

0.97 0.97

0.98 0.98

IS02xlmm ISO lx.5mm

0.95 0.97

0.96 0.98

HgAD

0.98

0.96

07454 ISO 8x4mm

ISO 8x4 mm

ISO 4x2 mm

ISO 2xl mm

ISO lx.5 mm

0.97 0.96 0.96 0.93

0.99 1.00 0.90

-

-

1.00

-

0.89

0.91

-

-

-

-

Sulfur and Trace Metal Results For the most commonly specified 1-2 mm fraction the observed reproducibility was 0.052 g/ cm 3. In October 2012, ISO revised the reproducibility to 0.03 g/cm3 which is more in line with the observed precision in RR19. For other size fractions, R varied between 0.035-0.054.

Sulfur analyses were performed by the following methods: 16 XRF, 2 AA, and 5 unreported. Sulfur is a very critical parameter for both quality and environmental reasons. Results for the sulfur content are summarized in Table VIII. The average of the reported values ranges from 1.21 to 4.46%, which reflects the wide range of cokes included in RR 19. The variability increases significantly as the sulfur content increases, which suggests that some labs have fewer reliable standards for high sulfur cokes since these fall outside the normal range that smelters typically analyze.

The repeatability for the GeoPyc equipment was comparable with the TBD equipment in RR 19. The reproducibility was better and should be further improved after development of a standard procedure for setting instrument parameters such as compaction force and number of consolidation cycles that all labs follow.

1066

Table VIII: Sulfur Content Coke

A B C HB S

Unit % % % % %

1% 1

Std, Dev.

Mean 1.50 4.46

Range 0.35

0.07 0.24 0.12

3.07 1.21 2.13

some issues between labs with Coke B which is a relatively hard and highly isotropic, low RD coke. Table X: Lab Averages for Real Density [g/cm'l

0.99 0.45 0.33 0.31

0.09 0.07

Coke A

B C HB S

Trace metals analyses were performed by the following methods: 16 XRF, 5 ICP, 1 AA, and 2 unreported. The majority of the RR19 participants analyzed the trace metals in conjunction with sulfur content by XRF. The results are summarized in Table IX.

C

Vanadium

ppm

97

593

392

Nickel

ppm

181

268

206

68

176

Iron

ppm

176

443

302

75

212

Silicon

ppm

69

155

157

44

255

Calcium

ppm

77

132

99

20

141

Sodium

ppm

49

102

47

35

63

Titanium

ppm

2

15

8

2

5

2.04

3

2.1)2

Manganese

ppm

A

B

S 147

Unit

2

4

3

1

Mean 2.078

24 24

2.067 2.065

1.995 2.065

Std. Dev. 0.008 0.012 0.007 0.009 0.007

Range 0.029 0.116 0.029 0.034 0.029

The greater variability and range of coke B is shown in Figure 5. Three labs calculated RD' s based on an Lc-RD algorithm and these are the high results shown with green markers. This method is clearly not suitable for isotropic cokes with signiticantly lower RD's. These lab results and the Lab 14 results are excluded from the average and standard deviation data for coke B in Table X.

Table IX: RR Lab Averages for Trace Metals Ippml Element

Count 25 25 25

HB 235

2.1(1

HI8

+ . - - - - - - - - - -........

2.06 j_ •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••.••- - - - - .

2.00

An example from the RR19 report [6] for vanadium is shown in Figure 4. It shows how each lab performed on an average basis for all five cokes. The graph highlights the consistency of the results with only three labs being clearly outside one standard deviation.

1.98

1.96 1.94 j

................................................................................................................................

.

Figure 5: RD lab averages for Coke B (Blue = not known); 25 labs, "Issue" means less than 5 cokes or 5 duplicates measured :l40

Lc, average crystallite height Lc is a measure of the average crystallite size and is directly proportional to the heat treatment the coke receives during calcination. All Lc results were performed using an XRD method and are shown in Table Xl.

300

Table Xl: RR average result for Lc [A] 200

Coke A

+......................................................................................................................................................................................................

240 ..J.......................................................................................................................................................................................................................

B C

,

Figure 4: RR Vanadium Averages (ppm) by Lab; 25 labs, "Issue" means less than 5 cokes or 5 duplicates measured

HB

S

Similar to sulfur, the variability of the trace metals (V, Ni, Ca, Si, Fe) increased significantly at higher concentrations and labs should consider acquiring a more extensive set of calibration standards.

Mean 30.4 30.8 27.7 29.3 27.4

Std. Dev. 0.5 1.0 0.7

Range

0.7 0.5

3.8 2.1

1.6 3.2 2.2

For sponge cokes, there is usually a good correlation between RD and Lc. The correlation for the five cokes in RR 19 was poor however with an R2 value of 0.25. Data are shown in Table XII. When the isotropic coke was removed (coke B) the R2 value increased to 0.77.

Real Density Real density is used as a measure of calcination level and results are shown in Table X. Real density is measured on a sample ground to -200 mesh (-75 /lm) and the density is measured using helium or xylene as the displacement media. The overall reproducibility between the 25 labs was acceptable, but there were

Regarding precision, the r&R observed in RR 19 for real density by He-pycnometry was 0.007 and 0.017 g/cm3, fairly close to the standard's r&R of 0.005 and 0.013 g/cm3. For Lc, the r&R

1067

observed in RRl9 was 0.9 and 2.2 A, somewhat higher than the standard's r&R of, 0.5 and 1.9 A.

ASTM D4292 or D7454 preparation methods. The results presented in RRl9 show that the equipment is versatile enough to use with most sample preparation methods.

Table XII: RD and Lc relationship Coke A

RDAverage 2.078

LcAverage 30.4

B

1.995 2.065 2.067 2.065

30.8 27.7 29.3 27.4

C HB

S

In the companion paper to this one [1], none of the bulk density tests or the Hg AD test stood out as being any better than the other for predicting baked anode densities. On the other hand, all the tests showed a strong correlation with optimum pitch level. Based on this, it seems reasonable for the industry to settle on bulk density tests with the best overall precision - particularly for cross-lab comparisons. When looking at the other tests in RR19, the precision for sulfur, trace element impurities, Lc and RD are all acceptable, notwithstanding previous comments about high S and trace metals levels and calibration standards. The only caveat to this is the calculation of RD from Lc results. This is not a recommended practice for any cokes or coke blends containing isotropic structures. Labs also need to take care with crushing and grinding harder cokes like coke B. This can contribute to additional iron contamination if non-tungsten carbide grinding equipment is used. Care also needs to be taken to ensure that real density samples are ground to 95% passing 200 mesh (75Ilm). Grinding times may need to be adjusted for harder cokes.

Discussion and Conclusions Overall. the results for the five samples in RRl9 showed reasonable consistency for tests other than bulk density. The isotropic nature and greater hardness of coke B drove some additional variation in RD results. The agreement between labs for sulfur and vanadium also deteriorated at higher S and V levels and this is believed to be due to a lack of suitable calibration standards. The detailed RRl9 report [6] shows a greater spread in results for trace metals like Si, Ca, Fe and Ni at higher concentrations and this is also believed to be due to a lack of calibration standards by some labs at higher concentration levels.

Recommendations Once the ASTM committee tentatively approves a new bulk density procedure as outlined above, a new RR will be initiated using prepared samples for most of the preparation methods, including 28x48, 20x35 and screened, naturally occurring fractions so that they can all be included in the ASTM precision statement.

RR 19 participants were able to analyze both the blended and single source coke samples with the same level of precision. Sulfur analysis by four different methods showed good precision among all labs, with only two labs being significantly outside of one standard deviation. Most labs are set up to measure bulk density using one method, and only two labs were able to run all bulk density tests. The VBD/TBD precision statements in the current standard procedures seem optimistic based on the RR 19 results but that is likely because some labs are not following procedures exactly as written including the use of non-standard equipment.

More labs are encouraged to participate in RR studies like this in the future so that consistency can be improved throughout the industry. ASTM started a new Proficiency Test Program in 2012 known as the ILS program. They will conduct industry wide round robins every 6 months with two calcined and two green coke samples as long as enough interest remains.

The above is not a new finding but it highlights the difficulty of running bulk density tests with complicated sample preparation procedures. For most tests, the within lab repeatability was much better than the between lab reproducibility. Sample preparation differences are clearly driving most of the variation between labs.

References 1. M. Lubin, L. P. Lossius, L. Edwards and J.Wyss, "Relationships Between Coke Properties and Anode Properties Round Robin 19," Light Metals, 2013

The ISO TBD test eliminates the difficult sample preparation steps involved with the two ASTM VBD tests and it offers significantly better precision than the ASTM D4292 test and about the same level of inter-lab precision as the semi-automated ASTM D7454 method - at least in the RRl9 study. The GeoPyc equipment improves the within-lab precision for most of the bulk density tests.

2. Petroleum Coke VBD Special Session, Light Metals, 2011, 925-963 3. M. Lubin, L. Edwards and J. Marino, "Improving the Repeatability of Coke Bulk Density Testing," Light Metals, 20 II 4. R. Barral, "Coke Apparent Density by Mercury Pycnometry," Aluminium Pechiney Standard Procedure, 1999, A.07.11.V06

The within-lab repeatability for measuring the bulk density on naturally occurring size fractions with the GeoPyc equipment showed the best overall repeatability in RR 19. This is perhaps not surprising given the fully automated nature of the equipment. The sample can be poured into the measuring chamber without the need for special vibrating feeders and the measurement is fully automated after this.

5. Lorentz Petter Lossius and Marvin Lubin, "ASTM-RCII-RDCHydro 2011 - Petroleum Coke Round Robin RRI9; Bulk Density and Hg Apparent Density", Report distributed to RR 19 participants June 21 st 2012 6. Lorentz Petter Lossius and Marvin Lubin. "ASTM-RCU-RDCHydro 2011 - Review of Petroleum Coke Properties Measured in RCII RR 19; RD, Lc, S and Metals", Report distributed to RR 19 participants March 27th 2012

Based on this, the ASTM committee has requested that a new bulk density procedure be developed using naturally occurring size fractions and the GeoPyc equipment. The procedure will be multifaceted and will allow measurement of naturally occurring size fractions like the ISO 10236 test or samples prepared using the

1068

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

A METHOD FOR THE RAPID CHARACTERIZATION OF PETROLEUM COKE MICROSTRUCTURE USING POLARIZED LIGHT MICROSCOPY Andris Innus, Alain Jomphe, Hans Darmstadt Rio Tinto Alcan, Arvida Research and Development Centre, 1955 Boulevard Mellon, Jonquiere, QC, G7S 4K8, Canada Keywords: Coke, Anode, Structure, Microscopy Abstract

Coke Texture Classification

Petroleum coke is used for the fabrication of anodes for aluminum reduction cells. There are ever-increasing economic and supply availability pressures for alternative and multiple sources and qualities for fabricating anodes, and therefore for cost effective characterization to make timely adjustments of the anode mixes. One characteristic historically acknowledged as having a significant detrimental influence on anode thermo-mechanical properties is the so-called isotropic structure. Traditionally such characterizations using optical microscopy are exhaustive and therefore cause a throughput concern. A method was developed and successfully applied to green and calcined coke reducing the interpretation time of a sample down to about twenty minutes. The sample preparation and interpretation methodology is described and examples presented, including a case study of how the method was used to guide raw material blend decisions across Rio Tinto Alcan's North American sites.

For metallurgical coke, a texture classification has been accepted by ASTM [5]. For petroleum cokes, however, several systems are used, among which the optical texture index (OTI, Table 1 and Eq. 1) has been used [6-8]. The on was found to correlate with the coke CTE [7]. However, quantification of the ten texture types required for the on corresponds to an interpretation time of several hours, which is not practical in Rio Tinto Alcan's context.

1 2 3 4

Introduction

5

The aluminum industry faces a shortage of anode-grade coke. However, just one quarter of the green coke available is presently considered anode-grade [1]. Some cokes of the remaining three quarters are not used since their texture is believed to be unsuitable for anodes. An important concern with isotropic and near-isotropic (granular) cokes is their high coefficient of thermal expansion (CTE) which makes anodes more susceptible towards thermal shock cracking [2]. However, in the present short coke market, calciners have started to use highly isotropic cokes in their blends [3]. Furthermore, successful tests with shot coke, also very isotropic, in medium-amperage pots were reported [4]. Apparently, certain quantities of isotropic cokes can be tolerated in anodes. This may open the opportunity to enlarge the coke supply base.

6 7 8 9 10

Table 1 Coke textures used to calculate the optical texture index (On) of petroleum coke [6] Texture Size on of texture Isotropic No optical activity 0 Fine-grained mosaics 1 10 /lm width Eq. (1) i

Where t; is the fraction of a texture and oni its texture index. An alternative could be an automated system [9,10]. However, based upon experience with image analysis systems it was felt that set-up of a robust automated method would require considerable effort and would not necessarily yield a significant time advantage. Furthermore, some cokes recently evaluated by Rio Tinto Alcan had "non-standard" textures, such as intermediates of sponge and shot coke textures, which might not be recognized by a fully automated system.

Another development is that the coke quality, including shipments from long time suppliers, fluctuates much more than in previous years. For example, between 0 and 60% isotropic cokes are now observed in samples from a Rio Tinto Alcan green coke +30 years supplier. In order to improve their economics, many refineries are buying an increasing share of their feed on the spot market. The varying feed quality translates to unstable coke quality. It is therefore necessary that the coke quality is closely followed by the coke user.

While on has been used, in general a wide range of methodologies have been used [11]. There is neither a recognized standard nomenclature nor a quantification technique applicable to the interpretation of petroleum coke texture; the methodology applied appears to be at the discretion of the laboratory as a function of analytical needs. In that respect the needs of the Rio Tinto Alcan organization are as follows:

The above discussion illustrates that a reliable characterization method for the coke texture is required. In an industrial setting, laboratory throughput is a major concern. Ideally, the interpretation time (i. e. sample observation and interpretation, excluding sample preparation) should be as fast as possible. A target of less than half an hour is preferred.

1.

1069

To screen samples of in-coming lots and of new candidate sources for the presence of undesirable

2. 3.

4.

texture, and microstructure, e.g. highly porous such as in Figure 1. To discern trends that could significantly affect anode performance. To provide such information in a timely fashion for decision making by management to take advantage of spot market conditions. To minimize the burden on available laboratory resources.

Examples of different textures of the simplified Rio Tinto classification system. Micrographs recorded with polarized I

Figure I (non-polarized light) showing calcined coke grain, termed Alcan. Rio Tinto Texture Classification System Considering the observation time in the order of six hours required to determine the OTT this technique was ruled out for routine analysis. Thus, instead of the ten texture types listed in Table 1, just three textures types are quantified in routine analysis by Rio Tinto Alcan. This system can be considered as a simplified on system since its texture types comprise several on texture types (Table 2). By the simplification some information is lost, especially with respect to anisotropic textures. Nevertheless, such a simplification still allows to sufficiently predict the anode CTE [12]. Finally, note that the nomenclature for the textures used by Rio Tinto Alcan differs from that used by other groups [12], due to the laboratory's characterization of other material types. For instance an amorphous material, e.g. Figure 2, has no optical activity, and is called isotropic instead of fine granular coke. Table 2 Coke textures used in simplified Rio Tinto Alcan classification system Texture Corresponding Size Figure textures in on system No optical Amorphous I 2 activity Length width, Isotropic 2-3 3 width, Anisotropic 4-10 4 >5 microns

Figure 4 Fine particles of calcined coke illustrating pronounced textures. taken at 25 X.

1070

Magnification Selection

It took several iterations before arriving at the characterization method presently used. Key considerations contributing to an analysis time of less than 30 minutes per sample are as follows:

• • • •

To discern if a coke grain is optically inactive, i.e. no features resolvable by optical microscope with polarized light, magnifications up to about 1000 X and higher if possible, with oil immersion lenses, are advised to see if there are features as small as about 0.5 microns. The Rio Tinto Alcan Arvida R&D Centre is not equipped with oil immersion lens capability. Even if it was, the use of such lenses was considered as a throughput hindrance considering the number of fields of view used for obtaining a reasonable portrait of texture in a sample, placing oil droplets, and delicately applying fine focus due to the short working distance between lens and specimen surface. Even with air lenses, high magnifications require more care when fine focusing. More care in focusing equals more time.

Establishing a visual limit for unacceptable texture Magnification selection An optimum particle size of the coke grains A visual assessment method

These items are briefly discussed in the next sections. Texture Type Limit As previously mentioned, it is well appreciated and widely accepted that the stronger the anisotropy of coke the better it is for anode performance.

As a further consideration for choosing a low magnification, the interest is to have a sense of what the 'forest' looks like and not to seek detail of the 'trees'. Coke grains can often have a mixture of textures; an example is given in Figure 6. As a result, by and large, the evaluations are done at 25 or 50 X with, on occasion, if needed, checks at higher magnification typically not exceeding 200 X. The working distance with those magnifications is such that the entire sample can be evaluated with little to no refocusing. However, to facilitate this process it is beneficial if the coke grains are within a specific window of dimension.

When viewed using a polarized light microscope, isotropic materials are considered to be optically inactive, i.e. no discernible features are observable. However, to discern if a material is optically inactive, high magnifications are typically advised which, due to inherent requirements that will be discussed in the next section, can slow down the evaluation process. Our desire was therefore to use lower magnifications. In so doing resolution is reduced and anisotropic textures, such as fine grained mosaics are not readily discernible. Even so, it can be argued that such fine-grained and even medium-grained mosaics, that are near isotropic, would not necessarily be significantly better for anode performance. In fact, all OTT texture types that contribute to the isotropic texture in the Rio Tinto Alcan system have low OTT values (between 1 and 3, Table 1), corresponding to high CTE values.

Coke Grain Size To be able to work with low magnification lenses, coke grain size becomes a consideration. If the grains are too small, more time is needed to look at each grain, assess, and note texture. This can be appreciated in Figure 4. If the grains are too big, the entire grain may not be within the field of view, and/ or that more time is needed to appreciate the various texture types within the grain. Based on these considerations, Rio Tinto Alcan performs routine analysis on grains between 4.75 and 1.7 mm (- 4 to + 10 mesh). It was confirmed that the texture of these particles is in general representative for all particle sizes.

As a result, in consideration that this is a screening tool, and to avoid high magnification verifications by the microscopist, the term "isotropic" applied in Rio Tinto Alcan results includes textures that are just apparent at the magnification used, e.g. up to Type 3 in Table 1. An example of this condition is given in Figure 5.

Figure 6 coke grain with various texture types, of which about taken at a 50 X.

Figure 5 Green coke grain with a predominant texture that is apparent at the taken at 50 X.

Visual Assessment Initially an assessment technique similar to that for OTT (Table 1)

1071

was done. Each polished mount took in excess of half a day to characterize. Another method tried used a lOx 10 grid graticule in a microscope eyepiece to obtain area occupation values using various scanning schemes. For instance at 50 X magnitication the entire grid would overlay an area viewed of 1 mm x 1 mm, or each division would cover an area 100 microns by 100 microns. The area occupied by a given texture type is determined by its area of occupation within the grid for each of the 100 respective fields examined. This process was again time consuming and, in the end, it was demonstrated that the results were not necessarily signiticantly different from simply looking at the microstructure and judging by eye (Table 3). The human eye-mind connection can be a very powerful analytical tool, so that once the eye knows what to look for, the mind of a microscopist can estimate within a few percent fairly reliably, at least to the extent needed for screening. Usually it takes just a few seconds per grain to register the mind's eye interpretation.

The particle presented in Figure 7 was identified by subsequent energy-dispersive X-ray spectroscopy (EDX) analysis to be iron oxide. Probably, it is the result of contamination during coke storage or transport. The particle shown in Figure 8 is believed to be an intermediate between sponge and shot coke. The isotropic spherical features, typical for shot coke [13], are easily recognizable in the micrograph. However, by visual inspection of this coke, no characteristic shot coke beads were found. This is important since some Rio Tinto Alcan suppliers occasionally produce batches containing shot coke. Shipments from these suppliers are routinely analyzed for shot coke by visual inspection. However, as shown here, the inspection might fail to identify "near shot coke" structures. As opposed to fully automated microscopic analysis, a microscopist can catch and flag these anomalies.

Table 3 Comparison of results using a lOx 10 grid graticule and visual estimation Sample Measured concentration of isotropic textures [%] a 10 X 10 grid Visual estimate 1 53 50 2 53 50 10 to 15 3 7 a No amorphous coke found, remamder IS anIsotropIc coke The estimation process has two rules of thumb to simplify the evaluation:

• •

If a given texture comprises 2: 70% of a grain's surface, that texture is assigned to the entire grain. For instance if a grain has at least 70% of an isotropic appearance, it is counted as an isotropic grain. If a given grain does not have predominantly one texture type, i.e. both constitute < 70% of the surface area, then half is assigned as isotropic and half is assigned anisotropic. See Figure 6 for an example.

Figure 7 Example of foreign material encountered during an evaluation, determined to be iron oxide in this case. taken at 25 X.

The full and half grain counts are totaled and percentages provided relative to the total number of grains assessed. If the grain size is such that fewer than about 100 grains are present in a mount, all are assessed. If there are more than 100 grains, 100 are assessed and, if warranted, more. In this way a reasonably good portrait of the coke sample is obtained within about 20 minutes of evaluation. Both green and calcined cokes are assessed routinely - over 200 samples in 2011 by a single microscopist in the Rio Tinto Alcan Arvida laboratory. Since the introduction of this method, additional refinements have been incorporated in the reported results with respect to the semiquantification of the stronger anisotropic textures, without significant penalty to the evaluation time. The value of this information is under assessment and may be the subject of a future publication.

Figure 8 Example of an unusual texture, possibly attributable to shot coke. taken at 50 X. Sample Preparation

An additional advantage of such an overview technique involving numerous grains is the greater probability of encountering foreign material, e.g. Figure 7, and unusual textures, e.g. Figure 8.

As outlined in the previous sections, considerable effort was spent to reduce the sample observation time. The sample preparation procedure, however, did not need to be changed. Shortening the

1072

traditional optical microscopy method for their characterization. The microscope time for the evaluation takes about 20 minutes. The method provides sufficient information for screening and seeing trends with respect to the presence of undesirable amorphous and isotropic textures, while at the same time lessening the human resources burden that would be experienced in using the more traditional techniques.

References I.

2.

3. 4. 5. Figure 9 A mosaic of 80 images showing the entire surface of a polished mount of calcined coke

6.

preparation time would have resulted in loss of micrograph quality, which is critical for sample evaluation.

7.

The coke particles are mounted in epoxy resin to cure overnight. The 38 mm diameter mounts are prepared using an automated polishing system with a multiple sample holder. They are first ground using a progressively finer sequence of SiC abrasives, then an intermediate polish with diamond abrasives, and a final polish with an alumina suspension. The total polishing time is in the order of 15 to 20 minutes. Figure 9 has a mosaic of images showing the surface appearance of an entire polished mount.

10.

Industrial Application

11.

The use of this evaluation technique has allowed Rio Tinto Alcan procurement and technical management to make decisions with respect to spot market opportunities, and to help follow the consequences of those decisions with respect to anode quality.

12.

8.

9.

The simplified method described here was used to quantify the concentration of amorphous and isotropic coke in all of its green and calcined coke sources. Most of them were found to contain only few amorphous and isotropic textures. However, some coke sources contained significant concentrations of these textures. Furthermore, important fluctuations were observed [3]. Shipments from certain suppliers are therefore now closely followed by microscopy analysis.

13.

It is known for laboratory anode tests and plant experience what concentrations of amorphous and isotropic coke are still acceptable. If necessary, coke allocation is modified ensuring that the concentration of these textures stays below critical values.

Concluding Remarks As a response to the market pressures for using alternative and multiple sources and qualities of green and calcined coke, Rio Tinto Alcan developed and is successfully applying a non-

1073

Stewart, M. "Petcoke Industry Overview: A Growing Market in a Shrinking World", 11th Annual Petcoke Conference, Orlando, Fl, USA, 2012. J. P. Schneider, B. Coste., "Thermal Shock of Anodes: Influence of Raw Materials and Manufacturing Parameters", Light Metals, 1993,611-619. L. C. Edwards et aI., "Evolution of Anode Grade Coke Quality", Light Metals, 2012, 1207-1212. L. C. Edwards et aI., "Use of Shot Coke as an Anode Raw Material", Light Metals, 2009, 985-990. "ASTMD5061, Standard Test Method for Microscopical Determination of the Textural Components of Metallurgical Coke", Document of ASTM International, West Conshohocken, PA, USA, 2007. A. Oya, Z. Qian, H. Marsh., "Structural study of cokes using optical microscopy and X-ray diffraction", Fuel, 1983,62,274-278. M. Hole, A. 0ye, T. Foosmes., "Relationship Between Thermal Expansion and Optical Texture of Petrol Coke", Light Metals, 1991, 575-579. R. 1. Tosta, E. M. Inzunza., "Structural Evaluation of Coke of Petroleum and Coal Tar Pitch for the Elaboration of Anodes in the Industry of the Aluminum", Light Metals, 2008, 887-892. 1. L. Eilertsen et aI., "An automatic image analysis of coke texture", Carbon, 1996,34,3: 375-385. D. Bellot et aI., "Automatic Measurement of Coke Texture by Image Analysis", Light Metals, 1992, 659-663. Escall6n, M. M., "Petroleum and Petroleum/ Coal Blends as Feedstock in Laboratory-Scale and Pilot-Scale Cokers to Obtain Carbons of Potentially High Value", Ph.D. Thesis, State Collage, Pennsylvania State University, State College, PA, USA, 2008. K. Neyrey et aI., "A Tool for Predicting Anode Performance of Non-Traditional Calcined Cokes", Light Metals, 2005, 607-612. H. Marsh, C. Calvert, 1. Bacha., "Structure and formation of shot coke - a microscopy study", Journal of Materials Science, 1985, 20, 289-302.

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

Improvements of Vibrated Bulk Density Analysis at VM-CBA and Petrocoque S.A Jean Carlos Pardo l , Edinaldo Pereira da Silva2 , Paulo da Silva Pontes l , Andre Nantes2 1 VM-CBA; Rua Moraes do Rego, 347; Aluminio-SP, CEP, Brazil 2 Petro co que S.A Ind. e Com.; Rodovia Conego Oomenico Rangoni, km 267,5; Cubatao-SP, CEP, Brazil Keywords: VBO; Coke Analysis; Calcined Petroleum Coke Lossius, Spencer, and 0ye summarized aspects that might impact agreement between labs [2]:

Abstract Anodes for aluminum production are composed of coal tar pitch (CTP) and calcined petroleum coke (CPC). The anode composition depends on the reduction technology. For S0derberg anodes, as used by VM-CBA, the coke and pitch contents range from 67 to 79% and from 33 to 21 %, respectively. Coke quality control includes sampling and analysis of chemical and physical properties. These tasks are associated with uncertainties, which may lead to wrong decisions. Since 2010, VM-CBA and Petrocoque S.A studied possibilities to reduce the standard deviation (STO) of the vibrated bulk density (VBO) analysis. The result of this work was a reduction of the difference between the VBO results of the two labs from 0.018 to 0.006 g/cm 3 .

• • • • •

The authors of reference [3] concluded that the type of crusher and the gap setting used can result in variations of the VOB. Variations can even arise when the same crusher is used to prepare different coke types. Cannova, Canada, and Vitchus [4] studied the influence of the crushing steps, particle size, and particle morphology. The authors monitored how the coke granulometry changes during ship unloading and measured the impact on the VBO. The influence of particle segregation on the determined VBO is presented in Figure 2. The data clearly show that segregation has a signiticant impact on the measured VBO.

Introduction VBO determination is described in standards procedures elaborated by the American Society for Testing and Materials (ASTM), the International Organization for Standardization (ISO), and specialized research centers. These organizations defined at least three different methods for VBO determination: ASTM 04292-10, ASTM 07454, and ISO 10236. The main differences among these procedures are the coke granulometry and the apparatus used.

!1.9211 0.910 0.900

.;} :'t 0,$00 !lci1 O.S711

VM-CBA and Petrocoque have been using the ASTM 07454 method. The VBO results of the two laboratories did not always agree. This issue has been addressed over the years through cross checks within suppliers, costumers, and specialized labs.

Q.poo U.800

f· """""'"""""'"""""'"""""'"""""'"""""'"""""'"""""'"",,,,"',,,,,,,,"',,,,,,,,"'"""""'"""""'"""""'",,"'::::,,,,,,,,"'.

a

W

211

•

•

00

00

7ll

00

00

>l:Ik-+ - 400

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

200

Sample No.

Figure 3 Sulfur content of the GPC samples

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Sample No.

The concentrations of metal impurities that have a negative impact on the anode quality fluctuated considerably between the samples (Fig. 4). The average concentrations and range for the different impurities are listed in Table 1.

Figure 6 Vanadium concentration in GPC and the corresponding CPC The behavior of sulfur during calcination is quite different than for the impurities discussed before. As shown in Fig.7, the sulfur content in CPC was lower as compared to the corresponding GPc. Apparently, during calcination a portion of the sulfur reacted and formed volatile products. In general, desulfurization increased with increasing GPC sulfur content. For some high-sulfur GPCs (e.g. cokes #10, #11, #15, #19, and #23), the sulfur content was reduced by more than 30%. This confirms that a significant portion of the GPC sulfur was transformed into volatiles and that partly desulfurization can be achieved by calcination.

1000 900 800 700 m 600 500 400 300 0. 200 100 0

8

0. 0.

.,

"'"

1

2

3

4

5

6

7

8

9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Sample No.

Figure 4 Concentrations of some metal impurities in the GPC samples

8.0

?f.

i

Table 1 Average concentrations and range for the different metal in the GPC samples (ppm) Impurity Average Range 149 Ca 6 61 Na 10 Ni 218 137 V 347 18 -

o

impurities

'J).

7." 6,0

5.0 -',0 3.0 2.0

I."

0.0 1

2

3

4

5

6

7

S

9

to 11

12 13 14 15 Hi 17 18 19 20 21 22 23 24

Sample No.

540 209 308 916

Figure 7 Sulfur concentration in GPC and the corresponding CPC

1080

Most of the sulfur present in coke is present in organic environment, such as in thiophenes, side chains of aromatic structures, or in cycloparaftins. Some sulfur is chemically adsorbed on the coke surface and very small quantities might exist in inorganic environment such as sulfates or pyrites.

14

·s

10

""

250 - 200 150

8

6

§

A portion of the GPC sulfur reacts during calcination and is mostly emitted as S02' Desulfurization depends on the initial sulfur content, the nature of the sulfur compounds, and the calcination temperature.

a

400

- 350 __________________________---=- 300 C.

12

4

.S

100 " - 50 Ii':

o

0 5

6

7

10

13

Sample No.

15

19

23

""":IIIItr-C02 Reactivity ......... Na+Ca in CPC,ppm

The higher the sulfur content the more sulfur is removed. An increase of the calcination temperature has the same effect since more GPC sulfur groups are decomposed at higher temperatures.

Fig. to CO 2 reactivities and combined Na & Ca concentration for high-S CPC samples 10 CPC samples with medium combined Na & Ca concentrations (200-400ppm) were selected to study the correlation between the CO 2 reactivity and the sulfur concentration. Fig. 11 shows a clear inverse relationship, i.e. the high-sulfur CPCs had a low CO 2 reactivity and vise versa, which means that sulfur strongly inhibits the CPC CO2 reactivity.

3. CPC CO 2 reactivity Fig. 8 reveals the correlation between the sulfur content and the CO 2 reactivity of the 24 CPC samples. Tn general, CPC samples with high sulfur concentrations had low CO 2 reactivities. 60 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

f ::3"

S I...i

f

4 3 I.-

2"

2

.S CF;.

10 I

2

3

4

5

6

7

S

9

9

10 II 12 1314 1516 17 IS 19 20 21 222324

Samples No. 1 ___

C02 Reactivity - . - S in ere

45 40 35 30 25 20 15

4.5 4.0 3.5 3.0

0

i

2.5

2.0 1.5 1.0

10

5 0

.«'" " .:; .:; « " V1v N rl

'"

'"

L

M",,1Il

Fig. 9: Green Anode Production of Line 1 & 2

Il'Ia,inellio

-----------------------,

Fig. 9 shows the anode quantities produced by the paste lines I & 2 over a period of 2 years. The production is in line with planned requirement with spare capacity still available. Fig. 10 shows the rejected anode percentage for both lines during this period, well within the acceptable limits.

ft

tUl

Conclusion

6.0

40

It was the first time for the industry, that a 100 t/h green anode

20 o.n

.......

;::: ,... ....... .b i5 t J::

...

:li

"

;::1 ;::1 ,}..

'"

::E

,.., ,... n rl

M

.:;

"ermeabHity (nPm)

Figure 2. New line baked anode density (BAD) and permeability after vibrocompactor retrofitting.

Figure 3. Correlation between vibrocompactor speed and coverweight displacement. Correlat:ionGAD vs Cover weight displacement

Vibrocompactor performance IS generally affected by the following parameters: • Compaction time, • Motor speed, • Forming temperature • Unbalanced weight setting (eccentric angle and eccentric weight), • Specific load of the cover weight, and • Vacuum and top pressure (if installed). Since our vibrocompactor is not equipped with vacuum or top pressure, the focus of our optimization plan was on compaction time, motor speed and unbalanced weight settings.

.

L638 1536

LM4 L632

-i, LMO

i(·.O.8 ,/0'

/f

"" ,..","",,"""

.2Jl L618

,.>
"

w

lim

HiO

Oi.plo••"",nt 1m",)

Figure 4. Correlation between cover-weight displacement and green anode density (GAD).

Optimization of Motor Speed and Eccentric Angle using Vibration Analyzer: Optimizing a vibrocompactor by monitoring anode density changes is not easy. The quality of the incoming paste can affect the density as much as the vibrocompactor parameters. It has been demonstrated that the usage of a vibration analysis system is a good tool for optimizing a vibrocompactor [2]. We found that the usage of a simple vibration analyzer used in machinery industries could give interesting results and help in optimizing anode compaction. The equipment comprises an acceleration transducer and a digital signal converter that measures the acceleration, velocity and displacement. During our trials, it was observed that there is a good correlation between the cover-weight displacements, the vibrocompactor motor speed (rpm) and anode density (see Figure 3 and Figure 4). This correlation was the main tool for our optimization process. Using the same equipment and principle, the optimum vibrocompactor speed (1,300 rpm to 1,050 rpm) and the best unbalanced weight setting (from 140' to 100') were set up. Vibration analyses are now being performed routinely and used as tools to monitor the performance of both the vibrocompactors.

As a result of this optimization, the density of new line anodes has improved by 0.01 g/cm 3 (see Figure 5).

1.585

,

1.SS0 ,,""""""""'" 1IIIk"""""""""""""""""""""""""""""""""""""""""""""

1575 G:' IE 1570

""""""""""""""""""""""">"

"""

i"""""""""""""""""""""""""", """"""",""

iL565

i

1.560 M 155$ 1.5:50 1.545

VI8

i

;. 1,'16 1.'14 :. 1.42 ,.s

f

lAO 1.33 1.36 1.341

t

§

132

:li!

LID

Figure 5. Impact of new line compactor optimization on anode density and permeability Improvement of Paste Mixing in New line Even after the optimization of the vibrocompactor, the permeability of the anodes from new line remained approximately double that of old line anodes. Petrographic analysis of paste, conducted in coordination with external consultants, revealed an inconsistent coating of pitch around coke grains in the paste samples. The study suggested insufficient or inconsistent mixing. With this information, our focus shifted to the mixing stage to find the probable cause of high and irregular air permeability.

1106

To avoid paste sticking in the mixer discharge chute, water sprayers were installed at the exit of the mixer. It was suspected that the pitch in the paste may be "freezing" after direct contact with water spray and creating a "heterogeneous" paste. To verify this assumption, trials were conducted and it was confirmed that suppressing the water injection at the mixer discharge significantly improved anode air permeability. A new discharge chute was designed, with wider cross-section and better alignment with the mixer discharge, so as to be able to run without the water spray. Thereafter, the permeability of new line anodes has significantly improved to a level comparable to old line (see Figure 6). Im:pact. p.uilte chute modification on anode Air permeability of new line

Figure 7. Timeline ofDUBAL coke blending implementation. Pilots Test Results:

lII!

Pilot tests were conducted in 2009, separately by external technical laboratories. Four different sources of coke were tested: two regular supply and two new sources. As an example, the figures below illustrate the impact of the new coke source B when blended with the regular coke A (see Figure 8 & Figure 9). Optimum baked density with maximum reactivity residue was achieved when coke B was used with coke A in the range of 33% to 67%.

Average a:r permeabiHty

Figure 6. Improvement of anode air permeability by modification of kneader discharge chute. Coke Blending

Blending Pilot tests: Density 1.67

Historically, DUBAL sourced its calcined petroleum coke from four different suppliers because of the following reasons: security of supply, cost competitiveness and supply logistics. As the dock facility of the Carbon Plant was not designed for coke blending, the cokes were used in the plant on campaign basis. When the coke supply changed, adjustments were made in operating parameters to optimize the anode quality. In a drive to compensate for the degrading anode quality of traditional cokes and improve anode quality and performance, DUBAL developed a strategy for coke blending with the aim of increasing baked anode density, with minimal impact on anode reactivity.

1.57 156

1.66

US

1.55

$1.64

J:; 11,f.i3

.fi

................................................................................

1.62

!1.61 15 1.59

1;

······························;L54$

+ ................................................................. +

J:;

1.53

1

LSI

i

+ L52 1.5

... '"

1,49

Figure 8. Density of bench scale anodes.

Implementation Plan. To achieve this objective, an implementation plan was developed with the following key steps: • Identification of potential suppliers, • Pilot studies. • Plant trials, • Installation a coke blending systems, and • Full scale implementation and process optimization. The implementation timeline is shown on Figure 7.

Blending Pilot tests: eRR lind ARR 97,0

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . , 112.0

96.0

,,,...............................

+

95.0

194,0 }

5

........................;/'., ................................................. ,.""",

........ ,gs

911) 90,0

sa.o

S6Jl

. L 1$4.0 il1.0 ,00.0

93 .0

920

9(H)

l

I

. H!s) 71W

c····················································· .............................................................................•

74.0

Figure 9. CO 2 reactivity residue (CRR) and Air reactivity residue (ARR) of bench scale anodes.

1107

Plant Trial Results:

Impact of coke blending on Baked anode Density

To confirm the results of the bench scale anodes, trial consignments of new coke sources were ordered. Trial anodes were produced and their performances in the pots were evaluated. Figure 10 and Figure II show the similarities between the bench scale and plant anodes. Optimum baked anode density was also achieved with maximum reactivity residue when 50% of coke B was blended with 50% of coke A. With the contidence of the bench scale and plant evaluation, the full implementation of coke blending was decided and blending ratios were detined.

l,sas

Plan BlendingTfials: Density 1.0:)0

MOO

:

13

1$20 ,."""""".......................................................................................................................... ,..................................................................................................................................

BitS

Ibl0 c..................................................................................................................................

Figure 12. BAD single versus blended coke.

¢j

5

1.500

t

, 1$70

1

Strategic Alignment with Supply Chain After two years of coke blending operation and having used six sources of coke, each presenting very distinct properties (density, impurity content, reactivity and calcination level), it became crucial for DUBAL to try to answer the following question: • How to evaluate the overall impact of a blend? • Which cokes are best to be blended together and at what ratio? • How to optimize coke blending based on the properties of a particular shipment or source? • Are we using the best coke sources for blending?

i 1.5&)

c,

c 1.550

100% Ceke 2$% Ceke 40% Coke SSO% C"k" II 15k Cck" A 8 II

Figure 10. Density of Plant trial anodes. Plan Blending Trials ;CRR acnd ARR 96,0

94.0

i·····.•

. · · · · · · · · · · · · · ·. . . . . . . . . . . . . . . . :

lILO

&1.0 7iH)

IS

76,0

9W ,

..............................................................................................................

'c:':k

Extensive data analysis has provided some answers, giving DUBAL the opportunity to develop an empirical model to predict anode quality from coke properties. The key correlations established by the study were: • The correlation between coke vibrated bulk density (YBD) and baked anode density (BAD) (see Figure 13). • The impact of coke CO 2 reactivity on anode CO 2 reactivity residue (see Figure 14).

.

74,0

,

no

00.0 . . ............................................................................................................................................. 70})

ARR

1.61 c····················································· ..................................................................................................................

Figure II. Reactivity of Plant trial anodes. Full-scale Implementation Results A Coke Blending facility was installed and commissioned at the doclc All the coke silos have the option to deliver coke as per desired targets, calculated on the basis of blending ratio. During the course of travel to the discharge silo, cokes from ditferent silos are mixed and then sent to Green Mill for anode production. After six months of operation, we re-assessed the impact of coke blending. The data obtained confirmed the results expected from the trials, with an overall increase of density of 0.0 1 g/cm 3 and minimal variation in anode reactivity (see Figure 12)

() ..9VV

omo

c.>k. veo 16/«)

Figure 13. Correlation between coke YBD and baked anode density.

1108

Anode Recipe Modification

Coke CO 2 read. vs Anode CR Residue

1 1«

DUBAL's dry aggregate comprises Coarse, Medium and Fines Fraction. Coarse Fraction is composed of butt material and the other two fractions are primarily made up of calcined petroleum coke. Grain size distribution (or dry aggregate curve) is a critical parameter for anode quality. While using different coke sources in blending, it was observed that the higher density cokes in the blend also had a higher percentage of bigger particles in the dry aggregate, which had a positive impact on anode density. These observations prompted the idea of increasing the percentage of coke with bigger particles (+ 4 mesh) in the dry aggregate, so as to improve on anode density. Several trials were conducted to evaluate the impact of increasing the amount of +4 mesh particles in the recipe and to determine the optimum percentage. Positive results from the trials led to the replacement of the screen deck to increase the percentage of +4 mesh in the dry aggregate from 22.5% to 25.5% as shown in Figure 17.

SO)

930

5 91.0

..f.\••'••••••••••••••••••••••••••

+.................... +

.............. .

.....................................................................................................

i 1)9.0 if 871l 2.,0

KO

8.0

mo

12.0

Cok. 002 R""ctivity 1%1

Figure 14. Correlation between CO 2 of coke reactivity and anode CO 2 reactivity residue. On the other hand, good correlation was also established between: • Baked anode density and anode air permeability (see Figure 15). • Vanadium in anode and anode air reactivity residue (see Figure 16).

2$

Anode Air Permeability vs baked density

lt7 26

x*:tW,$

f 23

:n 21

{U

2tt 157

'* Aif

1$$

159

1..6

HH

VjJ

MOW«)

Figure 17. + 4 mesh in dry aggregate.

Figure 15. Correlation between BAD and air permeability.

Immediately after the change, baked anode densities improved by 0.01 g/cm3 . The other part of anode density increased being attributed to increased VBD of the coke blend. The increase in density also resulted in an improvement of anode air permeability from 0.8 to 0.5nPerm (see Figure 18).

ARR. \/s vanadium in dubll!;! anodes 100 $IS

...::i

911

4

liS 1111

·t I

75

;

Recipe change: BAD and A!!' 'Perm.

711 ..S liill 55 SII

'"

..

Figure 16. Correlation between vanadium content in anode and anode air reactivity residue. Being able to predict four key anode characteristics from coke properties gives a good tool to estimate the variation in carbon consumption resulting from the usage of different coke blends. They are also being used to develop the procurement plan and shipping schedules to minimize the risks to DUBAL.

Figure 18. Baked anode density and air permeability before and after recipe change.

1109

Results and Discussion

Conclusion

The overall results of the optimization over a three- year period are illustrated in Figure 19 and Figure 20: • An increase of anodes density by 0.025g/cm 3 , which is equivalent to 20 kg additional weight per anode. • A reduction of anode air permeability from 1.2 to 0.48 nPerm. The usage of cokes with low sulphur, and high vanadium in the blend has marginally affected anode reactivity. Air reactivity residue (ARR) has reduced by 2% while CO 2 reactivity residue (CRR) has remained practically unchanged due to the blending of high and low sulphur cokes.

DUBAL has achieved a tremendous improvement in baked anode density and anode air permeability through a systematic approach based on process optimization and selective blending of cokes. The overall baked density has increased by 0.025 g/cm 3 along with a reduction of anode permeability by 0.7 nPerm. This has helped the potroom increase the average current by 20 kA and also reduced the number of purchased anodes. This achievement is the result ofteam work, commitment towards continual improvement and encouragement for innovation and creativity.

References

Core Density (g/ec)

u.w

l.

MOO 1$90 1SOO

2.

U10

UOO 1009

3.

Figure 19. Baked anode density improvement.

vt

Air Perm (nPm) loll

1.2 1.0

IE

%!

0$ OiG OA 0.2 0.0

2m

lQlQ

lOU

Figure 20. Air permeability improvement.

Anode Performance in Cells. The improvement in baked anode density and reduction in air permeability helped DUBAL increase the amperage to the desired level, along with a positive impact on carbon consumption. During the implementation of this anode density improvement program, DUBAL Carbon team worked closely with Reduction management to improve effective utilization of carbon in potroom by optimizing the butt thickness for individual technologies without affecting pot performance. In spite of an average amperage increase of 20 kA, these combined actions contributed to 4% reduction in gross carbon consumption across the plant compare to 2009. Saving in gross carbon also resulted in reducing the number of purchased anodes.

1110

Les C. Edwards, "Evolution of Anode Grade Coke Quality and Calcining Technology", Proceedings of lOth Australasian Aluminium Smelting Conference, Australia 9th to 14th October 2011, Editors Barry Welch et al. .T. Higley et ai, "Maximizing Vibroformer Performance Through Vibration monitoring," Proceedings of 10th Australasian Aluminium Smelting Conference, Australia 9th to 14th October 2011, Editors Barry Welch et al. Khalil Khaji and Hameed Abbas, "Baked Anode Density Improvement through Optimization of Green Anode Dry Aggregate Composition", Light Metals 2010, 10271030.

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

DEVELOPMENT OF AN ANALYTICAL DYNAMIC MODEL OF A VIBRO-COMPACTOR USED IN CARBON ANODE PRODUCTION Fatma RebaYne l, Mohamed Bouazara l, Daniel Marceau l, Duygu Kocaefe l, Brigitte Morais 2 lCentre universitaire de recherche sur l'aluminium (CURAL), Universite du Quebec a Chicoutimi, 555, Boul. de I'Universite, Chicoutimi, Quebec, Canada, G7H 2Bl 2Aluminerie Alouette inc., 400, Pointe-Noire Road, C.P. 1650, Sept-iles, Quebec, Canada, G4R 5M9 Keywords: vibro-compactor, dynamic, anode, frequency frequency thus allowing the system to vibrate. Figure 1 shows a schematic of the key components ofvibro-compactor.

Abstract The carbon anode quality has a signiticant impact on the production of primary aluminum. Their performance can be evaluated by their various mechanical, electrical, physical, and chemical, properties such as density, electric resistivity, CO 2 and air reactivities. The focus of this work is to study the various parameters of the vibro-compaction, which is one of the critical steps in the process of anode manufacturing. In this work, a dynamic model of a vibro-compactor is developed. The vibrocompactor is modeled as a rigid mass suspended on springs and dampers and subjected to harmonic external excitation. This model is used to identify the optimal conditions of the vibrocompacting process. These conditions are obtained through a correlation between the analytical vibro-compaction parameters and data from an industrial vibro-compactor. The use of optimum parameters will help improve the anode performance and, consequently, lead to better productivity and reduction on environmental impact.

Gnidingrod

Coverweig.ll

VibrMing table

•... .. -..., Eccell!rtc weights

Introduction

Vibration isolillors

The aluminum industries in Quebec consumes alone about 1.27 million tons per year of carbon for anode production. The quality of anodes is widely influenced by the quality of raw materials and the manufacturing process. Before 1980, the hydraulic presses were used for the fabrication of anodes but now most of the presses have been replaced by the vibro-compactors due to their greater efficiency [1, 2]. However, the control of the various parameters of vibro-compaction is quite relevant because the vibro-compactors require much more efficient control and maintenance. During the manufacturing process, the raw materials coke, recycled anodes, butts) are crushed, mixed with pitch and used for forming the paste. After that, they are placed in a vibrocompactor for shaping followed by baking and cooling. The efficiency of each of the manufacturing steps is important to improve the quality and mechanical properties of the anodes. Vibro-compaction is one of the most critical steps during the fabrication of anodes. If anodes are not well vibrated, their mechanical strength decreases and in turn this causes the premature crack formation. Proper vibration can be achieved by using the optimum values of time, frequency level. load and ensuring a uniform vibration inside the mould.

Figure 1: Schematic drawing of the vibro-compactor. Some vibro-compactors also include a vacuum, to have a better efficiency. With the use of vacuum, the apparent densities of green and baked anodes increase by over 0.02 kg/dm3 and 0.015 kg/dm3, respectively [3]. At the beginning of the vibro-compaction process, the mould is attached to the vibrating table and all the vibrations are propagated under the follower weight which is lowered freely on the paste. When the desired degree of densification is reached with a required vibrating time, the mould is separated from the vibratory assembly and the finished block of anode is pushed laterally outside the mould. Vibrating table The vibrating table is one of the most important vibro-compactor components. The vibration isolator for the table has changed in the course of time. During the initial stages of the development of the vibrating table, the vibration isolator was made of helical steel. Later, it was replaced by a solid rubber block. Now, the compressed air intlatable rubber is used as a vibration isolator [4] [5]. The vibration of the vibrating table is largely influenced by the calibration of the angle of the eccentric counterweights. Figure 2 shows the different vibration isolators.

Vibro-compactor The vibro-compactor consists mainly of a vibrating table supported on shock absorbers (vibration isolators), a mould where the anode paste is poured and a follower weight having a loading system which helps to compact the anode paste. Below the vibrating table, there is a rotating motor with eccentric weights that can create a rotation unbalance with a certain excitation

1111

tet) '" to sin (wt) Helical steel

Solid rubber block

M; y

Co:trlpres!sed air inflatable .rubber

Figure 2: Evolution of different vibration isolators' with time. Eccentric counterweights The vibro-compactor also consists of four eccentric counterweights located on the two shafts. Their balances can be changed by changing the angles of unbalance and used to increase or decrease the magnitude of the vibrating load. The two shafts must be in the same orientation to ensure that all the forces act in the vertical direction and the vibration is balanced. The figure 3 shows these eccentric counterweights.

Figure 4: I-D dynamic model ofa vibro-compactor. The total mass of a vibro-compactor can be represented by the equation (I). It represents the sum of the masses of vibrating table, anode, mould and total weight of the cover. Equations (2) and (3) present the external excitation of a vibratory compactor. The external force can be represented by a sinusoidal force acting on the whole system. (I)

where; M r : Total mass ml :

Vibrating table, mould and anodes masses equal to 3650 kg.

m2 :

Cover weigh mass equal to 6270 kg.

Figure 3: Eccentric counterweights drawing.

f(t)

There must be angles of unbalance at each corner of the vibrating table, sufficient to have reasonably a uniform vibration of the anode paste. A non-uniform distribution of angles of unbalance may cause the formation of non-homogeneous anodes.

=

fa sin (wt)

(2)

kT·x

(3)

and

fa

=

with: f(t): External excitation force (N).

I-D dynamic model

fa : Force amplitude (N).

w : Frequency (RPM). t: Vibrating time (s). kT : Total rubber support rigidity (N/m).

A vibratory machine. which is mounted on elastic supports always, has a characteristic resonance frequency which depends on the weight of the vibratory machine and the stiffness of the supports. If the machine with a rigid support weighs less, the resonance frequency will be high. Heavy machinery on highly elastic supports will have a low resonance frequency [6]. The second case is the condition used in this study.

bT : Total rubber support damping (N s/m).

The numerical values of the parameters of the dynamic model are shown in a Table T. These values were obtained based on data from the article ofM. Beilstein et al [I].

Figure 4 shows a l-D dynamic model of a vibro-compactor developed in order to study the motion of the vibro-compactor and the vibration parameters. The vibro-compactor is modeled as a rigid mass Mr suspended on springs kr and dampers bT and subjected to harmonic external excitationf(t).

Table 1: Dynamic model parameters. Parameters Total mass

1112

I

Values 9920 kg

Rubber support rigidity

18.2* 106 Nlm

Rubber support damping

5000N slm

Amplitude

5mm

Vibrating time

45 s

Frequency

1300 RPM

Analytical model

Numerical model

The equation of motion is solved using the following two methods for the model shown in figure 4. First, an analytical approach, which corresponds to a direct solution of the motion equation, is used. The second approach is based on the use of the Simulink interface of Matlab software. In this section, the analytical model is presented.

It is necessary to use numerical computation techniques to obtain an approximate solution of the equations of motion. In this section a second approach is presented based on the use of Simulink interface. This technique requires the construction of a schematic block representing the equations of motion. The figure 6 illustrates the dynamic model used in Simulink. The driving force of the system was the input as shown in figure 7 on Simulink interface.

The equation of motion can be written as follows:

(4) This is a non-homogeneous linear differential equation of second order with constant coefficients. The general solution of this equation can be obtained by the sum of the complementary solution of the homogeneous equation and the particular solution of the equation with the non-homogeneous term (forced response). If:

o Ext'i!rnal excitation

.A.dd1

M T . x : Inertia force. br

Gain

...... 0ain2

kr . x : Elastic force.

Then

Clo4:

..

Integ!,0

, .......... ,

us

700

TIme (II)

Figure 10: Impact of the variable properties on the temperature history at the center of the anode block (shown by red dot).

1100 1000

900

E 000 i!! 100 i

."e

(.00

900 {It 'lOO

l!! ??O 200 100

Figure 9: The anode surface temperatures after heating for (a) 75h, (b) 100h, and (c) 120h; the temperature profiles on four vertical planes inside the anodes for cases corresponding to above heating periods of (d) 75h, (e) 100h, and (f) 120h.

0 (J

%2

{J;l

0,4

{J,$

OIS

'W

0.1

Q,S

0,9

I'IQfmali$ed Pl$\lI.I1;

Commissioning of the first fire group commenced in February and the fourth and final fire group was completed in mid April. By May, the furnace was fully operational and supplying all anodes previously supplied by the existing closed baking furnaces. Early gas consumption for the new furnace was in the range of 1. 9 to 2.1 GJ/mt of baked anodes (Figure 9). Further opportunities have been identified to reduce consumption by tuning of the zero point and adjustment of the target temperature curves.

§

25

0

Gas Consumption and Off-Gas Emissions

u

i

8 15

The introduction of a wireless network communication between other ramps in the fire group was a new development for BSL. A second wireless network is in use for the CBF4 building, for communications from the furnace cranes. Testing of the strength and reliability of the tire control system network was performed with the cranes in operation and travelling to all areas of the furnace building to simulate all expected operating scenarios.

!i2.

•

40

Week 4

Figure 9: CBF4 natural gas consumption

1167

Table I - Early CBF4 furnace and anode property data. Standard deviation shown in brackets

Parameter

Units

Typical Value

°E

1,225 (40)

Electrical resistivity

flQ·m

57 (4.5)

Baked apparent density

kg/m3

1560 (7)

CO 2 Reactivity Residue

%

94 (2)

CO in furnace off-gases

mg/Nm 3

25 -40

°c

200 - 250

GJ/mt baked anode

1.9 -2.1

Baking level

Anode temperature time of unload

at

Natural gas consumption

The early results from CBF4, along with a small scale trial ofthe anodes on the Reduction Line gave BSL the contidence to commence a rapid de-commissioning of the existing closed baking furnaces. Further, the initial results have given the plant process engineers a basis for further optimisation of the target temperature curves and burner set-up to reduce variation in temperatures within the pit and reduce the natural gas consumption.

Summary This paper describes the results that can be achieved when a commissioning team comprised of experienced start-up and operations personnel is combined with a bake furnace fire control system with advanced process control functionality. It also demonstrates that the adoption of modern Wi-Fi communications technology and more stringent furnace safety requirements is no barrier to high system availability. BSL has entered into a new era of anode baking capability, and is well prepared for the present and future challenges that will no doubt arise in the highly competitive aluminium smelting industry.

References [I] J. Ameeri; K. M. Khaji; W. K. Leisenberg, The Impact of the Firing and Control System on the Efficiency of the Baking Process, Light Metals (2003) 589-594. [2] C. P. Hughes, Methods for Determining the Degree of Baking in Anodes, Light Metals 1996, p 521

1168

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

Laser Mapping of Carbon Bake Furnaces Ashley Tews 1 , Mike Bosse 1 , Robert Zlot 1 , Paul Flick 1 , Meaghan Noonan 2 lCommonwealth Scientific and Industrial Research Organisation (CSIRO) Queensland Centre for Advanced Technology; Pullenvale, Queensland, Australia 2Pacific Aluminium, Brisbane, Queensland, Australia Keywords: automation, carbon bake, analysis

Abstract The flue walls in carbon bake furnaces (CBFs) deform over time under cyclic heating and cooling, leading to difficulties in loading/unloading anodes, and inconsistent anode baking. It is useful to regularly measure the deformations to establish the rate of deterioration and assist in the prediction of flue wall life. Boyne Smelters Limited commissioned a new CBF in 2012. During commissioning, CSIRO utilised its 3D laser scanning technology to map the CBF. A sensor payload consisting of a scanning laser rangefinder sensor connected to a logging and processing PC was suspended from a crane and moved over open pits in a pattern to obtain sufficient coverage of internal pit surfaces, as well as some of the CBF floor and walls. The resulting data is useful for comparing the finished CBF to blueprint plans, and serves as a baseline for future scan comparisons to determine deformations of the flue walls and pit floor.

Introduction A carbon baking furnace (CBF) consists of interconnected, hollow refractory flue walls through which hot gases are drawn. The heat is conducted through the brick walls to the carbon anodes, improving the anode properties for the electrolysis process. The flue walls deform under this cyclic heating and cooling, leading to difficulties in loading and unloading anodes from the pits, process issues and/or structural instability. Regular maintenance prolongs the working life of the flue walls until replacement is required. The flue wall condition is monitored to assist with maintenance planning and predicting flue life. Assessments are typically conducted manually and therefore, can be subjective. Improved assessment systems use physical observations and measurements to calculate a risk score, but time constraints still limit the frequency of assessments and the level of detail captured. The concept of using a more automated approach to obtain three-dimensional (3D) maps

of refractory surfaces in the CBF is therefore attractive. CSIRO has substantial experience with developing 2D and 3D sensing systems [1-3] that are capable of creating data representations that can be used for mapping, navigation or dimension extraction of surfaces such as walls, floors and infrastructure. These systems consist of off-the-shelf hardware configured to provide data to the software that CSIRO has developed. The software integrates the data to produce maps in the form of 3D point clouds, where the accuracy, precision, density, and coverage depend on the sensor motion and technical specifications. Typically, the sensors used are 2D scanning laser range finders that provide time-of-flight range measurements to the nearest surface. Moving the sensors around an environment allows for measurements to be gathered from all surfaces observed by the scans of the sensors. Boyne Smelters Limited (BSL) commissioned a new sixty-six-section, Rio Tinto Alcan AP technology open baking furnace in early 2012. This presented an opportunity for access to a larger number of empty pits than is possible on a fully operating furnace. CSIRO proposed the use of sensing systems to gather 3D data relating to the shape and position of floor and pit surfaces. The project involved using 3D scanners to effectively 'map' the floor, walls and pits in areas of the CBF, with particular focus on the open sections.

Data Acquisition In this section, the equipment setup and data acquisition steps are described. Equipment 3D surface mapping equipment is readily available off-theshelf but mapping can be time-consuming and/or expensive. A common approach is to use surveying equipment statically set up on a tripod at pre-determined locations around the area to be mapped. The sensor rig is then moved to different locations to ensure adequate coverage. This is a highly manual process that can produce accurate

1169

results, and requires qualified operators to use the equipment and process the data. For a CBF layout, placement of surveying equipment would be particularly challenging both due to the geometry of the environment, and the difficulty of situating equipment on floor locations between the pits. As part of improving mobile robotics applications for industry, the CSIRO team have developed hardware and software systems for 3D mapping using off-the-shelf, relatively inexpensive equipment and in-house technology for processing the data into appropriate representations. For CBF scanning, a prototype sensor rig was developed for data acquisition.

Figure 2: The schematic of the sensor rig, showing the lasers and IMU location. (Figure 2), only the rotating laser was used for mapping purposes (the others were included to increase coverage and point density, but were later deemed unnecessary). The rotating laser consists of a SICK LMS 291 lidar on a spinning platform. The SICK lidar measures range values in a single plane with a 180 0 field of view and 10 angular resolution at 75Hz. By spinning the sensor using a rotation motor mounted to its base, the scan plane is no longer fixed, and the resulting configuration results in a hemispherical field of view. Due to radial dispersion, nearby objects will have a higher density of scan points as compared to objects farther away; however, by moving the sensor around the environment, local measurement densities increase over time. The noise level of the range measurements are predominantly determined by the laser hardware (2-3cm for the SICK LMS 291). Through postprocessing of the data, better accuracies are possible by averaging over multiple returns where sufficiently high point densities are obtained.

Figure 1: The sensor rig attached to a crane in the CBF. The lidar sensors can be seen under the rig's frame with the control and logging PC, batteries and cabling above it. The rig, shown in Figures 1, 2, and 3, consists of several scanning laser rangefinder (lidar) sensors and an Inertial Measurement Unit (IMU) connected to a logging and processing PC. The rig was made to be rugged and attached to a crane's hook by standard slings. The crane can then 'trawl' the rig over the areas to be scanned, including descending into the pits if required. The design is intended to enable simple setup and independent operation for acquiring the range data without requiring complex interfaces with existing equipment. While several laser scanners were mounted on the rig

Spinning laser Controller

Figure 3: The schematic diagram of the top view of the sensor rig. In order to generate a globally consistent point cloud model, it is necessary to accurately estimate the trajec-

1170

0 E .r:.

1::: 0

c

80 60 40 20

-5 -10

o

10

20

40

30

50

60

C ·E

(])

E :;:::;

70

east [m]

Figure 4: The crane path over sections open pits (black rectangles). The path is coloured by time according to the legend on the right. The entire scanning time for the 14 sections was approximately 90 minutes which was a limitation of maximum crane speed rather than as a requirement of the mapping system. tory of the sensor payload; that is, the position and orientation of the device at all times. The trajectory is estimated using software developed at CSIRO which uses the relative changes in lidar returns to infer the motion of the payload through the environment [2]. The IMU provides additional rotational rate and translational acceleration measurements which can improve the reliability of the trajectory estimate and ensure that the direction of gravity is known. Data Collection At BSL, one of the CBF cranes was made available for data collection. The sensor rig was suspended from the crane and a series of initialisation tests undertaken prior to the main data gathering exercise. For an initial calibration run, the sensor rig was lowered into a pit. From the initialisation tests, a crane path was determined over 14 new sections with open pits and then moved opportunistically around covered pits and the CBF to get a general coverage. For most sections of open pits, the crane made three lateral passes (two on the first pass and one on a return run) to allow for adequate coverage. Note that the sensor rig was developed to allow scanning inside the pits by passing over them rather than having to lower the rig into them. The path of the crane over the open pits from one sequence can be seen in Figure 4. The actual path of the sensor rig from a data sequence collected over the pits of a single section is shown in Figure 5. The colour change on the path represents the time progression from the start (blue) to the end (red). The wobble in the path is the swing after the rig was picked up or changes direction (which is correctly detected by the trajectory estimation software). The rig's trajectory is shown over the processed laser data points seen from an overhead view.

5.

Figure 5: Time-coloured path of the sensor rig over a section during a forward pass. The black dots represent processed scan points. The wobble in the path was the rig swinging under the crane. Note that the effect of the swing has been corrected in the registered points (i.e. the underlying points form straight lines around the pits). After gathering data from the open pits, the sensor rig was also trawled over other areas of the CBF to allow for supplementary data gathering. The logged data was then processed as discussed in the following section. Data Representation Data gathered from scanning is in a raw and unregistered form, and does not constitute a consistent point cloud model until compensation has been made for the payload motion. CSIRO's technology addresses the problem of estimating the sensor trajectory and the environment map concurrently, a problem which is known in the Robotics literature as Simultaneous Localisation and Mapping (SLAM). As the sensor payload moves, features in the environment appear to shift from the scanner's point of view. However, since the environment is assumed to be predominantly static, these apparent motions can

1171

be used to correct for the scanner trajectory. The trajectory corrections are derived from solving a system of constraints including surface patch correspondences, trajectory smoothness, and IMU measurement constraints. Given the globally registered trajectory, the laser range measurements can be projected from the appropriate sensor positions and orientations to produce a consistent point cloud. The point cloud can be saved in a file format for further analysis or visualisation in 3D viewing software. Further analysis is then possible such as extraction of dimensions or surface distortion properties. The remainder of this section discusses the results from scanning the CBF from a high level overview of the entire CBF down to analysis of a pit, as an example of the potential of the mapping process. Entire CBF Figure 6 shows a topdown view of the processed 3D data points over the end of the CBF that contained the open pits. From this high level view, the pits, floor and various infrastructure are clearly identifiable. The density of the data is evident by the solidity of the colours on the furnace features. The area around the open pits is particularly dense since it was the main area to be mapped, and the crane made three passes over most of those pits compared with only a single pass over the rest of the CBF. From this representation, major dimensions of the CBF can be determined.

Pit Analysis Since the floors and walls have many scan points to represent them in the data, they may be represented and statistically analysed in a number of ways. For example, the data points on a pit wall can be examined either individually or over small averaged windows. Given the potentially large amount of data (thousands of scan points on a single pit wall), merging points into larger reference windows offers a more appropriate representation. Given that the open pits have never been used, the walls should be relatively flat and parallel within some arbitrary level of accuracy. This is not an assumption taken by the mapping system as it does not place constraints on the data representation such as flatness or squareness. Since pit shape and uniformity is of importance for data analysis, the remainder of this section provides an example analysis of an open pit. The analysis method is chosen for ease of demonstration and others may be more useful to the end user. The figures in this section represent an exploded view of the pit, i.e. the pit has effectively been 'opened' like a box. The pit floor is shown at the bottom of the figure (at approx. x = -7 to -2, y = -6). The colours in each figure show the respective values represented in the legend on their right. Each figure identifies different aspects of the information available from the processed data. Figure 7 shows the density of scan points falling on the walls and floor. Since the sensor rig passed only over the top of the pit, fewer points fall lower down the pit, evident by the blue values the further down the walls. These values represent the number of scan points falling per 'pixel' where a pixel represents a IOxlO cm cell or window. 200 100

W Q)

Co

20 10

g

5

2 -6

-4

-2

1m]

Figure 6: A partial view of the CBF showing the pits in operation at the top of the figure and the empty pits (shown in blue) in the lower right. The colours pertain to height with blues being low and ochres being high. The CBF floor is consistently green indicating its uniform flatness. Firing equipment can been seen in yellow.

o

2

Q)

-0

4

Figure 7: Density of data points falling in IOxlOcm grid cells on each surface of a pit. Each surface is shown bounded by a black border with the four walls in the upper part of the figure and the floor at the bottom. Using the same pixel resolution, an analysis of the

1172

points that fall into each cell can show both the accuracy and precision of the data. Accuracy is determined by the taking the average value of all points in each pixel and comparing their distance to a plane fitted to all data points for that surface (wall or floor). A value of 0 would indicate the average value of all points falls on the plane. In Figure 8 , the accuracy is shown for each cell. Deviations represented by the extreme colours are more prevalent in the lower walls which correlates with the lower point density. Precision can be considered as the 'spread' of the points around the average value. For example, a cell with an estimated average value of 0 (therefore, accurate) may have a large spread of points around it. The precision value quantifies this spread. Figure 9 shows the precision estimate for each pixel.

Figure 8: Accuracy analysis calculated using the processed data. Green values represent no deviation and other colours represent indents (redder values) or projections (bluer values). These deviations may be either real or virtual artifacts caused by sensor noise and a sparsity of points in those regions.

Statistical Estimations of a Section The previous analysis provides an indication of the variance and quality of the processed data, assuming the surfaces of the pit are uniformly flat. A general analysis was conducted of the processed data to determine the outer dimensions of each pit including the variance on these measurements. The measurements were based on computing the average distance between planes fitted to opposing surfaces using the processed data as a guide to plane-placement. The calculated variance is estimated 'noise' or dispersion of the points used in the calculation. The resulting estimations were compared against blueprint plan dimensions as well as physical measurements taken by Pacific Aluminium staff. Variations are differences between estimated measurements and the actual dimension and can be due to: the walls not being built exactly to plans, walls not being perfectly planar, and/or sensor noise. Tables 1, 2, and 3 below show the dimensions for length, width and height of all eight pits in a single section. For the length and width, the measurements were made between opposing surfaces. For the depth, the location of the top of the pit was estimated and depth analyses were based on this offset. Hence, the errors may include a small constant bias in one direction. If the pits of the section were made to the dimensions provided (5100mm length, 795mm width, and 5710mm depth), then the errors from our estimation of the dimensions were: length 12-30 mm, width 0-5mm width, and depth 30-50mm. Note that these were estimates taken from further processing the data and do not represent the accuracy of the data points, only of the manner in which these dimensions were calculated. Summary

-6

-4

-2

2

4

1m]

Figure 9: Precision estimate for the processed points. Note that most are within 10 mm.

Lidar-based sensing systems have the potential to provide data-rich records of refractory condition which will likely assist maintenance planning and flue life prediction. A key opportunity existed with the newly constructed CBF at BSL to gather baseline data of pit surfaces and the CBF floor. During data gathering, the entire CBF was surface scanned with 14 sections of open pits allowing for internal scanning. The time for scanning an open section using the prototype sensor rig was less than six minutes, though this was due to the crane speed rather than a data gathering requirement. The entire CBF was scanned in approximately 90 minutes. This is substantially faster and more automated than existing methods considering the sensor

1173

Table 1: Estimated pit widths and variances. Ideal actual pit width = 795mm. Pit: Mean width (mm): Stddev (mm):

1 795.4 9.47

2 794.87 9.82

3 798.03 9.57

4 794.67 11.14

5 795.5 8.4

6 796.64 9.33

Table 2: Estimated pit lengths and variances. Ideal actual pit length Pit: Mean Length (mm): Stddev (mm):

1 5130.1 15.2

2 5122.1 16.96

3 5129.5 11.89

4 5117.4 18.79

5 5115.7 13.99

6 5127.5 12.89

7 800.43 11.29

=

8 798.29 12.385

5100mm. 7 5122.7 12.58

8 5124.5 14.97

Table 3: Estimated pit depths and variances. Ideal actual pit depth: approx 5710mm. Pit: Mean Depth (mm): Stddev (mm):

1 5740.7 8.8717

2 5745 8.5198

3 5751.9 9.1178

payload only needed to be turned on and picked up by the crane with no other setup. In future iterations of the project, the prototype sensor rig will be redesigned to be more ergonomic or may be replaced by a permanent sensor system mounted to the crane. The data can be periodically processed through the 3D mapping system and a client-based application can analyse it to create reports on pit distortions. Automating this process has the advantages over the current human-derived methods by being more consistent and less prone to human measurement errors, and alleviates the manual effort required. References [1] M. Bosse and R. Zlot, "Map matching and data association for large-scale 2D laser scan-based SLAM," International Journal of Robotics Research, vol. 27, no. 6, pp. 667-692, June 2008.

[2] - - , "Continuous 3D scan-matching with a spinning 2D laser," in IEEE International Conference on Robotics and Automation, May 2009, pp. 4312-4319. [3] R. Zlot and M. Bosse, "Efficient large-scale 3D mobile mapping and surface reconstruction of an underground mine," in Proceedings of the International Conference on Field and Service Robotics, 2012.

1174

4 5755.2 8.0469

5 5762.1 7.8063

6 5747.6 8.4923

7 5749.5 7.0358

8 5743.7 7.7731

Anode Quality and Performance SESSION CHAIR

Matvey Golubev Rusal- ITC UCRUSAL Moscow, Russia

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

PILOT SCALE ANODES FOR RAW MATERIAL EVALUATION AND PROCESS IMPROVEMENT Lorentz Petter Lossius 1, Chmelar 1, Inge Holden2 , Hogne Linga 1, Michal Tkac 1 I Hydro Aluminium Primary Metal Technology, Ardal, Norway 2 Hydro Aluminium Ardal Carbon, Ardal, Norway Keywords: anode development, pilot scale simulation, quality, cost been 60 to 90 batches with over 1400 pilot anodes produced. Quite a wide range of raw materials have been tested, and Hydro has gained good experiences through cooperation with coke, pitch and equipment suppliers.

Abstract Primary and secondary raw materials and carbon plant practices are of critical importance for anode quality. Frequent testing of cokes, pitch and production factors in a full scale plant would be feasible, but might be high risk and expensive. It could also be time consuming as the testing would be subject to the demand of production for priority. For over twenty years, Hydro has run systematic pilot scale tests in Ardal, Norway, and since 2005 the facility has been upgraded with intensive mixers and vacuum vibroforming. Today the pilot production simulates full scale operation, and pilot scale results are successfully implemented in carbon plants. The paper discusses the factors that ensure quality practices for pitch level evaluation, aggregate screening curves and baking level control. Examples are from tests of new material sources, a study of secondary raw materials quality related to carboxy reactivity, and studies of production parameters.

Pilot scale anode testing is common in the industry, and Hydro has run pilot scale testing and has several publications from this work [1,2,3]. The perhaps best known test equipment is the R&D Carbon bench scale system with roots as far back as 1978 [4]. Work utilizing pilot scale testing includes studies of isotropic cokes [5], pitch studies [6], reactivity studies [7] and even anode baking furnace optimization [8]. The relationship between anode and coke properties is studied using pilot scale testing in another paper in this conference [9]. Pilot Scale Anodes, Challenges Below is an overview photo showing important steps in the pilot anode production. Each step has challenges relative to quality, and to achieving a realistic simulation of full scale production.

Introduction, Anode Development

In the foreground (1) are many barrels and buckets and these represent half the number of fractions required in a recent study note the labels; building up and tracking inventory is critical.

Background. Hydro in the Anode Plant Hydro Aluminium is well known as an aluminium producer and developer of cell technology such as the HAL300 and HAL4e cell technology and auxiliary operations technology. In parallel with this, Hydro has run - less well-known - programs for developing anode production technology. Through the last decades, as the cell amperages increased dramatically, parallel anode improvement programs played an important role in this success. The results of the H0yanger smelter exemplifies this: In the beginning of the 80's most trials of increasing the current above 205 kA in H0yanger were a failure, and insufficient anode quality was one of the contributions to these failures. Today this line is operated at 285 kA and this success has to a high degree depended on the ability of the carbon producers to meet the quality required of anodes that must sustain current densities at 0.90 A/cm 2 and above. The amperage load for Hydro pots are world class for end to end cells. and this can not be achieved with second class anode quality.

On the left, the multideck screening machine (2) for preparing aggregate fractions is a straightforward devise - the challenge here is ensuring that out-of-the-drum or out-of-the-bigbag material is homogeneous throughout a study. Suitable mixers are a key to useful pilot scale testing and (3) is an Eirich intensive mixer, run with test portions of 40 kg - there are many challenges among which we can mention estimating the pitch level, simulating the correct temperature and not overdoing mix energy input.

For the anode producer, maintaining and improving anode quality is always a challenge, partly due to the very long cycle time in production. The feedback loop for baked anode properties is counted in weeks, making full scale studies of material and process changes very time consuming. To speed this up, Hydro has turned to pilot scale. Over twenty years the work has evolved from using dry pitch, sigma-mixers and atmospheric vibroforming of cubic blocks to today's use of liquid pitch addition, lab scale Eirich mixers, pre-heated aggregate and close control of energy input and temperature including a controlled cooling stage. For vacuum vibroforming both die and plunger are thermally controlled.

Figure 1: 1) Part of aggregate inventory; 2) Multideck screening machine; 3) Mixer Eirich RV08; 4) Vacuum vibroformer.

Pilot Scale Studies The pilot scale studies are important in the materials part of anode development. For the last six years the annual scale of testing has

1177

In the back we see a sophisticated vacuum vibroformer (4) design made by Siegfried Wilkening at the VA W Bonn lab; the vibroformer has oil heating both in die and load - similar to mixing, a challenge here is to run at correct temperature and not overdo the vibroforming energy input.

pilot with thermally equilibrated freshly oiled walls yields a nearly homogeneous test piece with very little gradients. In cracking studies, this is a disadvantage, but in most investigations a homogeneous pilot anode with good repeatability ensures good quality comparison.

Working Environment Compared to the anode plant, technicians running pilot studies will be more exposed to the raw material dust and fumes; HES issues need to be addressed and followed up closely. The laboratory has considerable experience in this area and the photo above shows five point suction devices for fume collection. However, some stages in the preparation still require use of breathing masks with appropriate filters and further equipment development aims to reduce all open material handling. This will also yield better temperature control in the production. In the lower right corner is a barrel on a wheeled stand; the pilot work can involve some heavy manual work, and lifting is now mostly motorized; and wheeled barrels are preferred for stock. Simulating a Full Scale Process with Small Devices

Figure 3: A Hydro vibroformer, making anodes 300 times larger than the lab scale vacuum-vib (insert).

Some steps are simple to simulate in laboratory scale, such as fractionation or preheating, and some are difficult, either due to scale, like the milling of fines, or to the complexity of the process itself like the mixing step.

The Baking Furnace The baking furnace is where the difference in scale is most apparent; the photo shows the 112 ktonne per annum Hydro furnace (ABF3 at Hydro Aluminium Ardal, when new) and the two cupola furnaces at the Ardal testing facility. The baking simulates full scale very well as the heat treatment is tuned using the equivalent temperature method, ISO 17499 [10]; it is raw material independent. This stable baking practice ensures equal heat treatment and allows repeatable baked density, coke yield and anode shrinkage, enabling us to do comparison of batches made months apart.

Anode Fines Stability in the anode fines production impacts positively on the whole downstream line; process stability; anode mechanical strength and crack resistance. The laboratory mill is a batch device, several hundred times smaller in volume than a windswept mill. To make the lab production of fines realistic the milling ball size distribution is important. Milled product is controlled with sieve analysis, grain size analysis down to 0.001 mm and with Blaine number.

Figure 4: ABF 3, Hydro Aluminium, Ardal Carbon runs at 112 ktonne per annum with a section load of 168 anodes; in the lab cupola furnace the equal heat treatment zone limits the load to sixteen pilots, by weight less than 1I25000th offull scale.

Figure 2: A full scale wind-swept ball-mill, a thousand times larger than the lab mill with capacity 7-kg per 2-hour milling. The vibroformer A limitation in simulation is the size of the pilot anode compared to a full-scale anode. Full-scale vibroforming creates gradients due to the cooling, drag and push of the steel walls; the sheer mass of the free-swinging paste in the center of the form, and unavoidable anisotropy of packing. Compared to this, a 3.80 kg

Scope of Hydro Anode Development

What is the place of the pilot scale testing in anode development? For Hydro anode development has been a wide field in the last two decades, from baking furnaces to vibroformers:

1178

•

Anode Baking Furnace capacity increased for the vertical flue Hydro ABFs in Sunndal and Ardal so that amperage increases were covered by retrofitting within existing infrastructure [ll]

•

ABF refractory qualities developed with suppliers and scientific institutions enabling designs with better heat distribution in all ABFs [12]

•

Safety in ABF operations improved through CE certified safety systems for closed top furnaces based on the Directive of Machinery and TEC standards [13]

•

Further HES improvements achieved with installation of RTO fume treatment systems which combined with a downstream electrostatic precipitator placed the Ardal plant in world class for low atmospheric emissions

•

Baked anode design improved, including deep sawing of slots and a unique system for drilling of stub holes based on Norwegian North-Sea oil-drilling technology [14]

•

scale the risk of cracking can, unfortunately, not be simulated. There are mechanical properties that can be measured that indicate strength and both thermal and mechanical properties that indicate shock resistance, but the direct feedback of strain through cracks is not possible. Can Anode Density be Tested? Care must be taken when testing for higher anode density. Often high densities are considered a gain, and in pilot scale it is no problem to supermix the paste and flatten the anode in the vibroformer to green densities of 1.70 g/cm3 and above. But that is not the purpose of pilot scale simulation of a real paste plant. For simulation, a too high density is equally unsuited as a too low density; high energy input will give high density, but will not aid production! Making Green Pilot Anodes A typical batch is 40+ kg with 35.0 kg aggregate, allowing four pilot anodes to be made with moderate spread in green density. The photo shows 20 pilot anodes from five batches, and illustrates the sequential numbering system, with a letter referencing the material. The green pilot anode is 3.S0 kg; this is kept constant due to an increase in the density ofO.OOS kg/dm3 per kg paste. The spread of the green density within the 4-pilot batch is low; in 2009 the average standard deviation within 59 batches was O.OOS kg/dm3, or 0.21 %rel. Previous to establishing a constant inweight, the standard deviation within a batch could be 0.026 kg/dm3.

In the paste plant, stabilizing the aggregate quality through coke blending, stabilizing the fines through better ball mill circuit and air classification; improving the mixing string by adapting Eirich mixers to existing lines, improving butts with eddy current separator [15] and developing the Hydro vibroformer with automatic control of the anode density using CarboMaster

From this list it is obvious pilot scale anode testing till today has been a small part in anode development in Hydro, but currently it is becoming a critical and essential part for anode quality. The reason is the wider range of raw materials being taken into use, and the need to test in ways that ensure the paste plants are prepared. What Can Be Tested? This is an interesting and debatable question, what can be tested, realistically, in pilot scale? A list of what has been meaningfully tested till today includes •

Qualification of a new raw materials; pilot testing gives good guidance for what anode quality can be expected when a new raw material is introduced in the full scale production

•

Reactivity studies on coke and butts - the butts level, butts contamination level, butts' fines; butts cleaning issues and interaction of sodium and sulfur and desulfurization at higher baking levels

•

Aggregate studies, closely reproducing full scale recipes including all fractions; making similar recipes for different cokes; introducing new cokes, varying the fines content and fineness; studying anode physical properties with higher or lower coarse grain content

• •

Figure 5: Five batches - each with four pilot anodes, individually marked. Mass is 3.S0±0.01 kg/dm3. Over 1400 have been made in the last six years. Mass and dimensions are measured pre- and post-baking to determine green density, green aggregate density (GAD), baked density and changes during baking such as coke yield (CY) and shrinkage.

Anode properties with unusual raw materials, both pitch and cokes. and even coal and tests with charcoals can be done in pilot scale with no risk of upsetting the anode supply

The Raw Materials: Coke and Butts Fractions The practical solution for large studies has been to collect ready fractioned reference material from Ardal Carbon, e.g. 2.5 metric tons all together of coke fractions, mill product and butt fractions. This is material for 60-S0 batches.

Evaluation of process equipment regarding the suitability for use with different raw materials And What Can Not Be Tested?

Pitch Level Correct pitch level for good comparison is a challenge to get right. The literature refers to pitch optimization procedures where anodes are made over a range of pitch levels and the resulting

The major limitation is the scale; a 4 kg pilot is only of an anode. The size-dependent gradients will not playa part for pilots; as the thermal strain during baking will not be as large as in full 1I250th

1179

peak of the baked density curve indicates an optimum pitch level. However, for Hydro, when running studies simulating full scale production this has not turned out to be a good solution. Tn full scale, the pitch level can be 1-3 % down on the left shoulder of the optimization curve, and in addition the anode plant pitch control is very tight; e.g. the ramping interval for pitch control might be 0.05 wt% or 0.1 wt% around an average of 14.0 wt%.

Figure 6: Four level of coarseness of aggregate.

When testing with paste plant fractions, the practice has been to use the same pitch level used in the plants, with a small addition due to a batch being made in a clean pan. When testing a new coke, the practice has been to select a likely pitch level based on coke type and inspect the paste for consistency and the vibroformed green anode for surface appearance. A next batch will be made with a small shift in pitch level. Work is underway to replace this practice with analytical procedures and methods for improved control including viscometry, wettability and even pitch distribution studies by quantitative image analysis.

This was a large study, with the plan shown below; four coke materials (blend/single source), four levels of coarseness, and three baking levels. For a pilot anode study this was relatively straight forward as all cokes were known and it was easily recognizable if results were away from reasonable values. Produced Sieving Curves & Batch numbers Week Four pilots per batch Fine

The Baseline Testing is started with establishing a good simulation of the current paste plant production in an experimental design for pitch and butts percentage, at several baking levels. Testing then proceeds - all the time with the advantage of having the original simulation as a baseline for effects. Standard testing are series with different recipes; series with blending different out-of-thedrum materials into the reference aggregate; alternative pitches; testing the effect of butts adjustments on reactivity and adjustments of process energy input. Tn these tests the system with the reference baseline composition is a great help in evaluation; and repeating the reference batch at any time is a help to check that the pilot line is stable and comparable.

Fine-Xi

Fine-X2

Baking Le.el °E Bulk 1.01 & dens

Coarse

o

'"

o

o

M

'"

M M

2009-w07 2009-w07

Coke B 2008-w50 2009-w1i 2009-w11 2009-w06

II•••-----==----,

Brand New Materials; Standardized Testing and Trends Sometimes little or no ready made coke fractions and fines are in the aggregate - then a set of standard parameters for aggregate recipe, preheating temperature, pitch temperature, mixing energy input and vibroforming energy input are used. The material will be tested at 2-3-4 levels. This is a less certain simulation of the anode plant; what is observed will not be as accurate. But a series' trend will be relevant, pointing the direction for full scale results. Figure 7: Four cokes; four recipes; each batch four pilot anodes; baking levels 1150° (underbake), 1230 0 E and 1330 0 E (overbake).

Baking Level The baking is done, packed in coke within a refractory container. in cupola furnaces of 16 or 8 pilot anode capacity, determined by the size of the zone of equal heat treatment. The zone of equal heat treatment was mapped using the equivalent temperature method [I, I 0]. The repeatability is better than 100 E enabling reliable comparison at different baking levels, typically normal 1230 0 E, underbaked 1150° and overbaked 13300 E, very useful for coke studies [2].

1.00

Correlations DensGreen Pilot [g/cm 3] vs. Aggregate Properties

0.80 0.60 DAD

0.20

Examples from Running Studies

000

Tn the following, some of the tools required for a pilot scale line simulating a full scale line are presented, and some issues encountered when trying to match full scale anode manufacture are discussed.

-0.20

8 11

-DAD

-0.60 -0.80

Example 1. Coarseness of the Aggregate Using fractions from the paste plant is very suited for recipe studies. A study was run in 2009 to look at effects of adding a higher percentage of +5.6 mm material, both coke and butts.

-1.00

Figure 8: Average correlation of the four coke materials across baking levels - the red indicates the +5.6 mm fractions.

1180

The results shown in Figure 8 are the correlation of the green density with several aggregate properties. This density is the average green density across coke materials and baking levels. An increase in the coarse coke fraction increased the green density.

Results from this test are shown in the Sharing of Results section below. Note that the purpose of this example is to illustrate systematic use an Excel dashboard type sheet for result presentation and distribution, not to show actual results. However, if the PDF is magnified details are visible. Figure 12 depicts a page with concentrated information, in this case with four groups of the three cokes - three pitch levels, low, medium and high, and the fourth group is a simulation of another paste plant shown in Figure 10 as "High" mix and vib energy input.

Example 2, Issue of linearity of coke properties A result from a small study of coke bulk density showing the effect on VBD when blending two coke materials, LS and NS. It is included to show the versatility of pilot anode testing. For this set the Blend followed the lower bulk density normal sulfur coke material and not the low sulfur, higher density, coke material.

Example 4. Dilemma of large experimental designs As a last example of running studies a special case of too many ideas is shown - the purpose was to test reactivity with coke and butts additions. The study stretched over too much time, invalidating overall comparisons. Individual segments were useful 2x2 factorials, but the scope grew past original planning with many additions such as an extra recipe, extra calcination levels, and this stretched the study beyond the stability of the coke materials.

1.2 1.1

'-'

El 0

(()

>

0.9 0.8 0.7 0.6 0

2

4

6

8

10

Grain size [mm]

Figure 9: VBD for many fractions; the LN and NS cokes were blended; the ANODE result shows bulk density of crushed anode. Example 3, Comparing Cokes Pilot scale allows realistic cost-saving pre-studies that can show if a new material will run normal, or they can give insight early of potential anode issues. One study was run in connection with and previous to renegotiation of a coke supply contract. Two potential cokes C and B were compared with Ref, the current coke. Mixing lnput

Vib input

Pitch Level

Pitch Leve! low

mid

high

Figure 11: Test scheme of a study that started simple but ran too far with many extra parameters. Na0503 means 503 ppm in the fine butts fraction.

High Low Low Low

Low

14.1

Current Issues

Low Low Low Low

B B B B

PlanlZ

High Low Low Low

Plant2 Planl2 Planl2 PI.nI1

Low Low Low High

Analysis Development Development of analysis methods is important in the work; both establishing new methods and learning to use existing analyses better. One must be aware that most standard analysis methods for anodes are made for mass production and are used to report an average for a lot, a week's or a month's anode production. For instance, the specific electrical resistivity has a within-lab precision when comparing two anodes of 1.2 /lnm at 95% confidence level; two anodes can not be distinguished if the difference is below 1.2/lnm [16]. Using this value of spread to estimate the uncertainty in the average of a lot represented by 25 samples gives 1.2/-V(25), or 0.24 /lnm; this means the average reported is a good estimate for the lot. But for pilot scale comparison using one measurements for each anode, the uncertainty is 1.2/lnm at 95% confidence level! Results are not

14.5 L()w Low Low High

13.5 14.1 14.5 14.1

Figure 10: Test scheme for evaluating two replacement cokes, see results in Figure 12. There are usually several "Change" batches to review effects of process parameters.

1181

significant if the difference is less so doing more than one analysis can be critical for the strength of conclusions in a pilot scale test program. Awareness of this is important when considering number of parallels, or the need for analysis improvement.

Acknowledgement Thanks to Hydro laboratory personnel who have assisted in the development of practices and production of the anodes; especially Kirsti Gulbrandsen, Audun Bosdal and Lars S0fhage. Also thanks to suppliers who have contributed to equipment and material development including Eirich Maschinenwarefabrik; Statoil Mongstad, Rain Cll Carbon Ltd, Koppers and RUtgers.

Sharing of Results As pilot scale testing grows the need for efficient procedures become apparent: systems for planning that are recognized by the team involved; systematic execution of test programs so the work can be delegated; sharing of the results - distribution of results. The Excel sheet in Figure 12 illustrates a system of compact information that is used to distribute results.

References 1. Lorentz Petter Lossius, Inge Holden, Hogne Linga, The Equivalent Temperature Method for Measuring the Baking Level of Anodes, Light Metals 2006, p609 2. Lorentz Petter Lossius, Keith J. Neyrey, Les Charles Edwards, Coke and Anode Desulfurization Studies, Light Metals 2008, p881 3. Bodil E. Monsen, Arne Petter Ratvik, Lorentz Petter Lossius, Charcoal in Anodes for Aluminium Production, Light Metals 2010, p929 4. A.M. Odok and W.K. Fischer, Application of Pilot Plant work in Prebaked Anode Manufacturing, Light Metals 1978, p269 5. P.J. Rhedey, Behaviour of Isotropic Petroleum Coke in BenchScale Anodes, Light Metals 1987, p491 6. E.R. McHenry, J.T. Baron and W.E. Saver, The Effect of Thermal Treatment on Industrial Pitch and Carbon Anode Properties - Part 2, Light Metals 1994, p525

Figure 12: Example of compact presentation of results from the test plan in Figure 10. Comparison of a reference coke, Ref, with two cokes Band C; the green bar is full scale anode results.

7. T. Muftuoglu, B. Steine and R. Fernandez, Anode Burning Behaviour and Sodium Sensitivity of Coke from Different Feedstocks: A Pilot Scale Study, Light Metals 1993 p543 8. Vinicius Piffer, Markus Meier, Paulo Miotto, Ciro Kato, Marcos Aurelio Silva, Peter Sulger, and Raymond Perruchoud, Process Optimization in Bake Furnace, Light Metals 1994, p959

Visiting Scholars The pilot facility is very suited for inviting students for summer jobs; this can be followed with related Project work and a MSc thesis. and even PhD level work. This is a boost for the research group and gives positive signals to the organization. Visiting work sometimes is of a very specialized nature and e.g. addresses special anode materials such as coal materials or charcoal, or unusual binders or other non-traditional applications [3].

9. Lorentz Petter Lossius, Marvin Lubin, Les Edwards, Julien Wyss, Relationship between Coke Properties ad Anode Properties - Round Robin 19, Light Metals 2013, In print 10. ISO 17499 (2006) - Carbonaceous materials used in the production of aluminium - Determination of baking level expressed by equivalent temperature

Pilot Testing and Full Scale Testing

II. Michal Tlmc, Anders Ruud, Inge Holden, Hogne Linga, Hydro Aluminium's Historical Evolution of Closed Type Anode Baking Furnace Technology, Light Metals 2013, In print

The tinal proof for anode quality is the performance in the pot room. The tinal result is therefore not seen before pilot findings are introduced in regular anode production. These are examples where pilot scale testing has played an important part.

•

Pilot scale to full-scale testing of anodes with new raw materials

•

Butts cleaning and limiting impurities that are circulated back into the anode aggregate e.g. by butts fines removal

•

Adjusting recipes for stability; fines level, fines fineness and pitch level control; limits to gluing together of anodes in the ABF

• • •

12. Anders Ruud, Hydro Internal Report TEK90/l00 with Christian Sch6ning, SINTEF 13. Inge Holden, Olav Sa::ter, Frank Aune, Tormod Naterstad, Safe Operation of Anode Baking Furnaces, Light Metals 2008, p905 14. Hogne Linga, Drilling of Stub Holes in Prebaked Anodes, Light Metals 2003, p541 IS. Juraj Chmelar and Hogne Linga, Use of Eddy Current Separator in Butts Processing, Light Metals 201 0, p959

Studies of raw materials versus anode properties

16. ISO 11713 - Carbonaceous materials used in the production of aluminium Cathode blocks and baked anodes Determination of electrical resistivity at ambient temperature

Plant optimization of aggregate, mixing improvements and forming improvements In close cooperation with electrolysis, improving rodding produced anode cover materials and general characterization and monitoring of ACM

1182

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

RELATIONSHIPS BETWEEN COKE PROPERTIES AND ANODE PROPERTIES - ROUND ROBIN 19 Lorentz Petter Lossius 1, Marvin Lubin2, Les Edwards 2, Julien Wyss3 iNorskHydro ASA, P.O. Box 303, 0vre Ardal NO-6882, Norway 2 Rain Cll Carbon, 2627 Chestnut Ridge Rd, Kingwood, TX, 77345, USA 3R&D Carbon, PO Box 362, CH-3960 Sierre, Switzerland Keywords: Calcined Coke, Bulk Density levels prior to the production of larger, 150mm diameter pilot scale anodes at the optimum pitch level.

Abstract This paper discusses the preparation and production of bench scale and pilot scale anodes with five different calcined coke samples prepared for a world-wide calcined coke round robin (RR). A key objective of the RR and anode testing was to look for relationships between calcined coke properties and anode properties, particularly coke bulk density/apparent density results and anode densities. The calcined coke RR was the 19th organized by Rain CIT Carbon, but this time it was a collaborative effort with Hydro Aluminium and R&D Carbon. Calcined coke results for RR 19 are discussed in greater detail in another paper in these proceedings. Bench scale anodes with the five cokes were prepared at R&D Carbon and used to select optimum pitch levels for production of pilot scale anodes. This is the first time a RR with such a broad scope has been coordinated and published and it has provided some useful data for the industry.

The primary objective of the above work was to look for correlations between the various coke bulk density and apparent density tests and anode properties. Correlations with baked anode density were of special interest but the work provided a good opportunity to look at correlations between all calcined coke properties and baked anode properties. The five calcined cokes selected for the round robin were chosen quite deliberately to represent extremes in terms of bulk density, structure and chemical analysis (primarily S and V). Three of the five cokes were single source, straight run calcined cokes (A, Band S) and cokes C and HB were blended calcined cokes. Calcined Coke Properties and Anode Production The properties for the five coke samples are shown in Table 1. Most of the results are based on measurements made at R&D Carbon. The results for 28x48 mesh (2.36-1.18mm) VBD, 20x35 mesh (0.85-0.425mm) VBD and Hg AD are all based on the industry average results reported in the companion paper.

Introduction During the 2010 TMS meeting in Seattle, the ASTM Standards Committee convened a special meeting to discuss coke bulk density testing. At least four different bulk density or apparent density test methods are used within the industry and there is no agreement on which method provides the best predictor of coke performance in anodes. Coke bulk densities have decreased on average over the last 10 years [1] and there is renewed interest in the relevance of the different test methods and their ability to predict anode densities. One of the actions from the meeting was to organize a dedicated session on coke bulk density testing at the 2011 TMS meeting. Seven papers were published in that session [2,3,4,5,6,7,8] and it was followed by a panel discussion on what could be done to reach better consensus within the industry.

Table 1: Calcined Coke Properties Property

Unit

Coke A

Coke B

Coke C

Grain Size+8mm

%

10.8

0.4

11.7

1.6

11.9

TBD (1-2mm)

kg/dm 3

0.73

0.98

0.81

0.86

0.83

VBD 28x48 Mesh

kg/dm 3

0.83

0.99

0.86

0.90

0.88

VBD 20x35 Mesh

kg/dm 3

0.79

0.95

0.82

0.86

0.84

Hg AD (Pechi ney)

kg/dm'

1.72

1.76

1.72

1.74

1.72

Grain Stability

%

65

77

80

76

83

Real Density

kg/dm 3

2.08

2.00

2.06

2.06

2.07

Crystallite Size Lc

A

29.9

29.3

26.6

26.6

28.2

Spec. Elect. Resist.

!lO.m

434

580

454

498

457

8.9

4.4

6.9

6.3

9.0

CO, Reactvity at 1000°C %

One of the recommendations from the panel was to conduct an industry wide, calcined coke round robin comparing the various bulk and apparent density test methods. A round robin (RR) was organized by Rain Cll Carbon and Hydro Aluminium and is the subject of a companion paper in these proceedings [9]. It was the nineteenth RR organized by Rain CIT and is hereafter referred to as RRI9. The preparation of the five coke samples is well described in the companion paper along with all the test results including withinlab repeatability and between-lab reproducibility data. Results reported in the paper include chemical analysis (S, V, Ni, Fe, Si, Ca, Na, P), real density, Lc, mercury apparent density, tapped bulk density (ISO 10236) and vibrated bulk density (ASTM D4292 and ASTM D7454). Bulk densities for the different preparation methods were also measured using the GeoPyc instrument [6]. A total of 28 labs participated in RR 19.

Coke S Coke HB

Air Reactivity at 525°C

%/min

0.06

1.15

0.38

0.21

0.35

Sulfur

%

1.46

4.37

3.05

1.16

2.16

Vanadium

ppm

94

619

404

144

241

Nickel

ppm

173

263

205

64

172

Iron

ppm

167

537

295

78

195

Silicon

ppm

70

159

129

54

228

Calcium

ppm

77

120

87

24

155

Sodium

ppm

42

121

45

29

62

Some general comments on the five cokes are as follows: • Coke A is a relatively low bulk density, low sulfur coke. It is always used in blends with other cokes. • Coke B is a high bulk density, highly isotropic coke with high Sand V. It is used at low levels «10%) in blends. • Coke C is a blended coke used routinely at some smelters for anode production. It has a moderately high S and V level. • Coke S is a relatively high bulk density, low sulfur coke. It is used primarily as a blend coke. • Coke HB is another blended coke used routinely at smelters for anode production.

In a significant expansion of the RR 19 work, 150kg samples of each of the five cokes were sent to R&D Carbon for the preparation and testing of bench scale and pilot scale anodes. The bench scale anodes were prepared to determine optimum pitch

1183

Four batches of 5.5kg were prepared for each coke by mlxmg according to the recipe and preheating at 200°C for 12 hours. The preheated coke and solid coal tar pitch were added to an Eirich mixer and mixed at 172°C and then cooled to 150a C by injecting water prior to forming. Forming was carried out via hydraulic pressing for 1 minute at a pressure of 500 bars. The green anodes were 146mm in diameter and 180-200mm in length.

Preparation of Bench Scale Anodes Samples of each of the five cokes were sized into different fractions including preparation of a fines fraction via jet milling (details in next section). No butts were used in the recipe. Small bench scale anodes (50mm diameter x 100mm length) were produced at 4 pitch levels using a 112°C Mettler softening point, medium Ql pitch. The cores were baked over 20 hours to 1l00°C. Graphs of baked anode density vs pitch content for the five cokes are shown below and were used to select the optimum pitch levels for pilot scale anodes.

The green anodes were baked to 1100°C in a pilot baking furnace which can bake up to 12 anodes at once. Three core samples were drilled from each baked anode with a diameter of 50 mm and length of 200mm. Testing was performed according to standard ISO test procedures.

Baked App. Density 156 , , . . . - - - , . . . . - -......- - - , . . . . - - - , . . . . - - -......--.........,

Pilot Anode Results Results from testing the pilot anodes are shown in Table 2. Without the addition of butts, densities are typically kg/dm 3 lower than full size production anodes. Mechanical properties, including permeability and electrical resistivity are also typically a little worse than production anodes, while thermal conductivity, thermal expansion coefficient and reactivity are similar.

152

1.48

Table 2: Pilot Anode Properties

1.44 12

1t1

15

16

17

Pitch Content

18

£%J

Pro perties

Units

Coke A

Coke B

Coke C

Coke S

Optimal P itch Co ntent

%

17.0

12.5

"E.O

15.5

"E.O

Green Apparent Density

kg/dm 3

1.53

1.53

1.52

1.59

1.56

Baking Loss

%

6.0

4.6

5.6

5.9

5

Baking Shrinkage

%

2.6

3.9

2.5

2.0

1.7

Baked Apparent Density

kg/d m 3

Spec. Elect. Resistance

Figure 1: BAD vs Pitch Content for Bench Scale Anodes Production and Testing of Pilot Anodes Pilot scale anodes were prepared at the optimum pitch content for each coke according to the scheme in Figure 2 and described previously [10].

Coke HB

1.47

1.52

1.47

1.53

1.49

62.1

53

62.5

65.4

62.4

Compressive Strength

Mpa

33

47

26

29

36

Flexural Strength

MPa

9.8

17

"Kl.9

8.6

9.9 3.8

Coef. Thermal Expansion

'K)-6/K

3.7

5.7

4.0

3.5

Air Permeability

nPm

2.7

13.2

2.5

1.8

2.6

Real Density

kg/d m 3

2.06

1.99

2.05

2.07

2.07

Thermal Conductivity

W1mK

3.06

3.19

2.85

2.91

3.12

C02 Reactivity Residue

%

92.1

95.8

93

91.3

90.5

Air Reactivity Residue

%

93.7

65

71.2

78.5

87

Impurities by XRF S

%

1.3

4.19

2.36

1.05

1.86

V

ppm

84

528

302

129

207

Ni

ppm

"El

232

156

62

154

Fe

ppm

205

504

328

135

309

Si

ppm

89

227

154

71

341

Ca

ppm

82

115

95

37

171

Na

ppm

50

134

42

29

61

The data generated during the RRI9 study was extensive with 28 labs providing data on calcined coke properties and R&D Carbon providing anode data on both bench scale and pilot scale anodes. It is not possible to review and present all the data in this paper so only selected data and correlations will be discussed for pilot scale anodes. More details on the precision of the various coke property tests can be found in the companion paper [9]. Pitch Level and Anode Density Correlations Of primary interest in this study were correlations between the various coke TBDIYBD results and optimum pitch levels and anode densities. Given the wide range in coke bulk densities for the five cokes (up to 35% for some tests), one might expect to see a wide range in optimum pitch levels and baked anode densities. R2, the Coefficient of Determination, is used in the discussion; each lab's RR 19 coke property result is correlated with the common anode property result. Labs with less than five cokes measured were not included in the comparisons. The bench scale anodes were useful for selecting optimum pitch levels and the levels ranged from a low of 12.5% for coke B to a

Figure 2: Production of Pilot Anodes

1184

high of 17% for coke A. Since coke bulk density is intended to be an indirect measure of coke porosity, it is reasonable to expect a good correlation with optimum pitch level. This was indeed the case, and the R2 correlations for many of the labs that participated were above 0.90 for the various YBO/TBO tests as shown in Table 3. The column titled "Count" refers to the number of results used for the correlations. Some obvious outlier results with R2 values 50%) due to thermal shock problems during anode heat-up in the cells. It works well as a blend coke and is used routinely in anode blends up to 10%. The CTE for the other anodes falls within typical industry ranges. The thermal conductivity of all the anodes also falls within normal industry ranges.

Figure 7: R2 Values for Coke Real Density vs Anode RD Anode Chemical Analysis All the coke chemical analysis results from RR19 showed a strong correlation (R2>0.90) with the anode chemical analysis results shown in Table 2. This is expected since only coke and pitch were used to make the anodes so the only chemical analysis change from coke to anodes was the dilution effect of adding pitch. As shown in the companion paper [9], the agreement between the 28 labs on all the chemical analysis results was generally excellent. These tests have good precision and are well established and well accepted within the industry.

Anode Reactivities Since the bench scale and pilot scale anodes were produced without butts material, there is limited value in drawing too many conclusions about the anode reactivity data. All the anode CO 2 reactivities are generally excellent with residues >90%. There is a good correlation between anode CO 2 reactivity residues (CRR) and coke and anode sulfur levels. There is also a good correlation between anode CO 2 reactivity dust (CRD) levels and coke and anode S levels. The R2 for CRD and coke S level has an average value of 0.85 across all labs and for CRR, the average is 0.75.The correlation between reactivity loss (CRL) is not as strong at 0.45.

The only caveat to the above, is that the agreement between analysis results deteriorated at higher levels of sulfur and trace metals. This is believed to be due to a lack of suitable high range calibration standards at some labs. The majority oflabs used x-ray t1uorescence (XRF) for sulfur and trace metals analysis and this method is very dependent on having a reliable set of calibration standards covering the full range of elemental concentrations being measured in unknown samples.

Measurement of coke CO 2 reactivities by the ISO 12981 test was not part of RR19 but the R&D Carbon results for coke CO 2 reactivity are included in Table I. The R2 correlations between coke CO 2 reactivity and CRR, CRD and CRL were respectively; 0.64.0.15 and 0.78.

Electrical and Anode Mechanical Properties One notable result from this study was the much lower electrical resistivity of the Coke B anodes, Figure 8. Similar results have been found in other unpublished pilot anode studies where significant volumes (> 10%) of isotropic cokes have been used in the aggregate recipe. The reasons for this are not fully understood but it may warrant further investigation. One thing the results

The air reactivity residue (ARR) of the coke B anodes was the lowest of the five cokes (65%) which is not surprising given the significantly higher V (619ppm) and Na level (12Ippm) of this coke. This was followed by the coke C anodes which were a little

1187

better with an ARR of 71 %. This coke had the next highest V level at 404ppm.

Acknowledgments The effort required to coordinate and complete this RR study was large. The production of bench and pilot scale anodes was a major expansion of the RR and the authors wish to acknowledge the large effort and excellent work by R&D Carbon to complete this part of the RR. Thanks are also extended to the 28 labs that participated in the RR.

Measurement of coke air reactivity was not part of RR 19 but the R&D Carbon results are presented in Table I for each of the 5 cokes. The R2 values for coke air reactivity vs anode reactivity residue (ARR), air reactivity dust (ARD) and air reactivity loss (ARL) range from 0.59 to 0.64. Despite the lower than average anode air reactivity results for coke C, this coke is used routinely at the 100% level at a smelter in the US. Further details can be found in another paper in these proceedings [14].

References 1.

Discussion and Conclusions The primary objective of this part of the RRI9 study was to establish correlations between the various coke bulk and apparent density tests and anode densities. A secondary objective was to look at correlations between other coke properties and anode properties. The study showed that all VBD/TBD procedures in common use through the industry do a good job of predicting optimum pitch levels. Their ability to predict baked anode densities is not as reliable and none of the VBD/TBD tests or the Hg AD test stood out as being any better than the others.

2.

3. 4.

The scope of this study was limited so it is not possible to draw definitive conclusions about coke and anode density correlations. The study does, however, highlight the relative complexity of the anode production process. Many factors affect baked anode density and coke porosity/bulk density is one of these factors. Anode plants have gotten much more capable over the last 20 years and many anode producers are making high quality, high density anodes from cokes that would have been regarded as high porosity cokes in the past.

5.

With the above in mind, it probably makes sense for the industry to use VBD/TBD tests with the highest level of repeatability and reproducibility. As discussed in the companion paper [9], this means tests without extensive sample preparation such as the ISO 10236 test and a new ASTM procedure being developed using natural fractions. These tests will at least allow reliable tracking of coke bulk density trends over time and between different coke sources.

8.

The anodes produced with coke B produced the most unusual results in this study which is perhaps not surprising given the highly isotropic structure of this coke. It is likely that the granulometry was not optimum for this coke and a lower percentage of fines would likely give better results for anode density and permeability. The significantly lower electrical resistance of anodes made with this coke was of interest and highlights the lack of correlation between anode ER and coke SER. The mechanical strength of these anodes was also significantly higher than the other anodes as was the CTE.

11.

6.

7.

9. 10.

12. 13.

14.

Most of the other anode properties were in line with expectations. Anode reactivities correlate quite well with impurities such as sulfur and vanadium but care needs to be taken when interpreting these results in the absence of butts and the typical increase in sodium and calcium levels this causes. It is unlikely that another RR on this scale will be attempted.

Further RR's are planned however to support the development of a new ASTM bulk density test.

1188

L. Edwards. N. Backhouse. H. Darmstadt. M-J Dion. "Evolution of Anode Grade Coke Quality", Light Metals, 2012, 12071211 J. Panchal, M. Wyborney and J. Rolle, "Historical and Future Challenges with the Vibrated Bulk Density Test Methods for Determining Porosity of Calcined Coke", Light Metals, 2011, 925-930 M-J Dion, H. Darmstadt, N. Backhouse, F. Cannova and M. Canada. "Prediction of Calcined Coke Bulk Density", Light Metals, 2011,931-936 F. Cannova, M. Canada and B. Vitchus, "Calcined Coke Particle Size and Crushing Steps Affect its VBD Result", Light Metals, 2011,937-939 L.P. Lossius, B. Spencer, H. 0ye, "Bulk Density Overview of ASTM and ISO Methods and Examples of Between Laboratory Comparisions", Light Metals, 2011,941-946 L. Edwards, M. Lubin, J. Marino, "Improving the Repeatability of Coke Bulk Density Testing", Light Metals, 2011, 947-952 F. Laplante and L. Duchesneau, "ATSM D7454 Vibrated Bulk Density Method - Principal and Limitations", Light Metals, 2011, 953-957 B. Spencer, L. Johnsen, D. Kirkpatrick, D. Clark and M. Baudino, "Vibrated Bulk Density (VBD) of Calcined Petroleum Coke and Implications of Changes in the ASTM Method D4292", Light Metals, 2011, 959-963 M. Lubin, L. Edwards and L. P. Lossius, "Calcined Coke Round Robin 19", Light Metals, 2013 A. M. Odok and W. K. Fischer, "Application of Pilot Plant work in Prebaked Anode Manufacturing", Light Metals, 1978, VoLl, 269-286 U. BUhler and R. C. Perruchoud, "Dynamic Process Optimisation", Light Metals, 1995,707-714 R.J. Akhtar and M. W. Meier, "Dynamic Process Optimization in Paste Plant" Light Metals, 2006, 571575 L.P. Lossius et al. "Pilot Scale Anodes for Raw Material Evaluation and Process Improvement", Light Metals, 2013 J. Gavin, W. Marcrum, A. Weber, L. Crabtree, L. Edwards, "Impact of Higher Vanadium Levels on Smelter Operations", Light Metals, 2013

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

APPLICATION OF THE ARTIFICIAL NEURAL NETWORK (ANN) IN PREDICTING ANODE PROPERTIES Dipankar Bhattacharyayl, Duygu Kocaefe 1, Yasar Kocaefe 1, Brigitte Morais2, Marc Gagnon 2 lCentre universitaire de recherche sur l'aluminium (CURAL), University of Quebec at Chicoutimi; 555 Boulevard de l'Universite, Chicoutimi, QC, G7H 2BI, Canada 2Aluminerie Alouette Inc.; 400, Chemin de la Pointe-Noire, Sept-iles, QC, G4R 5M9, Canada Keywords: Feed-forward, Back-propagation, Regression, Linear multi-variable analysis, Artificial neural network Abstract Carbon anodes are a major part of the cost of primary aluminum production. The focus of the industry is to minimize the consumption of anodes by improving their quality. Therefore, the determination of the impact of quality of raw materials as well as process parameters on baked anode properties is important. The plants have a large data base which, upon appropriate analysis, could help maintain or improve the anode quality. However, it is complex and difficult to analyze these data using conventional methods. The artificial neural network (ANN) is a mathematical tool that can handle such complex data. In this work, Matlab software was used to develop a number of ANN models. Using published data, linear multi-variable analysis and ANN were applied to assess the advantages of custom multilayered feed-forward ANN. Results are presented which show a number of industrial applications.

The limitations of regression techniques and differential equations have led researchers to explore alternative models. Thus, research on ANN [2] has become popular. Artificial neural networks (ANNs) have been increasingly used as a model for engineering, environmental, and other applications. ANN has earned popularity because of its ability to handle complex nonlinear functions. The concept of ANN was first introduced by McCulloch and Pitts in 1943 [3]. ANNs bear similarity with biological neurons and their interaction with each other in the brain. There are different ANN models such as perceptron [3], feedforward [4], recurrent [5], radial basis function neural networks [6,7,8], etc. The feed-forward neural network (FNN), also known as the multilayer perceptron (MLP) [9] is the most widely used ANN model. A feed-forward network typically consists of three layers of neurons, namely, an input layer, a hidden layer, and an output layer. The network sends information sequentially from the input layer to the output layer.

Introduction The carbon anodes constitute a significant part of the cost of the primary aluminum production. The variations in quality of raw materials such as calcined petroleum coke, coal tar pitch, recycled butts, green rejects etc. and operating conditions during different processes such as mixing, baking, cooling etc. affect the quality of baked anodes to a great extent. This, in turn, affects the anode consumption in the electrolytic bath. The goal of the industry is to produce better quality anodes in spite of the variations in raw materials and process conditions. This would have been easy if there was some distinct mathematical relationship between the input parameters and the properties of the baked anode. But, in reality, this relationship does not exist and the control of production is usually based on experience and intuition [1]. However, the plants usually maintain a large database. Although the data is complex, the proper analysis of these data can deliver significant information that can be used to control and improve the quality of anodes. The complexity of the data makes it difficult to be analyzed by conventional analytical tools.

With the advent of the back-propagation algorithm in 1980s, ANN has started to gain its popularity. The back-propagation algorithm made the training of an ANN model easier using experimental data. The parameters of the network are updated during each pass of the training with respect to the network's prediction error [10]. ANN is now used to solve problems previously thought to be impossible or very difficult with traditional methods [11]. There are many reasons behind the success of ANN. The structure of an ANN is generally flexible and robust. Unlike regression, it does not require a specific equation based on the system to relate the input and output variables. The general structure of an ANN can be applied to practically any system [12]. Also, ANN can handle situations when outliers exist in the data [13]. White et al. [14] described a feed-forward neural network with a sigmoid hidden layer as a universal function approximator. As a result, the artificial neural networks have been viewed as a powerful tool for predictions.

The linear multi-variable analysis and the nonlinear regression analysis are important tools for analyzing the relationship between multiple input parameters and an output parameter.

The major problem or limitation of ANN is in its development phase. Presently, there is no formal rule available for developing networks [15]. Thus, the development of a suitable ANN model is often time consuming [16].

For the linear multi-variable analysis, the dependence of the output parameter on the input parameters should ideally be linear. However, in real cases, it is hard to get a linear relationship for each and every parameter. The regression analysis can handle nonlinear relationships, but it is necessary to assume some mathematical relationship between the input and the output parameters. In case of anodes, the relationships between the parameters are highly complex and hard to generalize. Thus, it is difficult to apply those conventional methods in the case of anode quality control during production.

Different researchers have tried to introduce rules to reduce the time on trial and error. According to Sarle [17], the design of an ANN depends on a number of variables such as the size of the training data set, the number of input and output variables, the complexity of the underlying function, the amount of noise in the

1189

target variables, and the activation function used. A number of rules of thumb have been proposed [4,18,19], but these rules always try to over-simplify the problem and thus can lead to poor network performance.

c

In spite of various applications of ANN, a few works have been published regarding the application of ANN to predict carbon anode properties. Though carbon anodes are at the heart of primary aluminum production and account for a significant part of the production cost, yet regarding maintenance of anode quality, not many studies have been reported in literature related to application of ANN. Berezin et al. [I] developed a perceptron based artificial neural network to maintain anode quality at OKSA aluminum plant in Russia. The ANN model could predict and adjust variations in the production process with changes in the quality and quantity of raw materials.

(3) where, Kj denotes value of anode property Y at observation number j. IfP is the matrix ofcoefticients,

p=

(4)

In this article, a comparison of an artiticial neural network with the linear multi-variable analysis and the regression analysis in terms of their prediction capability will be presented.

then,

Method In this study, published data from the thesis of Chmelar [20] have been used for the analysis. He studied different formulations of anodes using 4 different cokes and 1 pitch. He also studied the physical and chemical properties of the raw materials, and some properties of the baked anodes were measured. Table 1, shown on the last page, summarizes all the 19 independent input parameters for 36 samples. Table 2 summarizes baked anode density, specific electrical resistivity, and Young's modulus for the baked anode samples. All three properties for 4 samples (4, 17, 30, and 32) were predicted using the corresponding input parameters by applying the feed-forward artificial neural network, the linear multivariable analysis and the generalized regression neural network. Those 4 samples were not used while optimizing the parameters by any of the methods. The remaining 32 data were used for calculating the parameters or to train the network.

Table 2: Properties of baked anode samples

The basic concept behind the linear multi-variable analysis is to express a property Y of the anode as a linear function of different independent parameters (X), X 2 ...... XN ), i.e.,

(1) If N is the total number of independent variables, M is the total number of experimental observations, then the input matrix B will take the following form:

B=

ZI.I

Z2.l

ZN.I

ZI.2

Z2.2

ZN.2

ZI.M

Z2.M

ZN.M

(5)

(2)

where, Zi,j denotes the value of input parameter i for observation number j. For M observations, the matrix for the property Y of the anode is represented as:

1190

Sample No.

BAD (glee)

SER (IlOm)

YM(GPa)

I 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36

1.45 l.47 1.43 1.35 1.36 1.33 1.39 l.43 l.41 1.38 l.41 1.38 1.395 1.489 1.45 1.374 1.4 1.378 1.39 1.42 1.395 1.42 l.5 1.481 l.429 1.493 l.47 1.345 1.378 1.356 l.41 1.44 l.42 1.405 l.499 1.475

77.3375 72.56 72.9 66.8 65.325 65.9 69.5 66.67 67.5 79.5 67.9 68.9 86.175 76.66 76.85 65.28 64.5 66 69.1 65.7 68.5 73.8 66.9 68 79.5 74.9 77.6 71.8 67.5 68.4 66.66 64.78 65.8 70.5 65.4 66.4

5.2 5.9 5.9 6.6 6.5 6.9 6.9 6.7 6.1 5.6 5.7 4.7 4.1 5.1 5.4 6.8 6.7 6.5 6.8 6.6 6 6.1 6 5.5 5.5 4.8 4.4 6.1 5.5 5.3 6.3 6.5 6 6.1 5.8 5.3

The 32 data mentioned earlier were used to calculate the coefficients. For the prediction of an output property, each input parameter was multiplied by the corresponding coefficient and the sum represented the predicted value of the property. Both the regression analysis and the artificial neural network model were done using Matlab 7.2.

Results and discussions

Figures 1, 2, and 3 show the correlation between the published and predicted values for electrical resistivity, Young's modulus and baked anode density, respectively. The figures show that the coefficients of determination are the smallest (0.093, 0.332, and 0.604 for electrical resistivity, Young's modulus, and baked anode density, respectively) in the case of linear multivariable analysis. The values for the coefficients of determination are medium (0.392, 0.846, and 0.816 for electrical resistivity, Young's modulus, and baked anode density, respectively) in the case of regression analysis. The values are highest (0.966, 0.989, and 0.947 for electrical resistivity, Young's modulus, and baked anode density, respectively) in the case of feed-forward ANN. For the feed-forward ANN, the configurations were different for the three cases. Table 3 lists the important parameters for the ANN models.

Matlab provides a built-in function 'newgrnn' for generalized regression neural networks [21,22]. It is a radial basis network that is often used for function approximation. The advantage of the method is that it can be designed very quickly. The method falls into the category of probabilistic neural networks. The normal distribution function is used as the probability density function. Each training sample is used as the mean of a normal distribution. The Euclidian distance between the training sample and the point of prediction, is used to estimate the position of prediction.

7'0

The 32 data for training were used to train the network, and the trained network was used to predict the values for the four test data set.

i

Two customized feed-forward neural network models with backpropagation training were tried using Matlab. They are the feedforward back-propagation and cascade feed-forward backpropagation networks. For both networks, one input layer, two hidden layers, and one output layer were selected.

:1 ,;

1 j

$il 6$

51

$I,

>1/

Various transfer functions such as logsig, tansig, purelin were associated with the hidden layers. The logsig function can be represented as logsig(n) = 1 / (l + exp(-n)). Similarly, tansig function can be represented as tansig(n) = 2/(l+exp(-2*n))-1. Purelin is a linear function represented as purelin(n) = n. The transfer functions process the input to a layer such that the output can be easily classified into groups of similar data. which is important for an efficient prediction.

1

.'11

J

65

&.

54

63

Initially, some random weights were associated with the normalized input parameters. The training of a neural network means the identification of optimum values of the weights associated with the input parameters. The training of the network was done using trainlm, trainbfg, and traingdm back-propagation functions. Those functions were based on Levenberg-Marquardt, BFGS (Broyden, Fletcher, Goldfarb, and Shanno) update of quasi Newton, and gradient descent with momentum back-propagation algorithms, respectively. The maximum number of iterations for training (epoch) was set to 1000. The weights were varied during each iteration based on the gradient descent learning algorithms (learngd and leamgdm) and a learning rate of 0.05 and momentum of learning of 0.9.

Figure 1: Predicted and published values of electrical resistivity

The networks were trained based on the measurement of error in prediction. The errors were measured in terms of mean squared error (mse) and mean average error (mae). The 32 data set were used for the training of the neural network. The trained network was used to predict the properties of the baked anode for the four test data sets.

S.t

'"

Pl.II1I_4

For all three cases, the predicted values were plotted against the published results for the four test data sets. The coefficient of determination for linear regression for each graph was used as the criteria for the quality of prediction. The closer the value of the coefficient of determination to unity is, the better the ability of prediction is.

",.1@4 {>I'$

Figure 2: Predicted and published values of Young's modulus

1191

Conclusions The artiticial neural network is an important tool for the prediction of anode properties. It can become an important tool for the quality control of anodes. The major advantage of ANN over the other methods is that it can efficiently handle highly nonlinear data with noises where there is no existing mathematical relationship. It is true that the development of an efficient ANN model is time consuming because it needs lot of trials and errors; but once it is developed, it can predict results for which no experimental data is available. For the training of an ANN, availability of large sets of data is important; but this is generally not a limitation in the case of industries. ANN, with its power of artificial intelligence, can save time and money for the aluminum industry.

i

j 1 i

I

Acknowledgements The technical and financial support of Aluminerie Alouette Inc., as well as the tinancial support ofthe National Science and of Canada (NSERC), Engineering Research Council Developpementeconomique Sept-TIes, the University of Quebec at Chicoutimi (UQAC), and the Foundation of the University of Quebec at Chicoutimi (FUQAC) are greatly appreciated.

Figure 3: Predicted and published values of baked anode density Table 3: Parameters for feed-forward neural network models used for prediction

Transfer function

Initialization

Property

Electrical Resistivity

Young's Modulus

Baked Anode Density

Network

newcf

newcf

newff

Layer I

logsig

logsig

logsig

Layer2

purelin

tansig

purelin

Training function

trainlm

trainlm

trainlm

Learning algorithm

learngdm

learngd

learngd

Error check

mse

mae

mae

Function

rands

rands

rands

Seed for rands function

4

7

10

References

Thus it can be seen that the customized feed-forward neural network model with back-propagation training was able to predict the output for the test data set better than the linear multi variable and regression analyses. The average percent errors in prediction are 0.8, 1.6, and 0.6 for electrical resistivity, Young's modulus, and baked anode density, respectively. In the production of anodes, there are numerous parameters that can influence the baked anode properties. Industries often maintain information about the input parameters and the output properties. These huge data can be utilized to train ANN. The industrial data is highly nonlinear in nature, and there is no mathematical relation available between those data. In such a situation, ANN has immense potential in quality control during anode production. In the case of variations in the properties of raw materials and processing parameters, ANN can predict the property of baked anode even before baking. In the case of variations in process parameters during the manufacturing of baked anodes, ANN could indicate the necessary changes in other process parameters to maintain the quality of baked anodes.

1.

1. Berezin, P.V. Polaykov, 0.0. Rodnov, V.A. Klylov, "Improvement of Green Anodes Quality on the Basis of the Neural Network Model of the Carbon Plant Workshop", Light Metals, (2002), 605-608

2.

R.S. Govindaraju, A. R. Rao, "Artificial Neural Networks: A Passing Fad in Hydrology?", 10urnal of Hydrologic Engineering, 5 (3) (2000), 225 - 226.

3.

W.S. McCulloch, W. Pitts, "A Logical Calculus of Ideas Immanent in Nervous Activity", Bulletin of Mathematical Biophysics, 5 (1943), 115 - 133

4.

M.R. Kaul, L. Hill, C. Walthall, "Artificial Neural Network for Corn and Soybean Yield Prediction", Agricultural Systems, 85(1) (2005), I - 18.

5.

1. T. Connor, R. D. Martin, L. E. Atlas, "Recurrent Neural Networks and Robust Times Series Prediction", IEEE Transactions on Neural Networks, 5(2) (1994), 240 - 254.

6.

Y. Hayashi, J. .T. Buckley, E. Czogala, "Fuzzy Neural Network with Fuzzy Signals and Weights", International loint Conference on Neural Networks, 2 (1992), 696 - 701.

7.

X.Yao, "Evolving Artificial Neural Networks." Proceedings of the IEEE, 87(9) (1999): 1423 - 1447.

8.

Z. R. Yang, "A Novel Radial Basis Function Neural Network for Discriminant Analysis", IEEE Transactions on Neural Networks, 17( 3) (2006),604 - 612.

9.

V. Cherkassky, J. H. Friedman, H. Wechsler, "From Statistics to Neural Networks", Springer-Verlag, Berlin (1993).

10. 1.K. Kruschke, 1. R. Movellan. "Benefits of Gain: Speeding Learning and Minimal Hidden Layers in Back-Propagation Networks." IEEE Transactions on Systems, Man and Cybernetics. Vol. 21. No.1 (1991): 273 - 280. 11. B. Cheng, D. M. Titterington, "Neural Networks: A Review from a Statistical Perspective", Statistical Science, 9(1) (1994),2 - 30.

1192

12. C.M. Zealand, D. H. Burn, and S. P. Simonovic, "Short Term Streamflow Forecasting Using Artificial Neural Networks", Journal of Hydrology, 214 (1999), 32 - 48. 13. J.W. Denton, "How Good are Neural Networks for Casual Forecasting?", Journal of Business Forecasting, 14 (1995), 17 - 20. 14. H. White, "Artificial neural networks: Approximation and Learning Theory", Blackwell, Cambridge, (1992). 15. G. Daqi, Y. Genxing, "Influences of Variable Scales and Activation Functions on the Performance of Multilayer Feedforward Neural Networks", Pattern Recognition, 36 (2003), 869 - 878. 16. A. Shigidi, L. A. Garcia, "Parameter Estimation in Groundwater Hydrology Using Artificial Neural Networks." Journal of Computing in Civil Engineering, 17(4) (2003), 281 - 289. 17. W.S. Sarle, "Neural Networks and Statistical Methods", Proceedings of the Nineteenth Annual SAS Users Group International Conference, (1994). 18. V.R. Prybutok, .T. Vi, D. Mitchell, "Comparison of Neural Network Models with ARIMA and Regression Models for Prediction of Houston's Daily Maximum Ozone Concentrations", European Journal of Operational Research, 122 (2000), 31-40. 19. C.X. Feng, X. Wang, "Surface Roughness Predictive Modeling: Nerual Networks Versus Regression", TIE Transactions, 35 (2003), 11 - 27. 20. .T. Chmelar, "Size Reduction and Specification of Granular Petrol Coke with Respect to Chemical and Physical Properties", (PhD Thesis, Norwegian University of Science and Technology, 1992). 21. D.F. Specht, "A General Regression Neural Network", IEEE Tras. Neural Networks, 2(6) (1991), 568-576. 22. P.D. Wasserman, "Advanced Methods in Neural Computing", Van Nostrand Reinhold, New York, (1993), 155-61

1193

'.0

34.00 34.00 34.00 36.00 36.00

35.00 35.00

17.10

17.10

17.10

20.50

20.50

20.50

16.80

16.80

16.80

21.10

7

8

9

10

II

12

13

14

15

16

34.00

35.00 37.00 37.00 34.00 34.00 34.00 36.00 36.00 36.00

20.50

16.80

16.80

16.80

2l.l0

21.10

2l.l0

17.10

17.10

17.10

20.50

20.50

20.50

24

25

26

27

28

29

30

31

32

33

34

35

36

36.00

37.00

35.00

35.00

36.00

36.00

20.50

20.50

22

34.00

23

17.10

17.10

21

17.10

20

37.00

21.10

18

19

34.00

37.00

37.00

35.00

2l.l0

17

37.00

36.00

37.00

21.10

2l.l0

5

6

37.00

35.00

16.80

2l.l0

3

4

35.00

35.00

16.80

74.00

74.00

74.00

85.00

85.00

85.00

66.00

66.00

66.00

87.00

87.00

87.00

74.00

74.00

74.00

85.00

85.00

85.00

66.00

66.00

66.00

87.00

87.00

87.00

74.00

74.00

74.00

11.00

11.00

11.00

10.40

10.40

10.40

10.20

10.20

10.20

24.50

24.50

24.50

11.00

11.00

11.00

10.40

10.40

10.40

10.20

10.20

10.20

24.50

24.50

24.50

11.00

11.00

11.00

10.40

10.40

85.00

10.40

85.00

10.20

10.20

10.20

24.50

24.50

24.50

CO2 reactivity mg/cm 2 h

85.00

66.00

66.00

66.00

87.00

87.00

87.00

HGIof Grain stability coke

16.80

')10

I

Coke Porosity

2

Sample No.

Table 1. List of independent variables

96.60

96.60

96.60

126.80

126.80

126.80

26.10

26.10

26.10

79.70

79.70

79.70

96.60

96.60

96.60

126.80

126.80

126.80

26.10

26.10

26.10

79.70

79.70

79.70

96.60

96.60

500.00

500.00

500.00

487.00

487.00

487.00

473.00

473.00

473.00

490.00

490.00

490.00

500.00

500.00

500.00

487.00

487.00

487.00

473.00

473.00

473.00

490.00

490.00

490.00

500.00

500.00

500.00

487.00

126.80 96.60

487.00

487.00

473.00

473.00

473.00

490.00

490.00

490.00

126.80

126.80

26.10

26.10

26.10

79.70

79.70

79.70

2.09

2.09

2.09

2.07

2.07

2.07

2.07

2.07

2.07

2.07

2.07

2.07

2.09

2.09

2.09

2.07

2.07

2.07

2.07

2.07

2.07

2.07

2.07

2.07

2.09

2.09

2.09

2.07

2.07

2.07

2.07

2.07

2.07

2.07

2.07

2.07

Specific Real Air Electrical Density reactivity resistivity of of coke mg/cm 2 h coke "n.m kg/dm3

0.48

0.44

0.36

0.48

0.44

0.36

0.48

0.44

0.36

0.48

0.44

0.36

0.48

0.44

0.36

0.48

0.44

0.36

0.48

0.44

0.36

0.48

0.44

0.36

0.48

0.44

0.36

0.48

0.44

0.36

0.48

0.44

0.36

0.48

0.44

0.36

Zn, ppm

0.48

0.49

0.51

0.34

0.35

0.37

0.22

0.23

0.24

0.44

0.45

0.47

0.48

0.49

0.51

0.34

0.35

0.37

0.22

0.23

0.24

0.44

0.45

0.47

0.48

0.49

0.51

0.34

0.35

0.37

0.22

0.23

0.24

0.44

0.45

0.47

Na, ppm

0.22

1.39

0.23

1.45

1.20

1.23

1.28

1.36

1.39

0.34

0.35

0.64

0.66

0.68

0.22

0.22

0.34 0.36

0.36

0.88

0.66 0.37

0.90

0.94

0.64

0.66

0.68

0.67

0.70

1.20

1.23

1.28

0.22

0.23 1.36

0.34 1.45

0.34

0.36

0.88

0.90

0.94

0.64

0.66

0.68

0.22

0.22

0.23

0.34

0.34

0.36

0.88

0.90

0.94

Ca, ppm

0.35

0.36

0.37

0.66

0.67

0.70

1.20

1.23

1.28

1.36

1.39

1.45

0.35

0.36

0.37

0.66

0.67

0.70

Ni, ppm

1.04

1.79

1.86

1.88

1.90

0.83

0.82

0.80

1.78

1.40

1.41

1.43

0.39

0.37

0.35

1.04

1.04

1.05

0.71

l.l0 1.80

0.69 0.70

1.09

1.40

1.41

1.43

0.39

0.37

0.35

1.04

1.04

1.05

0.71

0.70

0.69

1.40

1.41

1.43

0.39

0.37

0.35

1.04

0.01

0.03

3.02

14.40

14.26

2.00

2.05

2.13

3.02

3.10

3.21

0.54

20.00 0.03

0.02

0.02

0.02

0.02

0.02

11.20

10.98

10.65

14.40

14.26

14.05

20.00

0.01 0.01

20.00

0.01

16.80

16.72

0.01

10.98

10.65

14.40

1.21

0.56

5.10

9.84

10.20

4.00

4.10

4.25

1.60

1.64

1.70

4.80

4.92

1.60

1.64

1.70

4.80

4.92

5.10

9.60

9.84

10.20

4.00

4.10

4.25

1.60

1.64

1.70

4.80

4.92

9.60

1.25

0.58

9.84 9.60

5.10

16.60 0.01

4.00 10.20

14.05

0.01 l.l8

4.25 4.10

20.00

20.00

20.00

16.80

16.72

16.60

11.20

10.98

10.65

11.20

0.02

0.02

0.02

0.02

0.03

0.01

0.01

0.01

0.01

0.01

0.01

0.02

0.02

0.02

0.02

14.05 14.26

0.02

2.00

2.05

2.13

3.02

3.10

3.21

0.54

0.56

0.58

1.18

1.21

1.25

2.00

2.05

2.13

0.02

3.21

20.00 20.00

0.01

16.80 20.00

0.01

3.10

16.60 16.72

0.01

0.01

0.01

0.54

0.56

0.58

1.18

1.05

0.71

1.25 1.21

0.70

0.69

94.00 20.00

94.00 18.00

94.00 15.00

94.00 20.00

94.00 18.00

94.00 15.00

94.00 20.00

94.00 18.00

94.00 15.00

94.00 20.00

94.00 18.00

94.00 15.00

63.00 20.00

63.00 18.00

63.00 15.00

63.00 20.00

63.00 18.00

63.00 15.00

63.00 20.00

63.00 18.00

63.00 15.00

63.00 20.00

63.00 18.00

63.00 15.00

45.00 20.00

45.00 18.00

45.00 15.00

45.00 20.00

45.00 18.00

45.00 15.00

45.00 20.00

45.00 18.00

45.00 15.00

45.00 20.00

45.00 18.00

45.00 15.00

%. of Moisture, Pitch Si,ppm V, ppm S,ppm Ash, (Yo -63"m %. %. coke

1.10

1.86

1.88

1.90

0.83

0.82

0.80

1.78

1.79

1.80

1.10

l.l0

1.09

1.86

1.88

1.90

0.83

0.82

0.80

1.78

1.79

1.80

1.10

l.l0

1.09

Fe, ppm

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013 A MODEL FOR PREDICTING THE ELECTRICAL RESISTIVITY OF BAKED ANODES Dipankar Bhattacharyayl, Duygu Kocaefe l, Yasar Kocaefe l, Brigitte Morais2 , Marc Gagnon 2 lCentre universitaire de recherche sur l'aluminium (CURAL), University of Quebec at Chicoutimi, 555 Boulevard de l'Universite, Chicoutimi, QC, G7H 2B 1, Canada 2Aluminerie Alouette Inc.; 400, Chemin de la Pointe-Noire, Sept-lIes, QC, G4R 5M9, Canada Keywords: Electrical resistivity, Anode, Artificial neural network Tn spite of the significant contribution of carbon materials in anode resistivities, little work has been published to predict the resistivities of anodes. The focus of the present work is to develop a simple model which will take into account various process parameters and the properties of baked anode.

Abstract Carbon anodes are one of the key components of primary aluminum production. One of the desired properties of the anodes is low electrical resistivity. A proper understanding of the effect of different parameters on electrical resistivity can help produce better quality anodes. A model has been developed to predict the anode electrical resistivity. First, using the Kopelman model for the thermal conductivity of a composite material, the specific electrical resistivity was modeled for the solid part (coke/cokified pitch) assuming coke as the dispersed phase in the cokified pitch matrix. Then, the effects of the anode porosity, distribution of particles, and coke properties are incorporated into the model using an approach based on the work of Shimizu. A factor which is a function of particle size and other properties is introduced. This factor was estimated using the artificial neural network. Published data were used to validate the model.

Tn this model, an anode is treated as a composite material of calcined coke and cokified pitch. There are various models for the electrical resistivity of composite materials. Mclachlan [8] explained the electrical conductivities of composite materials in terms of percolation and Bruggeman's effective media theories. They also accounted for the shapes of the particles, namely, spherical and ellipsoidal. Ruschau et al. [9] presented the resistivity of conducting composites as a combination of a number of resistors connected in series and parallel. They considered the filler resistance as well as the constriction and tunneling resistances at the contacts of different particles.

Introduction More than half of the electrical energy input is dissipated as heat due to the electrical resistivity of various components in an electrolytic cell. The best technology uses 50% of the energy for aluminum production. Great effort is being spent to reduce these resistivities to decrease the energy consumption. Carbon anodes are an essential part of the aluminum electrolysis process. Reduction of the electrical resistivities through anodes and its components helps lower the energy dissipated as heat. Thus, losses from different parts of anode assemblies have attracted the attention of many researchers.

Sevkat [!O] applied statistical tools such as the Weibull distribution to predict the resistivity of a polymer-carbon fiber matrix. A statistical tool was employed to predict the cleavage of the fibers under stress. Chen and Chung [11] studied the effect of conducting carbon fibers and nonconducting silica fillers in a nonconducting matrix of cement on the electrical resistivity of composites. Merzouki [12] applied the Mamunya model in explaining the electrical resistivity of a polypropylene matrix filled with carbon black and acetylene black. The model was in good agreement with the experimental results when the fillers were present above the percolation threshold.

Some researchers have modeled the electrical losses in stub-anode connectors. Molenaar [1], Kandev and Fortin [2] applied a simplified thermo-electrical 3D finite element model to investigate the electrical losses in stub-anode connectors. Peterson [3] studied the variation in stub-cast iron resistance and temperature distribution in anodic connectors. Brooks and Bullough [4] investigated the variation in resistivity at stub-cast iron contact as a function of cast iron thickness.

Zhang and Wi [13] used Monte Carlo simulation to predict the effect of the distribution of short conducting fibers in a polymer matrix on the electrical resistivity. The literature review on the modeling of electrical resistivity of composites revealed that they dealt mainly with systems comprising of a conducting phase distributed in a non-conducting phase.

Andersen and Zhang [5] developed a 20 finite element model for energy consumption of an anode immersed in an electrolytic cell and studied the effect of geometry and anode-cathode distance. Adams et al. [6] showed that the resistivity of anodes depends on the pitch percentage.

The system of carbon anodes is different from those composites because it consists of conducting coke particles distributed in a conducting cokified pitch matrix. Thus, this work deals with conducting materials distributed in a conducting matrix. It also takes into account the porosity of the anode.

Chollier-Brym et al. [7] developed an equipment to measure the electrical resistivity of industrial anodes. Tn the article, they highlighted that carbon material itself is responsible for about 50% of the total voltage drop in the anode assembly. Depending on the variation of raw materials, forming and baking conditions, anode resistivities vary significantly between batches.

1195

Model The model has been developed in two parts. The first part deals with the prediction of the specific resistivity of anodes, without any pore, containing coke dispersed in a cokified pitch matrix. Kopelman model [14] for the measurement of the thermal conductivity of a composite material has been used for the prediction of specific electrical resistivity.

I

P

Pm

(4)

The volume fraction of coke particles (without porosity) dispersed in the cokitied pitch matrix of the anode has been calculated as follows. If BAD represents the baked anode density of anode, then the volume per unit mass of baked anode is IIBAD. If q> is the porosity of the anode, then the volume of the anode without porosity (volume of the solid part) per unit mass of baked anode is (1- q> )/BAD. If Wm is the weight fraction of pitch, then the weight fraction of coke becomes 1- Wm . If df is the real density of the coke, then the volume of coke per unit mass of baked anode (without pores) is (1- Wm)/df.

According to the Kopelman model, the equivalent thermal conductivity for a composite with particles dispersed in a continuous matrix is given as:

(I)

Thus, the volume fraction of coke (without pores) in the anode can be obtained by dividing the volume of coke (without pores) by the volume of anode (without pores):

where

Xt m

- 1

Q=

Kopelman model [14] is used to determine the thermal conductivity of a composite material. This model is applied for the estimation of the equivalent thermal conductivity of samples having a continuous and a dispersed phase. It is often used in food industries to measure the thermal conductivity of food items such as tortilla chips and French fries.

X2/3[If k

(3)

where

The second part deals with parameters like anode porosity, particle size distribution, hardness of coke and green anode density. This part has been handled in light of the work of Shimizu [15].

Q=

-[l-Q] Pm

1

(2)

(I-Wm)/d t (5)

(I-cp)/ BAD

k : equivalent thermal conductivity k m :thermal conductivity of the continuous phase (pitch)

where

kj

BRAD: real density of baked anode

Xf

cp

:thermal conductivity of the dispersed phase (coke)

:porosity of anode (= 1- BAD/BRAD)

Suffix m denotes pitch and f denotes coke.

:volume fraction of the dispersed phase

The real density of the baked anode was assumed to be 0.02 g/cm 3 higher than the real density of coke [17].

To develop a model for predicting the specific electrical resistance of anodes, the Kopelman model has been utilized by replacing the thermal conductivity terms by the specific electrical conductance. In this model, cokitied pitch has been assumed as the continuous phase and the calcined coke particles as the dispersed phase.

As the coke is a porous medium and the effective electrical resistivity of coke is measured including the effect of pores, Pj (which is specitic electrical resistivity of coke without pores) needs to be determined. Pi has been calculated based on the treatment of Meredith and Tobias [18]. Assuming that the electrical resistance of air in the pores is parallel to that of the carbon of coke. then:

When pitch is cokified during baking, the conductivity of pitch approaches to that of conducting graphitic carbon (specific electrical resistivity = 35 J.lQm) [16]. The average specitic electrical resistivity of calcined coke (with pores) is of the order of 500 J.lQm. The objective of the model is to show that it is possible to have an equivalent specific electrical resistivity of around 50 J.lQm in the presence of around 85% calcined coke.

Pf =1jI+(1-IjI)Pj

Pc

Pa

(6)

or

Thus, replacing the thermal conductivity terms by the corresponding specific electrical conductance terms and taking the specific electrical resistivity (p: specific electrical resistivity of only coke-pitch system in the absence of pores, Pm: specific electrical resistivity of pitch as the continuous phase, Pl specific electrical resistivity of coke as the dispersed phase) as the reciprocal of the specific electrical conductivity, the equations (I) and (2) can be modified as follows:

Pf

1

1

Pc

Pa

--(1-IjI)-

where

Pc : measured electrical resistivity of coke

1196

(7)

Table 3. Calculated values of P and [ 19]

lj/ : porosity of coke

Pa :electrical resistivity of air =

1.3 x

]022

/lnm

Following the approach proposed by Shimizu [15], the effects of parameters such as, particle size, hardness of coke, and green anode density are included through a correlation factor 't. The effect of anode porosity, q>, is also included in the calculation of the effective specific electrical resistivity, Perl, of baked anode as shown below:

pr PejJ

(8)

1- qJ

P

Sample No.

't

from the data of Chmelar

T

1

)tQ.m 81.04

0.66

2

78.60

0.65

3

77.04

0.65

4

93.89

0.46

5

90.30

0.47

6

88.03

0.48

7

81.77

0.57

8

79.27

0.58

9

77.67

0.59

where, T : a correlation factor

10

95.77

0.54

11

92.00

0.49

Pelf: effective specific electrical resistivity of baked anode

12

89.61

0.50

13

81.04

0.71

As P stands for resistivity of anode without porosity, the effect of anode porosity has been considered by dividing it by l-q>. As a rule of thumb, if q> increases Pelf should also increase. Here as q> increases, I-q> decreases, thus value of Pelf increases (Equation 8).

14

78.60

0.69 0.69

15

77.04

16

93.89

0.46

17

90.30

0.48

In the model, the effective specific electrical resistivity is determined using Equation 8. P is found using Equations 3, 5, and 7. The value of T is calculated based on the feed-forward artificial neural network with back-propagation.

18

88.03

0.50

19

81.77

0.56

20

79.27

0.56

21

77.67

0.59

22

95.77

0.52

Methodology

23

92.00

0.52

24

89.61

0.53

25

81.04

0.67

26

78.60

0.68

27

77.04

0.71

28

93.89

0.49

29

90.30

0.49

30

88.03

0.51

31

81.77

0.55

The ANN approach requires data for training the model. Part of the data published in the thesis of Chmelar [19] was used to train the ANN for 't, and the rest to validate the model. The researcher used four different types of coke and one pitch and mixed them in different proportions. He also varied the particle size of the coke and measured a number of properties of the anode. To study the effect of coke particle size, he mixed in the feed different percentages of coke having a particle size of -63 /lm. Table I, on the last page, shows the different formulations for anodes and the corresponding properties of the raw materials. Table 2, on the last page, shows the properties of the anode for each formulation. Using the values in Table 1, Pf, Xf and Q were calculated using Equations (7), (5) and (4). respectively. Then P values for all the samples were calculated using equation (3).

Pelf (1- qJ)

P The values of

't

79.27

0.56

77.67

0.58

34

95.77

0.49

35

92.00

0.51

36

89.61

0.52

A multi-layered custom feed-forward artificial neural network model with back propagation was developed to predict the values of T as the dependent parameter using the values of the HGI (Hardgrove Grindability Index) of coke and green anode apparent density and percentage of dust in the raw material as independent parameters. The network contained one input layer, two hidden layers and one output layer. The transfer functions associated with the hidden layers I and 2 were log-sigmoid and linear respectively. The network was trained using LevenbergMarquardt back-propagation algorithm available in Matlab 7.9. During training, the results of the samples 1, 5, and 8 (Table 1, gray rows) were not included; they were used only for the validation of the model. The trained network was finally used to predict the values of T corresponding to samples 1,5, and 8. The predicted values of T were used to calculate the effective electrical resistivity of the anodes using Equation (8).

From Equation (8) 't can be expressed as

T=-·:::c.·---

32 33

(9)

were calculated for all the samples assuming

Peff as the measured electrical resistivity of the anode samples using Equation (9). Table 3 shows the calculated values of P and r for all the samples from the data ofChmelar [19].

1197

small. Then with increase in HGI (35), the value of r increased signiticantly (average 0.68). With further increase in HGI (36 and 37) and porosity (20.5 and 21.1), the values of r decreased signiticantly (average 0.51 and 0.48 respectively). Thus the hardness and porosity of coke together are important controlling factors for the resistivity of anodes. The model can help find the optimum HGI and porosity of coke to produce anodes with lower electrical resistivity.

Results and discussion The model was used to predict the specific electrical resistivity data of the baked anodes. Table 3 shows the calculated values of p and T for the anode samples from the data of Chmelar [19]. It can be seen that the factor T varied within the range of 0.46 to 0.71 for the data analyzed. This variation in T can be attributed to the properties of coke such as the particle distribution and HGI. As it can be seen from Equation 8, if the value of r is small, the resistivity is small.

The resistivities of samples 1, 5, and 8 were calculated using the predicted values of r . Table 5 shows that the predictions are in good agreement with the experimental values with a percent average absolute error of 1.65.

An effort was made to analyze the influence of some parameters on the value of r . The effects of HGI, porosity of coke and percent of -63 /lm particles on the value of T are summarized in Table 4.

Conclusions The model presents a simplified approach to predict the electrical resistivity of anodes. The deviation from theoretical value has been handled using a correlation factor r which can significantly control the resistivity of anodes. Other parameters remaining the same, a smaller value of r indicates a smaller value ofresistivity. Properties of coke such as HGI and porosity significantly govern the value of r and in turn the resistivity of anode. The use of neural network to predict the value of r has been a key factor in predicting the electrical resistivity. For better predictions, more data are required to train the network. Thus with a large quantity of industrial data, this technique can playa signiticant role in the quality control of anodes.

Table 4: Study of the effect of different parameters on T HGI 34 34 34 34 34 34 34 34 34 35 35 35 35 35 35 35 35 35 36 36 36 36 36 36 36 36 36 37 37 37 37 37 37 37 37 37

particle %

Porosity, %

T

45 45 45 63 63 63 94 94 94 45 45 45 63 63 63 94 94 94 45 45 45 63 63 63 94 94 94 45 45 45 63 63 63 94 94 94

17.1 17.1 17.1 17.1 17.1 17.1 17.1 17.1 17.1 16.8 16.8 16.8 16.8 16.8 16.8 16.8 16.8 16.8 20.5 20.5 20.5 20.5 20.5 20.5 20.5 20.5 20.5 21.1 21.1 21.1 21.1 21.1 21.1 21.1 21.1 21.1

0.57 0.58 0.59 0.56 0.56 0.59 0.55 0.56 0.58 0.66 0.65 0.65 0.71 0.69 0.69 0.67 0.68 0.71 0.54 0.49 0.50 0.52 0.52 0.53 0.49 0.51 0.52 0.46 0.47 0.48 0.46 0.48 0.50 0.49 0.49 0.51

-63/Lffi

Average T

0.57

Standard deviation ofT

0.01

Table 5: Predicted values

0.68

0.51

0.02

Sample No

Calculated

Predicted

Measured value of resistivity I-lil.m

Predicted value of resistivity I-lil.m

1

0.66

0.69

77.3375

80.32

5

0.47

0.47

65.325

65.37

8

0.58

0.57

66.67

65.99

r

r

0.02

Acknowledgements

0.48

The technical and financial support of Aluminerie Alouette Inc. as well as the financial support of the National Science and Engineering Research Council of Canada (NSERC), Developpement economique Sept-TIes, the University of Quebec at Chicoutimi (UQAC), and the Foundation of the University of Quebec at Chicoutimi (FUQAC) are greatly appreciated.

0.02

References 1. D. Molenaar, K. Ding, A. Kapoor, "Development of Industrial Benchmark Finite Element Analysis Model to Study Energy Efficient Electrical Connections for Primary Aluminium Smelters", Light Metals,( 2011),985-990.

It shows that, for similar values of HGI and porosity of coke, r does not change significantly with the variation in percentage of the -63 /lm particles (see standard deviations of only 0.01, 0.02, 0.02 and 0.02 in Table 4). Thus the percentage of -63 /lm particles does not have a significant contribution to the value of T and in turn to resistivity of anode. The combination of HGI and porosity of coke have a signiticant intluence on the values of r . It was observed that for small values of HGI (34) and a medium value of porosity (17.1) of coke, the values of r (average 0.57) were

2. N. Kandev, H. Fortin, "Electrical Losses in the Stub-Anode Connection: Computer Modeling and Laboratory Characterization", Light Metals, (2009), 1061-1066. 3. R. W. Peterson. "Studies of stub to carbon voltage", Light Metals, (1978), 367-.378.

1198

4. D.G. Brooks, V.L. Bullough, "Factors in the Design of Reduction Cell Anodes", Light Metals, (1984), 961-976.

19. J. Chmelar, "Size Reduction and Specification of Granular Petrol Coke with Respect to Chemical and Physical Properties", (PhD Thesis, Norwegian University of Science and Technology, 1992).

5. D. H. Andersen, Z. L. Zhang, "Carbon Anode Modeling for Energy Savings in the Aluminium Reduction Cell", Light Metals, (2011), 1009-1014. 6. A. N. Adams. J. P. Mathews, H. H. Schobert, 'The Use of Image Analysis for the Optimization of Pre-baked Anode Formulation", Light Metals, (2002), 545-552. 7. M . .T. Chollier-Brym, D. Laroche, A. Alexandre, M. Landry, C. Simard, L. Simard, D. Ringuette, "New Method for Representative Measurement of Anode Electrical Resistance", ", Light Metals, (2012), 1299-1302. 8.D.S. McLachlan, M. Blaszkiewicz, RE. Newnham, "Electrical Resistivity of Composites", .T. Am. Ceram. Soc.,73(8) (1990), 2187-2203. 9.G. F.L. Fiuschau, S. Yoshikawa, R E. Newnham, "Resistivities of Conductive Composites", J. Appl. Phys., 72 (3) (1992), 953959. 10. E. Sevkat, J. Li, B. Liaw, F. Delale, "A Statistical Model of Electrical Resistance of Carbon Fiber Reinforced Composites under Tensile Loading", Composites Science and Technology, 68 (2008),214-2219. ll.P.W. Chen, D.D.L. Chung, "Improving the Electrical Conductivity of Composites Comprised of Short Conducting Fibers in a Nonconducting matrix: The Addition of a Nonconducting Particulate Filler", Journal of Electronic Materials, 24(1) (1995), 47 -5l. 12. A. Merzouki, N. Haddaoui, "Electrical Conductivity Modeling of Polypropylene Composites Filled with Carbon Black and Acetylene Black", ISRN Polymer Science, (2012),1-7. 13. T. Zhang, Y.B. Yi, "Monte Carlo Simulations of Effective Electrical Conductivity in Short-Fiber Composites", .T. Applied Physics, 103 (014910) (2008),1-7. 14. S. Sahin. S.G. Sumnu, "Physical Properties of Food", Springer, (2008). 15. S. Shimizu, T. Yamaguchi, Y. Fujishiro, M. Awano, "Effect of Microstructures on the Conductivity of porous Journal of the Ceramic Society of Japan, 117(8) (2009) 895-898. 16. .T. Xue, .T. Zhu, Y. Song, "Electrical Resistance of Graphitic and Graphitized Cathode Materials at Elevated Temperatures", Light Metals, (2010), 829-834. 17. D.Sulaiman, R Garg, "Use of Undercalcined Coke to Produce Baked Anodes for Aluminium Reduction Lines", Light Metals, (2012),1147-1151. 18. M. Kandula, "The Effective Thermal Conductivity of Porous Packed Beds with Uniform Spherical Particles", Journal of Porous Media, 14 (10) (2011), 919-926.

1199

Table 1: Properties of raw materials for anode samples (the shaded values were used for the prediction of resistivity)

Table 2. Properties of anode samples (the shaded values were used for the prediction of resistivity)

Pitch wt%

Real Density kg/dm 3

Specific electrical resistivity of coke

Porosity of coke %

-63)tm particle wt %

HGI of coke

9

20

2.07

487

17.1

45

34

10

15

2.086

500

20.5

45

36

11

18

2.086

500

20.5

45

No.

No.

Specific electrical resistivity of baked anode )tn.m

Green anode density, kg/dm 3

Baked anode density, kg/dm 3

9

67.5

1.489

1.41

10

79.5

1.44

1.38

36

11

67.9

1.5

1.41

68.9

1.483

1.38

12

20

2.086

500

20.5

45

36

12

13

15

2.07

490

16.8

63

35

13

86.175

1.515

1.395

14

18

2.07

490

16.8

63

35

14

76.66

1.583

1.489

15

20

2.07

490

16.8

63

35

15

76.85

1.576

1.45

16

15

2.065

473

21.1

63

37

16

65.28

1.418

1.374

17

18

2.065

473

21.1

63

37

17

64.5

1.485

1.4

18

20

2.065

473

21.1

63

37

18

66

1.471

1.378

69.1

19

15

2.07

487

17.1

63

34

19

1.5

1.39

20

18

2.07

487

17.1

63

34

20

65.7

1.571

1.42

21

20

2.07

487

17.1

63

34

21

68.5

1.544

1.395

22

15

2.086

500

20.5

63

36

22

73.8

1.498

1.42

66.9

1.585

1.5

23

18

2.086

500

20.5

63

36

23

24

20

2.086

500

20.5

63

36

24

68

1.589

1.481

25

15

2.07

490

16.8

94

35

25

79.5

1.515

1.429

26

18

2.07

490

16.8

94

35

26

74.9

1.583

1.493

27

20

2.07

490

16.8

94

35

27

77.6

1.576

1.47

28

15

2.065

473

21.1

94

37

28

71.8

1.418

1.345

29

18

2.065

473

21.1

94

37

29

67.5

1.485

1.378

68.4

1.471

1.356

30

20

2.065

473

21.1

94

37

30

31

15

2.07

487

17.1

94

34

31

66.66

1.5

1.41

32

18

2.07

487

17.1

94

34

32

64.78

1.571

1.44

33

20

2.07

487

17.1

94

34

33

65.8

1.544

1.42

70.5

1.498

1.405

34

15

2.086

500

W.5

36

34

35

18

2.086

500

W.5

36

35

65.4

1.585

1.499

36

20

2.086

500

W.5

36

36

66.4

1.589

1.475

1200

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

THE ROLE OF ELECTRODE QUALITY IN METAL PURITY Stephen J. Lindsay Alcoa, Inc.; Primary Metals; 300 N. Hall Rd. MS S-29, Alcoa, Tennessee, 37701-2516, USA Keywords: anode, impurities, properties not as apparent in this example is that a large fraction of the "process iron" contamination also has to do with anode properties.

Abstract Anode quality excursions are well known as factors that can have substantial impact upon metal purity and the ability to produce specific metal products. But, there is more to managing anode quality, as delivered to the pot rooms, than avoiding a major excursion. In this paper the author reviews the key factors from the paste plant through the rodding shop that affect the amount of impurities that are delivered to the metal. Conclusions include a summary of sub-process and design activities that can be managed to minimize the impact of impurities to primary metal production.

Visual Mass Balance for %Fe

Introduction

too/;

IYS

0.00

When it comes to the purity of metal from reduction cells the anodes play an important role in a number of ways beyond the impurities that are included in coke and pitch. From aggregate formulation, through mixing, forming, baking and rod ding there are multiple factors that can affect metal purity especially with regard to iron and silicon content of aluminum.

0,05

0,10

0.15

0.20

0.26

030

0.35

0.40

0.45

(),50

Figure 1 - Visual mass balance for Fe for a pot line The majority of process iron comes from the castings that are used to make the mechanical and electrical connections between stubs and anodes. Tracking the amount of iron consumed each month in the rodding shop helps to illustrate the magnitude of this area of opportunity. Many tons are typically consumed annually.

Anode-to-anode consistency as assembled and delivered to the pot rooms is the key. This evokes many details in the realm of work practices and process control. Some of these may not have an obvious connection to metal purity. There are also factors that have to do with maintenance of equipment and the design of systems to separate and to isolate impurities. There may also be design considerations for the anodes themselves.

Not all of this iron consumption is due to anode related factors. Attack by liquid bath and other factors also account for a large portion of "process" contamination. However, when the average thickness of anode butts change, or the variability in anode butt thickness increases, or air burning on butts increases one must examine the contributions that come from anode sub-processes.

In order to manage any parameter it first must be measured. Data systems must do more than be able to only track trends. Metrics on variation can be just as important, and even more so. This may require gathering and presenting information differently.

Anode Assembly Quality - It's a matter of millimeters when it comes to spent anode thickness and the corrosive effect of bath on iron castings and steel stubs. Refer to figure #2.

There are also some physical properties of the anodes that can have greater impact on iron levels of metal than the chemical properties of the anodes.

B,utt Thickness vs. Fe in Meted

The ambition of this paper is to touch upon the variety of anode related factors that can affect impurities in metal. Above average results in the pot lines and at the cast house rely upon above average attention to many of these details in the electrode plant.

720

!E: no I/O!!

----

ilisO •.........................,........................ f·····························",-"cc·········· ,....................

Discussion

•....................

•

.5 670

If

Raw Materials - It is understandable that those concerned with metal purity will focus on the levels of impurities found in anode coke and pitch. They affect the metal produced from every reduction cell.

boO

__

b:'V •..............•.................., .....................•..................•..................•....................

200

210

220

240

260

Ov"rallBult Thkk,,,,,.!mm)

Figure 2 - Median Fe in metal vs. anode butt thickness When we examine the mass balance around iron in metal it is apparent that a significant fraction comes from the impurities contained in the anodes. Figure # 1 shows an example in which 12% of all Fe in pot room metal comes from the anodes. What is

A change of a centimeter or more of anode butt thickness can be profound in its impact on typical iron levels in metal. This is especially true when the overall thickness is less than 175 mm.

1201

With modern reduction cells and large, multi-stub anode assemblies the perpendicularity of the anode rod to the top surface of the anode becomes more important that it was in the days of small, single stub, anode assemblies [1]. A deviation of 1 degree away from being perfectly square from the base of an anode that is 1.5 meters long means that one end of the anode will be 26 mm lower in the bath than the other end. The outer edge of the iron casting on one stub may be 2 cm lower than the elevation of the casting on the opposite end of the same anode assembly. Nevertheless the risk of attack on cast iron by bath is amplified almost as much as if the spent anode butt were to be 2 cm thinner.

Baked Anode Density

01,

,1

1,$70'

U6S

us. uso

•

u •• 1,$40

iSiS

1.$1. U10 UIS

1.>10

Deviations of this magnitude from may seem extreme. But, when this parameter is not routinely measured or controlled it is not unusual to find ::::15% or anode rods are 1 degree or more out of perpendicular to the top surface of the anodes. Refer to figure #3.

III OIl

J!!c

III

"....

If.

III

.::

1:::I

E

100%, 90%

Figure 4 - Example of variation of baked anode density A I % change in the density of the anodes or other factors that affect the rate of consumption can change the thickness of the anode butts by 4 to 5 mm. Figure #4 shows that changes of> I % are not necessarily uncommon.

Anode Assem bly Perpendicularity Deviation

Likewise the reactivity of the anodes in CO 2 or in air can be of concern as can factors that are related to the degree of anode baking such as Real Density or Le. All of these can change by more than a few percentage points over periods of months and all can affect individual anode butt thickness.

80%

60% 50% 40%

In real world situations anyone factor, or a combination of factors may change by significant amounts. The anode setting cycle in the pot lines may not be changed in response. Ergo, metal purity becomes a response variable in such a situation. But, following shifts in average properties is only the half of it.

.30% 20%

:::I

U

o

10

20

30

4G

Deviation in millimeters

While average spent anode thickness matters, the thinnest anode butts and those that have exposed parts of castings via air burning will contribute the most to iron contamination. The amount of variation in key properties can be more significant than changes of a few millimeters in average anode butt thickness.

Figure 3 - Example of deviation of assemblies from perpendicular Perpendicularity has been cited as a quality reference in order to illustrate that attention must be paid to more than just the anode blocks themselves. Many factors are important when it comes to metal purity. Electrode sub-processes including the assembly of the anode to the rod require attention to detail. This avoids adding root causes for iron to come into contact with bath and metal.

Variation in Anode Properties - The uniformity of anode butt thickness is a parameter that is not overly difficult to measure. But, it is often over-looked. It can tell a lot about variation in anode sub-processes and in pot room operating conditions.

There are other obvious iron contamination concerns that have to do with anode assemblies. These include iron over-pours or splatter on the bonnets of anodes. These typically impact Fe in anode cover unless 100% are removed, usually by manual means. A thin "skin" of iron may also form during pouring against the steel stubs to till the stub holes. This "skin" often ends up oxidizing and also finds its way into the pots via anode covering material. Then there is cast iron composition itself. It carries more than just Fe into the metal. Silicon, manganese, and phosphorus will also enter when cast iron is dissolved by bath.

Figure 5 - Examples of variation in anode butt thickness

Anode Properties - Even if anode surfaces are perpendicular to the anode rods the average thickness of the butts may vary. When metal purity is determined by a matter of millimeters a number of anode properties and the process control parameters that affect them come into consideration.

Figure #5 tells a story. These are butts taken from the same pot line under the same operating conditions. The anodes were from two different suppliers. The type A butt thicknesses are obviously not uniformly distributed. The range of thicknesses is +/-3 cm around the mean. The type B butts are much more uniform in thickness and encompass a range of +/-1 cm in around the mean.

The baked density of anodes is an obvious concern when average butt thickness is a surrogate for metal purity. See figure #4.

Type A anodes came from an older anode plant that did not sell anodes to third parties. Type B anodes came from a new and well controlled facility that only sold anodes to third parties.

1202

In the case study of Type A anodes there were many known issues with; coke blending, aggregate control, mixing and anode baking. The anode butt thickness distribution was multi-modal since many things were going on that all had some impact on anode consumption rates. The take-away may appear to be that purchased anodes are better than anodes that do not have to be sold to third parties. Type CAnode Butts

The starting point for inquiry is close examination of the process data, examples of which are shown in figure #7. The parameter may be density, reactivity residue, or a host of others that relate to green and baked anode manufacturing. Study of the entire distribution for any parameter brings the right questions into focus. Is the distribution bi-modal, or multi-modal? Why? What factors cause the upper and lower tails of the distributions to be as they are? What is necessary to obtain more uniform results, and more uniform spent anode butt thicknesses? The examples shown in figure #7 are real. The impact upon butt thicknesses and metal purity is left to the reader's imagination.

Type D Anode Butts gO

§40i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 20

ButtThiekness (em)

Unfortunately, conclusive links between specific anode properties and butt thickness variation metrics are left to future research.

Butt Thickne .. (em)

What does not have to be left for future consideration is that parallels in the fraction of anode butts that are; quite mealy, soft, or thin have been made to the fraction of baked or green anodes that have had less than desirable characteristics.

Figure 6 - Other examples of variation in anode butt thickness Now refer to figure #6. This location makes two types of anodes for in-house use only. There are no sales to third parties. The facilities are not new, and they are well maintained. Process control is held to the highest standards through the paste plant and the anode baking furnaces. The type C and type D butts respectively encompass ranges of +/-2 and +/-4 mm around the mean values for anode butt thickness.

The conclusion that begs additional research is that variations in specific anode properties do cause variations in anode butt thickness and metal purity. This is particularly so with regard to iron contamination. Spent Anode Butts - Butts represent a significant fraction of the most anode aggregates. But, they are seldom required to pass the demanding standards that maybe set forth for other anode raw materials. This may be particularly so when they are produced and used in-house.

This location also has excellent and very predictable results with metal purity, especially with regard to iron content in metal. These results are not surprising. The distributions for Type C and D anode butts are an order of magnitude tighter than for Type A butts. It is not uncommon in our industry to tind that the range of anode butt thickness is +/-2.5 cm around the mean value. There is plenty of opportunity for improvement in this area with benefits well beyond metal purity.

It is important to acknowledge that the content of Na, Ca, Fe, Si,

ash and other impurities are higher in butts than in new anode assemblies. Such impurities tend to greatly concentrate in the finer butts fractions, especially in dust that is gathered during crushing of butts. Some, such as sodium and calcium also affect the reactivity of new anodes. Ultimately this affects the thickness and amount of air-burning on the next generation of anode butts.

Baked Anode Density

As shown in figure #8 the iron level of butts is typically higher than that of new anode assemblies. However, the Fe content of fine butts is much higher. This is due to a number of potential contributing factors. Iron Coneentration in Anode Fraetions 1.SS

1.56

1.6

1.6Z

1.64

3000

1.66

Baked Anode Density {gm!cm 3 1

2500

2000

CO 2 Reactivity Residue

IE 1500 Q. Q.

1000 500

Fine Butts Butts Avg.

00

a

COatS6 Butts

Baked Anod$s

Coke

Pilch

Figure 8 - Tron concentration in anode & butt fractions M

00

•

•

•

•

COl Reactivity Residue {% of sample}

Fine butts will contain bits of scale from the points of contact between the iron castings and the anodes. Tron sulfide and some forms of iron oxide formed under high temperature are not

Figure 7 - Cumulative distribution examples of anode properties

1203

magnetic. Fortunately, these are often physically attached to particles of magnetic oxide or metal. This scale is generally quite friable and can easily be broken into small particles or even into dust. These can be quite difficult for magnetic separators to remove. With a "good" level of efficiency in removal of magnetic material being 70%, multiple passes of magnetic separation may be required to access this hard-to-get material.

direct financial penalties, such as those applied by independent anode producers. The financial penalties will be inherent to damage caused to anode quality, the baking furnace, and even to metal purity. Thus, specifications for butts should be set and adhered to with rigor. It is one of the few components of anode impurities that manufacturers will have under their direct control going forward as properties of anode coke continue to decline.

Butts are often cleaned with steel shot blast of 0.4 to 1.3 mm in diameter. A sphere is the most difficult shape for magnetic separator effectiveness. Steel spheres of low mass and small diameter are particularly difficult to remove from a stream of crushed butts.

One option that can help to significantly reduce the level of impurities in next generation anodes is to cull a certain fraction of the fine butts material [2]. This should not be used as a substitute for proper removal of shot or magnetic separation. But, at some level the very fine impurities that can not be easily removed by other means can be separated and culled. This may be limited to dust collector material. In some cases low-value, but usable endmarkets have been found for butts fines fractions avoiding need for land-filling.

The size, or perimeter length, of castings also figure in to the steel shot equation. The voltage drop between the casting and the anode generates heat. The casting also conducts heat upwards when anode butts become thin. Conditions become favorable to air burning of a groove around the top perimeter of the castings. Without exceptionally good equipment and procedures to recover steel shot that comes to rest in these grooves the iron level of the fine butts fraction is likely to be negatively impacted. Refer to figure #9.

There are also noteworthy concerns about silicon contamination. Locations that use delayed coke to pack anodes in the baking furnaces may also recover some of this coke to anode production. This can be a large mistake if there are metal product quality concerns that are related to Si. The refractory material that flakes off refractory, mortar and insulation accumulates in the anode packing material. It normally should not be used for anode production. Likewise there are small streams of fine coke that should never be used, such as the dust collector catch on anode bake tending cranes. As with butts, impurities tend to migrate to fine material. Dust collector catch has been measured at up to 2% Si by weight. The Ball Mill - As noted above, butts and especially butts fines can be a significant contributor to contamination of Fe and Si to baked anode production. There is another contributor that is often at par with the contribution of butts that quite often goes unnoticed. Steel balls and the steel liners of most ball mills often contribute> 120 ppm Fe to the total iron found in baked anodes. Credible mass balance studies made internally by Alcoa and ppm of the Fe Elkem have found that it is quite common for found in pot room metal to have originated in the ball mills that are used to grind coke. Roughly 60% of this sum comes from the erosion of the steel balls themselves. The remaining 40% comes from the steel liners of the mill. This is presumed to vary somewhat with the grade of steel used to make the balls, the fineness of the coke grind, and the percentage of fines that are used in the dry aggregate formulation.

Figure 9 - Shot blast trapped in air burn grooves around castings The other factor of concern that is also related to the perimeter of the castings is the total surface area of each casting. As flutes and extremities such as wings are added, the total area available to form oxide and sulfide scale increases. With each new cycle of anode casting thin layers of oxides and sulfides will form. The amount is in direct proportion to a few factors including the total surface area of the castings. Some fraction of this non-magnetic and mildly magnetic material then finds its way into the butts fraction, usually as fines.

There have been some attempts in industry to capture and remove some of this very finely ground metal. But, no standard magnetic separation equipment appears to be effective. Most attempts have fallen below 10% in removal efficiency. If we consider a paste plant that uses coke with 150 ppm Fe in it and anode butts that include most of the butts fines or dust we could easily find that 33% of Fe in anodes comes from coke and pitch, 33% comes from butts, and the balance is from the contribution of the ball mill.

Let's return to the concept of butts as a raw material that is often generated in-house. As a raw material the various fractions of butts should have certain specifications. These might include levels of; Na, Ca, Fe, Si, and ash in addition to characteristics on sizing. Control of deviations will not have the convenience of

What might be done about this? The use of rubber liners for ball milling can eliminate 40% of this contributing factor. Some have

1204

reported on this technology in the literature [3]. The technology does face some challenges for it to be successful. But, it holds the potential to immediately reduce Fe to pot room metal by ppm.

eliminated and common causes of variation to be better controlled using common statistical process control methodology. An external focus on raw materials is necessary to control the input rate of impurities to pot room metal. But, many producers commonly over-look the potential that an internal focus can also reveal. Examples for managing to rigorous specifications for recycled anode butts and placing some focus on control of the iron contamination from ball milling or anode packing materials management have been given.

Counter-Measures to Impurities - Some counter-measures are obvious and have already been specifically discussed or referred to in sections ofthis paper. In the Green Mill - Uniform aggregate control, thorough mixing and cooling of paste, uniform filling of mold boxes, rubber ball mill liners

The overall summary is that Electrode Quality is much more important to metal product purity and quality than is commonly recognized. Those that have lead the way and that already excel in these areas provide hope and a pathway to those that will follow.

In the Anode Bakes - Uniform baking of each anode and each pit, uniform anode heating rates, avoid use of packing material and dust in anode manufacturing In the Rodding Shop - Perpendicularity of the anode assemblies, over-pouring of cast iron on anodes and on stubs, removal of steel shot after shot blast cabinets

References

In All Areas - Tracking of key properties and variation of key properties, efforts to understand and eliminate causes of special variations in properties, efforts to understand and reduce common causes of variation in properties.

I. Kalban, A. J. M., BinBrek, A. S. S., & Sachan, G. S. "Rod ding Room Upgrade at Dubal", Light Metals 2001, pp. 651 - 656 2. 0ye, Harald A., "Materials Used in Aluminium Smelting", Light Metals 2000, pp. 3 - 15

There are also some design factors to take into consideration when metal purity is an important outcome. These include:

3. Malcolm, Ned, Marvel, Donna & Roberts, James, "Conversion of Alcoa, Inc. Ball Mill Liners from Rubber to Steel", Light Metals 2002, pp. 90 - 93

Shaping the profile of the anode bonnet to protect castings from air-burning

Acknowledgements

The design of casting including their perimeter, total surface area, and number of flutes

- Alcoa, Inc. for support in the preparation and publication ofthis paper

The design of the rodding carousel to assure perpendicularity of anode assemblies plus adequate placement of each stub to the proper place and depth

- Merino, Dr. Margarita R. (Ph.D. - Florida State University) - for her encouragement, dedication, and support.

Entire systems to prevent attack of bath on stubs and castings may also be implemented. The use of stub protection collars is not uncommon in the industry. Producers of very high purity metal may employ wholesale spraying of the anode tops with an aluminum mantle. This can greatly reduce the risk of air-burning that might otherwise expose cast iron.

Conclusions The contribution of the Electrode Plant as a whole to impurities found in pot room metal extends well beyond the impurities that arrive in anode raw materials. The impact of butt thickness on iron contamination can be measured in millimeters of anode butt thickness. This extends to deviations that may be caused by; the perpendicularity of anode assemblies, green and baked anode factors that influence consumption rate such as density and reactivity, and the amount of air-burning that factors such as casting and anode bonnet design contribute to. Measurement of these key parameters does not only involve following of trend information. As shown in examples included in this technical paper one must also study metrics related to process variability. This allows special causes of variation to be

1205

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

ELECTROCHEMICAL CHARACTERIZATION OF CARBON ANODE PERFORMANCE Rebecca Jayne Thorne l , Camilla Sommerseth l , Espen Sandnes 2, Ole Kjos 3, Thor Anders Aarhaug3, Lorentz Petter Lossius2 , Hogne Linga2, and Arne Petter Ratvik l I

Dept. of Materials Science and Engineering, Norwegian University of Science and Technology, NO-7491 Trondheim, Norway 2Hydro Aluminium, Ardal, Norway 3STNTEF Materials and Chemistry, NO-7465 Trondheim, Norway Keywords: Anode, carbon, electrochemistry, gas analysis, characterization, impurities 'electrocatalysis' by artificially distributing dopants (metals, metal oxides and salts) throughout a coke before production into green anodes. However, artificially added dopants are probably not fully incorporated into the coke bulk structure. Fewer publications (such as [IS]) have studied the natural variation of impurities in single cokes and their effect on electrochemical performance.

Abstract Coke used in the manufacturing of anodes is commonly a mix of cokes from several suppliers to meet the customer's specifications. This variation in coke composition from suppliers may lead to considerable deviation in anode performance. The present work, still in its early phase, aims to develop a method with which to characterize electrochemical performance of anodes and relate this to the anode material properties. To verify the experimental approach, laboratory anodes were produced from various single cokes with different impurity levels. Voltammetry was performed and polarization curves were recorded to investigate current-voltage characteristics of these anodes. Gas analysis was also executed in order to study the CO 2:CO ratio and calculate Pearson-Waddington current efficiencies. The reaction overpotential from polarization curves was found to decrease with increasing total metallic and sulphur impurities, indicating that blended cokes may behave differently on a microscopic scale and between individual anodes in a cell if the anodes come from different production batches. Contrary to the polarization curves, metallic and sulphur impurities were found to not significantly change the CO 2 :CO ratio or Pearson Waddington current efficiency. Experiments of this type aim to develop fundamental understanding of how single coke properties affect electrochemical performance.

Anodes made from single cokes may give a better elementary understanding of the coke properties with the most pronounced effect on the overall anode performance. This study attempts to develop a method with which to distinguish the electrochemical performance of anodes made from different single cokes, and correlate this to anode properties such as impurities. The method is based on determining reaction overpotential; an important factor approximating to the energy required for the anodic reaction to occur. Although difficult to separate out from the measured anodic potential, the anodic reaction overpotential should be the only variable between experiments (as detailed in Figure I) when comparing potentials at the same current density (CD). This means comparative values of reaction overpotential between a series of materials can be given. Voltammetry and gas analysis are also used to support and explain the electrochemical results, and provide a measure of the CE using the PearsonWaddington (P-W) Equation [16].

Introduction CE(%) = 100% - 0.5 [%CO(g)] Anodes are traditionally made by using a blend of various single source cokes [1, 2]. However, although the blends from the coke supplier are within certain analytical specifications, anode quality can still vary widely [3]. Reasons for this are not fully understood and more fundamental studies are required.

Anode potential

50% + 0.5 [%C0 2(g)]

Concentration Dverpotentlal

CONSTANT AT GIVEN CD

1

Reversible potential for reaction CONSTANT

[1]

Voltage

rj,o/iR

MEA\

Many papers have therefore studied how changes in coke properties, such as structure and density, affect anode performance [4-6]. The effects of coke impurities on anode performance have additionally been widely studied: iron, vanadium, calcium and sodium are found to enhance the combustion of carbon in air and its reaction with CO 2 [4, 7-9]. Studies indicate these impurities can also affect the anodic reaction, causing increased electrolytic anode consumption and/or decreased overpotential [8, 10-13]. Sulphur acts as an inhibitor of these catalysts possibly due to metal-sulphide formation [8], and has therefore been shown to retard CO 2 and air reactivity of anodes [9]. It has additionally been associated with lowering current efficiency (CE) [14] and varying degrees of anode dusting [8, 14]. Most researchers have investigated this field of

=

drops

/CONTROL

Reaction Dverpotential

WANT TO INVESTIGATE

Figure 1) The measured potential (Ej ) is composed of numerous terms, including the reversible potential (E rev ), concentration overpotential (l1c), reaction overpotential (l1a) and voltage drop (iR)

1207

Each anode was lowered into the melt to a depth of 1.5 cm (although accurate immersed areas were measured after experiments), giving an approximate active electrode area of 5.5 cm 2 . Electrochemical Impedance Spectroscopy (EIS) was used to determine the ohmic resistance at the Open Circuit Potential (OCP), the value of which was used to iR compensate all electrochemical measurements during experiments. Although the additional resistance with current flow (for example, bubble resistance) was not compensated, this would be approximately the same for each material at any given CD. Linear sweep voltammograms were produced at 1 V S·I before one polarization curve was recorded for each anode by slowly sweeping the anode potential (relative to the aluminium reference electrode) from the OCP to 2.6 V at 2 mV S·l, and measuring the responding current. All electrochemistry was performed using an Autolab PGSTAT 20 (with built in Frequency Response Analyzer (FRA) and 10 A booster, all from Eco Chemie)

Materials and Method Experiments were performed in a cryolite melt (cryolite ratio = 2.3 (cryolite from Sigma Aldrich, purity >97%), excess AlF3 = 9.8 wt% (industrial grade AlF3 sublimed in-house), alumina concentration = 9.4 wt %, (y alumina from Merck). The melt was contained in a graphite crucible (Svensk Specialgrafit AB, Sweden), in which alumina crucible shields and a copper-plate cathode were inserted (schematics in Figure 2). An aluminium reference electrode in an alumina assembly was fabricated according to [17]. Four types of single cokes were used to produce pilot scale anodes using fraction size 2-0 mm. The production of the anodes varied only in the coke type, all other parameters were kept constant. Anodes were labeled 1-4, core drilled into cylindrical rods with dimensions 60 x 10 mm, and screwed into stainless steel rods. Parallels of anodes types 1-4 were additionally analyzed using XRF to determine the impurity content (including sulphur and total metallic elements).

a)

Stainless steel rod for anode contact AI reference (in alumina assembly)

Alumina spacer

To relate the polarization curves/voltammetry to reaction products, mass spectrometry (MS, ProteaProMass) was used in combination with gas chromatography (GC, Agilent IlGC series 3000) to measure CO and CO 2 content in the gas outlet, based on an adapted method from Kjos et al [17]. GC and MS techniques had different sampling rates and varying degrees of accuracy; MS has a low sampling time and volume but drift due to vacuum renders data less quantifiable. GC produces quantifiable data, although collection is slow due to elution times of four minutes. Thus, to determine concentrations, the slow throughput (but accurate) GC data was used to continuously calibrate the high throughput (less accurate) MS data.

outlet steel pipe) lid

--I-----'M

Alumina crucible lid

The order of the anode materials tested was randomized to eliminate possible changing characteristics of the melt over time, and two parallels were performed for each material (with new anodes) in the same bath. Bath samples after each anode experiment were taken for subsequent oxide content analysis (LECO analyzer model TC-436DR).

shields Alumina radiation

Furnace base Argon inlet -

rod for cathode contact

Results and Discussion

b)

Electrochemical measurements The measured polarization curves showed a steep increase in current relating to the main anodic reaction during aluminium electrolysis where CO 2 (and possibly some CO) is formed (see Figure 3 for a typical polarization curve). Due to the relatively large reaction overpotential involved, the potential where a large increase in current occurred was approximately 300 mV higher than the standard potential for CO 2 formation. No diffusion limitations were observed in the potential range studied, probably due to the fact that the melt was saturated with alumina. The produced polarization curves were used to get a comparative value of the reaction overpotential of each anode type (containing different cokes). LECO measurements of all the melt samples showed little variation in alumina concentration over the course of the experiments, meaning all potential differences could be considered due to the materials only. Owing to the fact that all anodes had identical distribution of grain size and pitch type/level, these variations in potential must relate to differences in the coke properties such as impurity levels.

_ _ _ _ _ AI reference (in alumina assembly)

III-_++_ _ _ _ _ _'>'"'in"Ie sssteel rod for anode contact

Alumina crucible lid

. - - - Alumina shielding

.00000-+-t---l---l!IIi-------l!1Ii-----

Anode Electrolyte Graphite crucible

- - - Copper plate (cathode)

0--------- Graphite contact block L------==mF=------ Steel rod for cathode contact Figure 2a) The location of the crucible inside the furnace tube, and Figure 2b) A close up of the crucible and its contents (gas outlet removed for simplicity), detailing how potential is applied to the anode and current measured

1208

The CD chosen for comparison was 1 A cm- 2 , as this is close to industrial conditions. Comparison of the potential at 1 A cm-2 in Table I showed a standard deviation (STDEV) of 0-50 mV for repeats of the same material, with reaction overpotentials between different materials varying by 40-200 mY. As the experimental STDEV was lower than the average material differences, resolution of reaction overpotential trends could be observed between the anodes. The order of reaction overpotential of the materials (high to low) correlated with increasing levels of total metallic impurities and sulphur. As metallic impurities can catalyze air and CO 2 reactivity, it would also be expected these can additionally catalyze the anode reaction and reduce anode overpotential [10]. This trend correlates with previous studies which showed that metallic dopants such as vanadium, calcium, sodium and iron could lower anodic overpotential and increase electrolytic anode consumption [8, 12, 13]. Although the order of overpotential also correlated with increasing sulphur, a known catalyst inhibitor [8, 9], it is possible that the sulphur in the anodes requires time to accumulate at the surface to act as reaction inhibitor. Additionally, some studies showing sulphur as a catalyst inhibitor used artificially doped sulphur, with most sulphur residing in the binder coke - a different situation than with naturally occurring coke sulphur [8]. However, due to the fact that many other properties vary between coke types, further important factors towards overpotential could include surface morphology (porosity) and structural composition of the coke; both not discussed in this paper. m

Similar to the polarization curves, the main feature on the linear sweep voltammograrns was the large increase in current relating to the main anodic reaction of COziCO formation (Figure 4). Additionally, there was a smaller peak at lower potentials which, although difficult to identify, has previously been related to adsorption [18]. Since the electrolyte was saturated with alumina, the sweep rate dependency of the current response was very limited. With no mass transport limitation, the system showed kinetic control. There were therefore similar correlations between the reaction overpotentiallCD and the anode types as found from analyzing the polarization curves. For example anode I (with lowest metallic and sulphur impurities) had the lowest CD from voltammetry and highest reaction overpotential from polarization curves, whilst anode 4 (with highest sulphur and metallic impurities) had the highest CD and lowest reaction overpotential.

30

5

i$ ."

I Q

H 2.0 15 H)

M M .()s

It'ii!liib!e:

o.a

1$

H

Lt

;$

10

22

2A

26

pta,oo

Figure 4) Linear sweep voltammetry curves of the anode materials tested, showing variations in CD.

10

5

2,S;

::";

'""'"'" " 15

Gas analysis Gas analysis showed that the main anodic product was CO 2, but substantial amounts of CO was also detected. CO arises mostly from the back reaction, but could also have been formed electrochemically or via the Boudouard reaction. As expected, the concentration of CO and CO 2 started to increase exactly where the current started to rise rapidly (Figure 5). The CO2 :CO ratio and p- W CE were found to increase with potential (Figure 6), possibly due to a decrease in the retention time of the produced CO 2 in the melt, leading to a decrease in back reaction.

."

"

Q

M

4.5

qa

to P"(,,nttell V VIX.! AI

Figure 3) A typical polarization curve showing a scan from OCP to 2.6 V w.r.t AI. Detailed are the standard potential for CO 2 formation and the initiation of gas evolution. Table I) Average anode potential and STDEV/ V w.r.t Al (at I A cm- 2) and anode impurities. Material

Anode I Anode 2 Anode 3 Anode 4

Average potential IV

Order of anodic overpotential (high to low)

Content of impurities (low to high) Total metallic S impurities

1.94 ± 0.00 1.90 ± 0.03 1.84 ± 0.00 1.74± 0.05

P"wnnall V wxl AI

Figure 5) A typical polarization curve with measured CO and CO 2 concentrations over the potential range studied.

1209

Generally, all anodes produced similar gas ratios and PW C-E, although the anode with the lowest sulphur content produced a lower CO 2 :CO ratio than the others. Understanding the relationship between single coke properties, physical anode properties and electrochemical characteristics (such as overpotential) is crucial to avoid issues resulting from blending. Future work in this project will continue to strive to develop a better fundamental understanding of how coke properties affect the anode performance.

Acknowledgements Potentia! i V wxt AI

This work was financed by Hydro Aluminium and the Research Council of Norway. Thanks are due to Aksel Alstad at the NTNU workshop where fabrication of experimental parts was required, as well as to Hydro Aluminium lab engineers at Ardalstangen for performing XRF testing. Additionally, the contributions of Egil Skybakmoen and others at SlNTEF are gratefully acknowledged

Figure 6) Typical CO 2 :CO ratios and P-W CE over the potential range studied. Generally, all materials produced similar amounts of CO 2 and CO at I A cm-2 , giving similar P-W current efficiencies. The concentration of CO 2 was usually around 6000 ppm (at I A cm-2 and constant carrier gas flow), with a theoretical value of 9000 ppm. Using the P-W Equation and assuming all CO is produced from the back reaction, the CE was calculated as 90% for the materials (Table 2). In contrast, previous studies have indicated that high levels of metallic impurities such as vanadium lower the CE, possibly due to the presence of multivalent oxides and higher non-electrolytic anode consumption [IS]. The fact that CE did not correlate with metal impurities in the current study could be due to the scale of the experiments, i.e., laboratory vs. industrial. Interestingly, anode 1 (with the lowest sulphur content) produced gases with a smaller CO 2 :CO ratio than the other anodes tested. Previous studies have found increasing sulphur increases dusting [8]; others found that levels of Boudouard reaction, dusting, and current efficiency do not change significantly within a range of 23.8% anode sulphur content [14]. Discrepancies are not easy to explain, but indicate a complex influence of impurities on the anode process, difficult to comment on at this stage.

References 1. 2. 3.

4. 5.

Table 2) A summary of the gas ratio and P-W CE for each anode type (at I A cm-2) and STDEV

6.

Average P-W CE / %

7.

Material Anode Anode Anode Anode

1 2 3 4

Average ratio CO 2 :CO 3.0 ± 0.3 4.1 ± 0.0 4.3 ± 0.0 4.4 ± 0.2

87.6 ± 90.3 ± 90.6 ± 90.5 ±

0.9 0.0 0.0 0.6

8.

9. Conclusions 10.

In contrast to studies where dopants are artificially added to raw cokes or cokes are blended to achieve differing impurity levels, this study investigated four anodes containing a single coke type (all other parameters constant). An electrochemical method was then used successfully to obtain comparative measurements of the reaction overpotential of each anode, which were related to anode impurities. Results indicated that the anodic reaction overpotential varied by 200 m V between the materials tested, and decreased with increasing total metallic impurities and sulphur levels. Whilst metallic impurities are known to exhibit electrocatalytic behaviour, sulphur is a known catalyst inhibitor.

II. 12. 13.

1210

Edwards, L., Responding to Changes in Coke Quality, in Presented at Australian Aluminum Smelting Technology Workshop. 2007. Edwards, L., Backhouse, N., Darmstadt, H., and Dion, M.l, Evolution of Anode Grade Coke Quality Light Metals, 2012. Edwards, L.C., Neyrey, K.1., and Lossius, L.P., A review of Coke and Anode DesulJurization. Light Metals, 2007. Leach, C.T., Brooks, D.G., and Gehlbach, R.E., Correlation of Coke Properties, Anode Properties and Carbon Consumption. Light Metals, 1997. McClung, M. and Ross, l.A., A Method to Correlate Raw Material Properties to Baked Anode Core Performance. Light Metals, 2000. Wilkening, S., Maintaining Consistent Anode Density Using Varying Carbon Raw Materials. Light Metals, 2009. Rolle, lG. and Hoang, Y.K., Studies of Vanadium and Sodium on the Air Reactivity of Coke and Anodes. Light Metals, 1995. Kuang, H., Thonstad, 1., and S0flie, M., Effects of Additives on the Electrolytic Consumption of Carbon Anodes in Aluminium Electrolysis. Carbon, 1995. 33(10). S0flie, M. and Eidet, T., The Influence of Pitch Impurity Content on Reactivity of Binder Coke in Anodes, in Light Metals. 1998. p. 763-768. Thonstad, l, Fellner, P., Haarberg, G.M., Hives, l, Kvande, H., and Sterten, A., Aluminium Electrolysis 3rd edition ed. 2001, Breuerdruck, Germany: AluminiumVerlag Marketing & Kommunikation GmbH. Feng, N., Zhang, M., Grjotheim, K., and Kvande, H., Carbon, 1991. 29 (I): p. 39-42. Liu, Y., Wang, X., Huang, Y., Yang, lH., and Wang, W., in Light Metals. 1995. p. 247-251. Haarberg, G.M., Solli, L.N., and Sterten, A. in 7th Slovak Aluminium Symp. 1993. Bankska Bystrica.

14.

Pietrzyk, S. and Thonstad, J., Influence of the Sulphur

Content in The Carbon Anodes in Aluminium Electrolysis - A Laboratory Study, in Light Metals. 15.

2012. p. 661-667. Schmidt-Hatting, W., Perruchoud, R., and Durgnat, lE.,

16.

Influence of Vanadium on Anode Quality and Pot Performance. Pearson, T.G. and Waddington, 1., Discussions. Faraday

17.

18.

Soc, 1947: p. 307-320. Kjos, O.S., Aarhaug, T.A., Gudbrandsen, H., Solheim, A., and Skybakmoen, E., Fundamental studies of Perfluorocarbon Formation, in 10 AASTC 2011: Tasmania. Jarek, S. and Thonstad, l, Voltammetric Study of

Anodic Adsorption Phenomena on Graphite in CryoliteAlumina Melts. Electrochemical Science and Technology, 1987. 134(4).

1211

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

HIGH CAPACITY THERMO BALANCE ANODE REACTIVITY TESTING Nick Janssen l , James Baker l , Frank Cannova l , Dr. Barry Sadler2 i BP, 5761 McFadden Avenue, Huntington Beach, CA, USA 2Net Carbon Consulting Pty Ltd, PO Box 286, Kangaroo Ground, Victoria 3097, Australia Keywords: Carbon Anode, Air Reactivity, Thermogravimetric analysis, TGA will significantly impact results unless the test is conducted at temperatures below 400°C; however reaction rates at these temperatures are too slow for practical reactivity testing.

Abstract As raw material quality changes and Potline Customer requirements become more onerous, anode reactivity has become an increasingly important quality measure. Conventional anode reactivity testing procedures require dedicated and expensive instruments that have relatively low sample throughputs. Smelters are often unable to test the reactivity of all the core samples they take due to reactivity test capacity imitations. This reduces the ability of plants to identify any changes in anode reactivity that may require countermeasures. A commercially available, multiple sample ThermoGravimetric Analyzer (TGA) has been successfully adapted to measure anode air reactivity. This paper describes the modifications made to the instrument and the standardization of a procedure for anode air reactivity testing. Results obtained from plant anode samples baked under different conditions are also discussed.

These complexities mean that all air reactivity tests are somewhat of a compromise and they do not measure the real chemical reactivity of the anode carbon to air, e.g. the most common temperature range for air reactivity tests is 500 - 550°C [4]; which is in the Zone IIIZone III transition range of Figure 1 and hence mass transport (e.g. in-pore diffusion) and not chemical reactivity, is the dominant factor impacting the reaction rate in this temperature range. one III

Introduction Anode grade petroleum coke quality has changed significantly in the last 10 years and this trend is expected to continue [1,2]. These changes have included an increase in catalytic impurities such as Vanadium, which would be expected to increase the anode air reactivity. This does not appear to have had a widespread impact on anode performance to date, as in-cell airburn is controlled more by the degree of protection from air access afforded to anodes in the cell, than it is associated with anode quality [2,3]. Despite this, monitoring anode air reactivity can be important to detect changes in anode quality and signal the need for countermeasures. One of the factors that influences how quickly a change in anode quality can be detected is the number of samples tested. At present, this is limited by the relatively low throughput of conventional anode air reactivity testing equipment. This means that many plants are unable to test the air reactivity of all of the core samples they take; this represents a loss of data that can delay the identification of significant shifts in anode quality. The use of a high throughput air reactivity test could avoid this limitation.

Figure I: An schematic Arrhenius plot for the reaction between Carbon and Oxygen [from 3, P. 467], showing (log) reaction rate (Y axis) plotted against the inverse of reaction temperature (X axis - note that temperature increases to the left). In Zone I, reaction rate is controlled by the intrinsic chemical reactivity of the anode carbon. In Zone II, reaction rate is increasingly controlled by the diffusion of air through the anode pore structure, but intrinsic reactivity contributes at lower temperatures in the zone (i.e. in transition zone a). In Zone Ill, reaction rate is so fast that all oxygen is consumed immediately it reaches the anode surface, so the reaction rate is largely controlled by how fast the air reaches the anode and the anode geometry, with little contribution from other anode properties. Zone a (Transition from Zone I to II) occurs at about 400°C [4] and Zone b (Transition from Zone II to III) has been variously reported at around 500°C [3] and 800°C [4]. (See [3] for further discussion of anode air reactivity/mass transport.)

Anode air reactivity testing has a number of complexities that make the development of a "technically ideal" test (i.e. a test that measures the intrinsic chemical reactivity of the anode carbon without mass transport effects, see Figure 1) for plant use somewhat difficult: • The reaction rate is very dependent on temperature - as temperature increases, so does the reaction rate. • The reaction between Carbon and air (Oxygen) is highly exothermic which makes temperature control of samples during testing very difficult. • In addition to the intrinsic chemical reactivity of the anode carbon, the reaction rate is highly dependent on the rate of mass transport of air to, and through the sample (Figure I). This means that the air flow conditions of a reactivity test

Since the development of a technically ideal test is not realistic for plant applications, the focus can be directed to tests that, while not technically ideal, still produce data that is practically significant, i.e. the results make sense from what we know about anode reactivity and the way anodes are consumed in cells. However, the development of a practically significant test is not straightforward - different parts of a single anode in a cell is exposed to temperatures >800°C and ;.100

ij!

All experiments were performed with heat-treated CU6sNi20FelS powder milled for 40 h with 1.5 wt.% stearic acid. The heattreated powder was sieved to exclude the powder fraction with a particle size> 90/lm (corresponding to less than I wt.% of the powder).

675

..

2.5

AI 3 +/Al

Figure 1. Cyclic voltammograms on cermet anode in 22wt%K3AIF 6-50wt% Na3AIF6-28wt%AIF3-AI203 (sat) melt. Temperature: 830°C; sweep rate: 0.05 V S-I.

2.0

In order to obtain more information on the anodic reactions, Tafel plots were drawn using the data from both the forward and reverse sweep of the current-voltage curve (see Figure 2). The observed reversible potentials (RPs) for copper anodic reactions are easy to obtain since the anodic and cathodic branches meet at a given value of potential [1]. Reversible values are 1.68 and 1.92 V respectively from the forward and reverse sweep, which can be attributed to the two steps of anodic oxidation of copper in the surface layer of the cermet anode:

4.0

1.68

1.5

(b) 3.5 -

....+

:;;:

3.0 -

en

> 2.5 -

ffi 2.0 -

1.92

(2) 1.5 -4.00 -3.50 -3.00 -2.50 -2.00 -1.50 -1.00 -0.50 000

(3)

IgU/A cm-2 )

The anodic dissolution of Cu as CUF2 could happen since the calculated reversible potentials in Table I are very close. However in the present systems we studied, copper oxides were preferred, especially in the oxide saturated melts [21, 22].

Figure 2. Polarization curves for cermet anodein 22wt%K 3AlF 6 50wt%Na3AIF6-28 wt%AIF 3-AI 20 3 (saturated) melt. Temperature: 830°C; sweep rate: 0.05 V S-l. (a) forward sweep (b) reverse sweep.

These observed RP values are different from the calculated ones, 1.73 and 2.05 V (Table I). One possible reason is that different melts could have different IR drop, which would cause a shift in

Similar results were obtained by Russell [1] and Lorentse et al. [21], whereas Windisch et al. [22] found that upon oxidation of a NiFe204+ 17wt%Cu cermet anode, a layer of CU20 was formed on

1296

the copper surface, and subsequently this layer reacted with alumina in the electrolyte, forming CuAl0 2 on the outer part of the CU20 layer, but not on the metal surface. They also found that at low concentrations of alumina, the metal surface was not completely covered by CU20, or the CU20 layer was unstable. In the present work, the amount of Cu was much lower, and even in the alumina saturated melt, the metal oxide was still unstable. When the anode was polarized, the exposed metal grains dissolved anodically with the formation of copper cations, which subsequently transferred to the cathode, followed by reduction and dissolution into aluminum. Since the metallic phase did not form a continuous network in the anode, the electrochemical process should slow down rapidly when the exposed Cu in the outer region of the anode had dissolved. The ceramic matrix of the cermet anodes, being in an electrochemically stable state, underwent only chemical dissolution. After the metallic phase in the surface layer was depleted, the anode corrosion process should be controlled by chemical dissolution of the oxide phase. Chemical and electrochemical attack at the cermet/melt interface cause cermet anodes to corrode in cryolite-alumina melts. 1.5 0.60

3.0

3.5

i = ioeant'l (6) where j is the redox current density, jo is the exchange current density at formal potential, n is the electron transfer number, 1] is the overpotential, which is defined as E-E, f is FIRT and a is the electron transfer coefficient. Equation (6) is usually presented as a Tafel plot. From the Tafel plot, the exchange current density can be extracted from the intercept Igjo. For the oxygen evolution reactions on cermet anodes in alumina saturated melts. the cyclic voltammograms indicated high irreversibility, i.e. the reactions were slow and mass transfer effects were probably small. We could therefore estimate the apparent exchange current densities jo by a simple Tafel plot analysis on the rising portion of the CVs (Figure 4). From extrapolation of the Tafel linear variation of Igj with the electrode potential, E, to the standard oxidation potential (E=E), we obtained the apparent exchange current density jo = 0.19 A cm- 2, which indicates the electron transfer rate on cermet was slow. This could be attributed to the porous cermet surface [24] caused by the depletion of copper when anodically polarized, which blocked the oxygen evolution.

4.0

forward scan

0.40 c.J

2.5

The Butler-Volmer equation is usually used to describe the relationship between electron transfer rate and potential. When there is no mass transfer effect, at sufficiently large overpotential, e.g. at the potential range of oxygen evolution in the present study, the Butler-Volmer equation takes the following form [23]:

(a)

0.50

E

2.0

copper dissolution reaction both decreased almost to zero. This further supported our suggestion on the corrosion mechanism of cermet anodes.

--AI 0 saturated 2 3 - - no AI2 0 3 added

0.30 0.20 0.10

0.00 , - - - - - - - - - - - - - - - - - - - - - - ,

0.60 , - - - - - - - - - - - - - - - - - (b) 0.50 reverse scan

-0.10


- CO{g)

1.0

1800

.·.

1400

--+-AIS(g)

2000 Tem,pe,ratu.re COC) d) Predicted gaseous phases at 0.001 atm

Temperature ("C) c) Predicted condensed phases at 0.001 atm

Figure I: Predicted equilibrium phases in the Ah03+3C+3HzS system at T = 1000DC to 2000 DC: a) condensed phases at I atm. b) gaseous phases at I atm. c) condensed phases at 0.001 atm. and d) gaseous phases at 0.001 atm. Experimental Investigation on AlzOrC-HzS Reaction Systems W,,:er

Experimental Procedure Experimental investigation on carbosulfidation of Alz03(s) by using C(s) and dilute HzS(g) (5% HzS - 95% Ar) at different temperatures (1100 to 1600 DC) and reaction duration were carried out using a horizontal tube resistance-furnace (Nabertherm RHTV 200-600). Mixtures of Alz03(s) and C(s) powders (I to 6 molar ratio) were pressed at 84 MPa pressure to form pellets with diameter of Ilmm. Excess carbon was added to ensure that enough carbon is present at the reaction temperature. Three pellets were placed in an alumina boat and inserted into the furnace. The furnace was then sealed and put under vacuum before purging with the dilute H2 S (5% HzS - 95% Ar). A schematic diagram of the experimental setup is shown in Figure 2. The gas was injected directly on top the sample by using a gas injection tube (made of high purity alumina) throughout the experimental cycle. The pellets were heated up under HzS-Ar atmosphere and were held at the target temperatures (lIOODC to 1600DC) for different time intervals (3 to 12 hours) before being cooled down to room temperature.

c",,:.cd

limlgC$

COOl:buSt:iPH

htmt

Pn::.s:wre Rillise

Figure 2: A schematic diagram of the experimental set up using a horizontal tube furnace The samples, after the experiments were analyzed using XRD (Xray diffraction). SEM (scanning electron microscopy). EDS

1301

(energy dispersive X-ray spectroscopy), ICP-AES (inductively coupled plasma) and chemical filtration. XRD analyses of the samples were carried out using BRUKER DS Advanced X-ray Diffractometer with a graphite monochromator using Cu K" radiation (Theta range 2.5° to 45°, step size 0.02°, Ie = 1.5406). The samples were ground by using mortar and pestle before putting into the sample holder of the XRD machine (approximate weight for each sample is I g).

in the samples. However, it can also be seen clearly that there is a gradual decrease of the intensity with increasing reaction time. Figure 5 shows the results of XRD from samples obtained from the experiments carried out at 1500°C at different reaction time. Similar to the results from samples at 1400°C, significant Al2 S3 was observed after 6 hours of reaction. At 9 hours of reaction, the Al 2 S3 peaks are sharper. The results in Figures 4 and 5 confirm the formation of Al 2 S3 at 1400°C and 1500°C, and that the amount increases with increasing of reaction temperature.

The microstructures of the samples and the qualitative elemental composition were examined using SEM and EDS techniques in a variable pressure SEM (FE SEM Carl ZEISS SUPRA 40VP) with an accelerating voltage of 20 keY. rcp - AES analyses were carried out to quantify the elements in the sample after the experiments. The aluminum content was determined by borate fusion followed by nitric acid dissolution. The resultant solution was analyzed using Varian 730ES Inductively Coupled Plasma. The carbon and sulfur were analyzed using LECO CS200 combustion analyzer.

.'"' ,$ §

!

100

0 200

X-ray Diffraction Analysis The first set of experiments was carried out for 3 hours at temperatures of 1100 to 1400°C at 1 atm total pressure. The X-ray diffractograms of these samples are shown in Figure 3. The results confirmed the formation of AI 2 S3 in the temperature range studied as indicated by the presence of peaks labeled 2 in the difractograms. However, at this 3 hours reaction time it appears that only a small amount of AI 2 S3 is formed. It can be seen from Figure 3 that there are major traces of unreacted corundum Al 20 3 (peaks labeled I) and graphite (peaks labeled 3).

m

_

•

•

•

•

•

•

•

•

m

N

•

Angle :I-Theta

Figure 4: X-ray diffraction pattern of the samples after 3, 6, and 9 hours experiments at 1400 0C. (I = corundum (AI2 0 3), 2 = aluminum sulfide (AI2 S3) and 3 = Graphite (C))

00f) 00()

$00 {)

t$OO

tOOO ';;'

1: :I 0

J.

..J

00() {)

$00 00()

$00 {)

Angle 2-Theta

Figure 5: X-ray diffraction pattern of the samples after 6 and 9 hours experiments at 1500 °c. (l = corundum (AI 2 0 3), 2 = aluminum sulfide (AI 2 S3) and 3 = Graphite (C))

t$OO 1000 $00 0

•

•

•

•

•

•

•

•

•

•

ro

•

Angle 2-The!!!

SEM and EDS Analysis Secondary electron images of the typical microstructure of the sample after the experiments are shown in Figure 6. It can be seen from Figure 6(a) that the surface of the sample consists of packed re-crystallised grains. Figure 6(b) shows the surface at higher magnification. The boundaries between the grains appear to be fused and the edges are smooth and rounded. Some small globular particles are also present on the surface of the grains. These shapes of the particles and surface morphology are characteristics of a surface that undergo a melting and recrystallisation during the process. It should be noted that AI 2 S3 is a liquid above II OO°C at I atm.

Figure 3: X-Ray diffraction pattern of the samples after 3 hours experiments at temperatures of 1100 to 1400°C. (I = corundum (AI 2 0 3), 2 = aluminum sulfide (AhS3) and 3 = Graphite (C)) Further experiments at longer reaction time were carried out at 1300 to I 500°C. Figure 4 shows the comparison of XRD pattern of the samples after experiments at 1400°C for three different times (3, 6 and 9 hours). Ah03 and AhS3 peaks are marked by "I" and "2", respectively. As shown in Figure 4, significant aluminum sulfide (AI 2 S3) was detected after 6 and 9 hours of reaction. This is indicated by the higher and sharper Al 2 S3 peaks at 6 and 9 hours compared to those from at 3 hours. Al 2 0 3 peaks are still present, indicated that some Al 2 0 3 remains and unreacted

EDS analyses were carried out on the surface of the samples to evaluate the presence of different elements. Figure 6(c) shows the

1302

EDS spectrum of the surface indicating the presence of AI, S, 0, and C. This result, along with the X-Ray diffractogram, suggests that S from H 2 S reacted with the surface forming A1 2 S3, and that some unreacted Ah03 is present below this surface.

Determination ofn (Conversion) The percentage of conversion from AI 2 0 3 to AI 2 S3was determined by chemical dissolution and filtration. As pure AI 2 S3 completely dissolves in hydrochloric acid (HCI), a portion of the experimental samples were dissolved in HCI (36% w/w aqueous solution) and the solution was then filtered out. The amount of mass that dissolves in HCl represents the formed Al2 S3 while the residues are the unreacted Al 2 0 3 and C. Table I: Results of ICP-AES and LECO analyses of samples reacted at 1300°C, 1400°C and IS00°C Temperature Duration wt% wt% wt% S wt%O (hours) Al C COC) 1300 1400 1500

6

38.5

4.6

25

31.8

6

38.8

41.1

6.1

13.9

9

37.3

45.8

6.5

10.3

3 9

31.7 40.3

17.8 45.9

24.4 0.8

25.9 12.9

Table II: The conversion of Al 20 3 to Al2 S3 from selected samEles at 1400°C and IS00°C %of Temperature Duration Weight of Conversion (0C) (hours) Sample (g) (11) 0.2012 75.4% 6 1400 77% 9 0.2051 78.9% 6 0.2186 1500 81.6% 9 0.2060 Approximately 0.2g of ground samples from each experiment was put into 20 ml HCI (36% w/w aqueous solution) in an Erlenmeyer flask. The solution was then stirred vigorously for few minutes and left for approximately 3 hours to allow all AhS3 to be dissolved. The solution, along with the residue was then filtered using filter paper. The filter papers with the residues were then dried and weighed. The dried filter paper with the residue was then weighed to determine the weight loss of the sample due to dissolution. From these dissolution and filtration processes, the percent of conversion (11) of Al 20 3 to Al 2 S3 was calculated using following equation:

SpecllUm 2

(c)

@A

1.4

ZA

i!

00%

valid for the temperatures higher than this point due to the melting of phosphorous content phase ofFe3P'

100 00 00 11)

t

I I

00 00 4iII :It :It I@

•

I,.

1100

111)@

ilq Figure 2: Comparison of solidification amount ofHPGT for different cooling rate

Figure 4: Thermal expansion coefficient ofHPGT vs. temperature graph indicates an irregular reduction in expansion at 977.8 °C due to the melting of Fe3P phase.

In order to validate the results obtained from the thermodynamic calculations for the solidification of HPGI, DSC test was carried out on HPGT samples during heating and cooling process as shown in Fig.3. DSC measurement clearly shows that the melting temperature of HPGT begins at about 988°C while its solidification ends at about 926°C. Delay in solidification ending temperature compared to the start of melting arises from the undercooling effect required for solidification [5]. Nevertheless, the Factsage calculations indicate the solidification ending temperature about 35°C higher than that of DSC experiment.

As cast rod ding microstructure Rectangular graphitic block (120 mm x 75 mm x 200 mm (L WH)) with a slot size 20 mm x 20 mm machined at the bottom surface was used for electrolysis tests. Steel collector bar was machined to 10 mm x 10 mm and was centered in the slot using ceramic plates. The entire block with collector bar was then put in a heat treatment furnace to be preheated at 400°C before casting. The effects of preheating temperature and the methods applied in industrial conditions playa significant role in final microstructure of rodding cast iron and particularly in the contact pressure (thermal stress) made on cathodic block during the cell start-up. However this subject is beyond the scope of this study and much research is needed to investigate this effect. HPGT alloy was melted by induction furnace and poured into slot already preheated. Metallographic sample cut from the center of bar was polished and etched by Nital reagent (a solution of 4 % acid nitric and 96 % alcohol) to investigate the microstructure at HPGT-bar interface in as cast condition.

StAlrt of melti!'!il

9S7,g

Figure Sa shows the microstructure at the interface HPGI-bar casting process. Figure Sa indicates that the rapidly cooled microstructure of HPGT is mainly similar to the white cast iron [7] containing cementite (Fe3C), very fine pearlite and a network of iron phosphide (Fe3P) as depicted in Figure 5b.

Figure 3: DSC measurement in cooling and heating of HPGI at the rate of 10°C min-]

However, the precipitation of graphite inside the HPGI region with ferritic microstructure was also observed only near to the interface of HPGT-steel as shown in Figs. Sa and 6. On the other side of the interface, inside the steel region with a typical ferritic microstructure, a layer about 70 microns containing fine pearlitic microstructure was detected near to the interface. This phenomenon may occur due to the diffusion of carbon from HPGT into the steel fonning ferritic and pearlitic microstructures on both HPGT and Steel sides of the interface, respectively.

Figure 4 shows the thermal expansion test of HPGI carried out to justity the identified peaks in DSC test during heating process. It is clearly observed that the HPGI sample expands with a constant rate by increasing temperature up to 873.1 °C (892.4 °C at DSC test) where a contraction occurs [6] as a result of the iron phase change from ferrite (a) to austenite (y). Subsequently the alloy continues to expand for a second time up to 977.8°C. With further increase of the temperature, the expansion graph abruptly declines unlike other ferrous alloy with low phosphor content. This is indeed the point that the alloy begins to melt as shown in DSC test at 987.8°C. The expansion measurement will not be

1306

Microstructure after electrolysis tests

Two electrolysis tests were separately carried out for 3 and 9 hours at 960°C using the graphitic blocks with sealed HPGl. RTA cryolitic electrolyte was used for the tests. The detail of the electrolysis process can be found in our previous TMS publication [8]. After electrolysis tests, the cast iron attached to the bar was removed from the block and a sample for metallographic purpose was cut at the center of the bar. Figure 7 compares the metallographic surface of HPGI-bar etched by Nital reagent after two electrolysis tests. The layer formed around the bar and pointed by arrows in Figure 7 shows the zone where carbon diffuses from HPGI into the steel bar. This layer appears immediately after etching the surface with Nital reagent. It is clearly observed that the carbon diffusion layer becomes thicker (1.3 mm and 2.6 mm, respectively) with electrolysis time as expected.

Figure 5: (a) Microstructure at the Interface of HPGI-Steel bar after casting and (b) white cast iron microstructure of at the center ofHPGI away from the interface

Figure 7: Nital etched metallographic surface of HPGI-steel bar after two electrolysis test, arrows point to the diffusion layer.

Figure 6: Precipitation of graphite with ferritic matrix near to the interface while the rest of HPGI is white cast iron.

1307

Figure 8: Microstructure of HPGI after electrolysis for (a) 3hrs and (b) 9hrs showing the width of diffusion layer.

Figure 9: HPGT and steel are integrated together after both electrolysis tests without having interface.

Figure 8 shows the microstructure of HPGT-steel bar after two electrolysis tests. The microstructure of HPGI contains irregular form of graphite precipitated from decomposition of cementite existing in as-cast microstructure dispersed in a ferritic matrix as well as a network of phosphor containing phase. The microstructure of HPGT close to the interface where carbon migrates toward steel is quite different compared with the rest of HPGT microstructure as shown in both Figure 8a and 8b with dashed lines. Graphite size in this layer is extremely bigger than those out of the layer. Phosphor also diffused to this layer and made a large and interconnected network ofFe3P particles. Figure 8a demonstrates that this layer can be divided into 4 regions in terms of Fe3P distribution. Large Fe3P particles are observed in the first region. Tn the second region, there is only ferritic matrix without presence of any Fe3P particle. Third region once more shows the Fe3P particles but fairly smaller than those in the first region. In the last region only ferritic matrix is again observed. It is worth noting that the original interface position must be situated right next to the most distal graphite particles at the left side of the region 4 as shown in Figure 8 while the pearlitic microstructure has been formed at the other side of the region 4. By close attention to this region using a SEM image shown in Figure 9, it is observed that the HPGI-steel collector bar interface has been completely disappeared implying the integration of both alloys.

Figure 10: Diffusion of P into the pearlitic layer of steel after 3hrs electrolysis test. Figure 10 shows the diffusion of P in the pearlitic layer of the steel bar at the interfac position after 3hrs electrolysis. It is therefore concluded that both C and P elements diffuse from HPGI into the steel collector bar forming a particular layer in both HPGI and steel bar alloys close to their interface. However the diffusion of C is more significantly deeper than the P into the steel.

By comparing Figure 8a and 8b, it is also observed that the thickness of the diffusion layer after 9hrs electrolysis test which reaches about 700 11m is at least three times bigger than of 3 hrs. Moreover, there are no Fe3P particles in the third region for higher electrolysis time as shown in Figure 8b. As it was mentioned in Figure 7, large pearlitic zone forms inside the steel bar due to the diffusion of carbons from the HPGI layer shown in Figure 8. It is interesting to note that the diffusion of P element into the steel was also detected for both samples.

Labrecque et al. [6] have also investigated the microstructure of both HPGI rodding and steel collector bar before (as-cast) and after the period of electrolysis operation in the plant. They characterized the microstructure of as-cast HPGI as white cast iron without pointing out the microstructure at the interface HPGT-Steel.

1308

References

After service, they showed the interface microstructure containing the integration of both metals (disappearance of interface) as well as the formation of phosphide parallel to the interface due to the diffusion of phosphor element. They have also pointed to some cavities planes formed at the interface. However the goal of their study was to compare two different types of cast iron i.e. HPGI and FDI (ferritic ductile iron). They have not sophisticatedly characterized the microstructure before and after electrolysis process.

Conclusions The study of solidification behaviour of HPGI (high phosphorous gray iron) used as a rodding of cathode to the steel collector bar showed the decrease of the freezing point down to about 960 DC and 925 DC according to the Factsage calculation and DSC measurement, respectively. The thermal expansion test also confirmed the melting temperature of HPGI at about 978 DC (988 DC by DSC test) by the abrupt decline of the expansion vs. temperature graph. The microstructure at the interface of the cast HPGI and steel collector bar was also studied in as-cast and after 3 and 9 hrs electrolysis tests. The rapidly cooled microstructure of HPGI was mainly white cast iron containing cementite (Fe3C), very fine pearlite and a network of iron phosphide (Fe3P) except at the interface where graphitization has occurred. A diffused carbon layer with a width of about 70 microns containing pearlitic microstructure was detected around steel side of the interface.

1.

C. W. Bale, E. Belisle, P. Chartrand, S. A. Decterov, G. Eriksson, K. Hack, 1. H. lung, Y. B. Kang, J. Melancon, A. D. Pelton, C. Robelin, S. Petersen, "FactSage thermochemical software and databases - recent developments", Calphad: Computer Coupling of Phase Diagrams and Thermochemistry 2009, 33 (2), 295-311.

2.

John Campbell, Complete Casting Handbook (Butterworth Heinemann, Elsevier, 2011), 1220 p.

3.

M. Keppert, G. M. Haarberg and S. Rolseth, "Electrochemical behavior of aluminum phosphate in Proceedings-Electrochemical cryolite-alumina melts" Society, PV 2004-24 ( 2006),237-245.

4.

V. Raghavan, "C-Fe-P (Carbon-Iron-Phosphorus), "Journal of Phase Equilibria and Diffusion", 25 (6), (2004),541-542.

5.

J. R. Davis, Cast Irons. (ASM International: Technology & Engineering, 1996), 494 p.

After 3 and 9 hrs electrolysis tests, a layer was formed next to the interface position of both HPGI and steel due to the diffusion of carbon. A layer with the microstructure entirely formed from the fine pearlite was detected at the interface of the steel with ferritic microstructure. This layer can be visually observed on the surface of the metallographic samples right after the etching process by Nital in the form of a white layer. The thickness of the layer was measured to be 1.3 mm and 2.6 mm respectively. The layer formed at the interface of HPGI had a width respectively about 220 /lm and 700 /lm containing bigger graphite with less particle fraction in comparison to the graphite particles out of the layer. On the other hand, extremely large and interconnected network of Fe3P particles was observed in this layer. After electrolysis tests, the interface line of HPGI-steel observed in as-cast microstructure was completely disappeared and both alloys were integrated together. This phenomenon facilitates the current passing through the interface during the electrolysis process.

Acknowledgement This work was made possible with financial participation of Rio Tinto Alcan (RTA), "Conseil de Recherches en Sciences Naturelles et en Genie du Canada" (CRSNG) and "Fonds Quebecois de la Recherche sur la Nature et les Technologies" (FQRNT). Authors gratefully acknowledge the personnel of "Centre de Caracterisation des Materiaux". at Universite de Sherbrooke for material characterization.

1309

6.

L. Caruso, K. A. Rye, and M. Sorlie, "Experimental comparison of cathode rodding practices", TMS Light Metals, The Minerals, Metals and Materials Society, (2007), 827-831.

7.

C. Labrecque, M. Gagne, D. Lavoie, A. Levesque, and B. Murphy, "A new technology for cathode rodding used in aluminium electrolytic cells, TMS Light Metals, The Minerals, Metals and Materials Society, (2003), 661-667.

8.

M. Brassard, M. Lebeuf, A. Blais, L. Rivoaland, M. Desilets and G. Soucy, "Characterization of carbon cathode materials by X-ray microtomography", TMS Light Metals, The Minerals, Metals and Materials Society, (2012),1325-1329.

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

PREPARING AI-Sc-Zr ALLOYS IN ALUMINUM ELECTROLYSIS PROCESS Yi Qian, Jilai Xue, Qiaochu Liu, lun Zhu School of Metallurgical and Ecological Engineering, University of Science and Technology Beijing Xueyuan Road 30, 100083 Beijing, China Keywords: Al alloy, aluminum electrolysis, rare metals conventional industrial Hall-Heroul cells. Scandium oxide (>99.99 wt%) and zirconium oxide (>99.99 wt%) were also added into this cryolitic melt. All these chemicals before mixing were subject to dry treatment in a muffle furnace. The electrolyte composition was of Na3A1F6- 4 wt% MgF 4 - 2 wt% CaFr 2 wt% SC203 - (0.05 wt%-I wt%) ZrOb in which there was no Al 20 3 addition involved at this stage of investigation.

Abstract AI-RE Alloys have attracted much attention in recent years due to their great potential in many advanced applications. Tn this work, Al-Sc-Zr alloys were prepared in a laboratory electrolysis cell. Effects of SC203 and Zr02 additions and the electrolysis time on Sc-Zr contents and their ratio in the alloys were investigated in Na3A1F6 based melt at 960°C. SEM and ICP-AES show that Sc and Zr contents in the alloy produced were 0.25-0.32 wt% and 0.26-1.24 wt%. SEM-EDS analysis reveals that AI 3(Sc,Zr) particles form in Al alloy prepared by electrolysis. Cyclic voltammetry results demonstrate that Zr4+ proceeds a two-step process electrolysis at -0.7V and -1.05V, respectively, while no single peak appears for Sc deposition.

Set-up for Aluminum Electrolysis Figure 1 is the schematic drawing of the experimental setup used as a laboratory cell. A inner diameter 50 mmX 80 mm graphite crucible embedded with an inner diameter 45 mm X 108 mm corundum crucible served as cell lining. A small hole of diameter 5 mm was drilled at the bottom of the corundum crucible that was crossed with a small graphite plug to carry the electrical current from the cell bottom to the liquid aluminum (99.99 wt%). The corundum crucible containing 160 g electrolyte and 15 gAl metal was placed in a vertical tube furnace with temperature controller.

Introduction It is known that a small addition of Sc in aluminum can

significantly improve its properties [1-3], and a combined addition of Sc with other transition metals, for instance, Zr can offer even better properties and lower production costs. The addition of minor zirconium to aluminum alloys with scandium enhance positive effect on their operational properties, due to the formation of extremely fine, coherent Ah(Sc,Zr) particles with Ll2 structure, which can substantially inhibit recrystallization and dislocations [4-5]. These alloys have good welding performance, superior corrosion resistance and strong toughness, etc [6-8], which can be used in transportation, construction, aerospace, and other engineering fields. There are a number of publications on preparing AI-Sc alloy in molten salt [9-11], while few in open literature are dealt with electrochemical preparing AI-Sc-Zr alloys. Generally, AI-Sc-Zr alloys are prepared by melting Al-Sc and Al-Zr alloys together in a furnace [12-13]. The preparation of these alloys with master alloys is expensive due to high price of Sc and Zr raw materials and smelting costs. Therefore, it is low costs to prepare Al-Sc-Zr alloys directly using their metallic oxides in aluminum electrolysis process.

I-Stainless steel anode rod; 2-Gas outlet; 3-Cooling coil; 4-Thermocouple; 5-Corundum crucible; 6-Graphite anode; 7-Molten electrolyte; 8-Aluminum cathode; 9-Stainless steel cathode rod; 1O-Gas inlet Figure 1. Schematic drawing of experimental setup for electrolysis During the whole experiment process, a flow of 30 ml/min of argon was kept through the reaction tube in the furnace to provide a protective atmosphere. The laboratory electrolysis process was carried out at a constant current with the cathode current density of 1 A/cm2 and duration of 2 h at 960°C. The testing temperature was measured by a thermocouple and the cell voltage was monitored by a voltmeter, both continuously logged into the data system in a Pc.

Tn this work, we are intend to directly prepare AI-Sc-Zr alloys in aluminum electrolysis cell by adding SC203 and Zr02 into cryolite based electrolyte. Variation of Sc and Zr contents in aluminum alloys were investigated by changing the electrolysis time parameter and the oxide addition in electrolyte. SEM-EDS and voltammetry techniques were applied to characterizing the microstructure of the alloy and electrochemical behaviors. Experimental

Apparatus for Voltammetry Measurement Figure 2 illustrates a brief schematic drawing of the experimental apparatus for electrochemical study. Various electrochemical measurements were carried out using an Tm6eX electrochemical workstation (Zahner Co., Ltd.).

Chemicals and Electrolyte Sodium cryolite (CR=2.4, industrial grade), MgF2 (analytical grade) and CaF 2 (analytical grade) were mixed and melted at 960°C to form molten electrolyte similar to that used in

1311

The electrolyte used in voltammetry study in was the cryolitic melt similar to that mentioned above with addition of scandium oxide and zirconium oxide. A platinum wire was used as a reference electrode, a molybdenum wire as the working electrode, and a graphite rod (6 mm) as the counter electrode. Before the test, the Mo wire was polished with SiC paper and cleaned in dilute hydrochloric acid and then in ethanol by ultrasonic to a shine finish. The graphite electrode was pre-cleaned by boiling it in 5 wt% dilute hydrochloric acid for I h, and then washing with deionized water. All electrodes had been dried before test in oven to eliminate possible influence from the moisture.

current density changing. This may be caused by uneven temperature distribution in the cell when graphite anode inserted into the electrolyte melt. After that, the voltage was fluctuated around 3.35 V. The course of dissolving and electrolysis of scandium oxide and zirconium oxide in the molten salts may lead to small change in cell voltage. The recording of cell voltage against time demonstrates a relative stable operation and good control state during laboratory electrolysis, which is considered important in a possible large scale testing in future.

Temp erature: 960 0 C Cmrellt density: 1 A/em2

I-Corundum tube; 2-Mo wire; 3- graphite electrode; 4-Pt wire; 5-Molten salt: 6-Reaction chamber Figure 2. Schematic illustration of the experimental apparatus for electrochemical study in cryoltic melt

o

20

40

GO

Time! min

80

100

120

Figure 4. Variation of cell voltage vs. electrolysis time (2 wt% SC203 addition, 0.1 wt% Zr02 addition)

Characterization of Al Alloys Figure 3 is the photograph showing a typical alloy product obtained by electrolysis for 2 h at 960°C in the sodium cryolitic melt. At the end of the electrolysis process, the crucible containing the alloy product was moved down to the bottom of the furnace for a fast cooling. After that, the aluminum alloy was removed from the crucible, and the solid electrolyte on the surface of the alloy was cleaned carefully.

Contents ofZr and Sc in Al Alloys Produced Table I shows the effect of electrolysis time on Sc and Zr content in Al alloys produced by electrolysis in the sodium cryolitic melt at 960°C. All electrolysis tests were performed at the current density of I A/cm2 , while the additions of SC203 and Zr02 were different from test to test. It was obvious that the content of Zr in alloy increased with the

prolonged electrolysis time and the maximum value could reach 3.27 wt%, while at the same time Sc content appeared to be almost unchanged. To understand further such phenomenon, another set of experiments were arranged with varied addition of Zr02 into the cryolitic melt. Table 1. Sc and Zr Content in Al Alloys Produced with Varying Period of Time in Electrolysis Current Time Zr Sample SC20 3 Zr02 Sc Density No. (wt%) (wt%) (h) (wt%) (wt%) 2 ( Alcm )

Figure 3. Photograph showing aluminum alloy obtained by electrolysis (in sodium cryolite (CR=2.4) mixed with 4 wt%MgF 4-2 wt%CaFr2 wt%Sc203-0.05 wt% Zr02) In order to know more information about the characteristics of the alloy, the alloy sample was cut into half A number of specimens were taken from one of the half parts, and was dissolved to determine Sc and Zr contents using ICP-AES technique. Another half was sanded with SiC paper, polished, etched in Dix-Keller solution for SEM observation.

ASZ-I

2

I

1

0.5

0.25

0.75

ASZ-2

2

I

1

1

0.28

3.27

ASZ-3

2

I

I

2

0.26

1.24

In Table II, the contents of Sc and Zr in the Al alloys produced are listed with varying Zr02 additions while a constant addition of 2 wt% SC203. Two electrolysis experiments were performed at 960°C for 2 h and a thermal reduction experiment (ZW-l) was done for comparison under identical testing conditions without electrical current. It is noticed that Sc content in the thermal reduction is similar to the values in Table I, suggesting Sc in this

Results and Discussion

Relation Between Electrolysis Time and Cell Voltage Figure 4 displays the relation between cell voltage and electrolysis time. At the beginning of electrolysis process in the first few minutes, the cell voltage reduced from 3.64 V to 3.42 V without

1312

case may not come from the electrolysis process. This may explain the fact that Sc content is almost unchanged with variation in electrolysis time. Table IT. Sc and Zr Content in Al Alloys Produced with Various Zr02 Additions in Electrolyte Current Sc Zr Sample SC203 Zr02 Density Ratio (wt%) (wt%) (wt%) (wt%) ( No. 2 A/cm ) ZW-I

2

0.1

0

0.25

0.30

I: 1.2

ZW-2

2

0.1

I

0.32

0.45

I: 1.4

ZW-3

2

0.05

I

0.27

0.26

I :I

From thermodynamic calculation, the decomposition voltage of Zr02 (2.40V) is lower than that of SC203 (2.70V) at 960°C. This gives Zr an advantage in electrochemical deposition, leading to its increase with increasing electrolysis time and Zr02 concentration in cryolitic melt (see Table I and Table II). Zr content can vary with changing Zr02 concentration in cryolitic melt and electrolysis time as well. This may provide a technical possibility in controlling the alloys composition. It is an interesting phenomenon, however, that the change in the ratio of SC203 to Zr02 has almost no effect on the ratio of Sc to Zr in Al alloys produced. This may also imply that it has to look for other ways to control Sc content and the related composition of Al alloys prepared using electrolysis method. The obtained results, so far, suggest that the investigation has still long way to go towards this direction.

(c)

3()

20

Characterization of Al Alloys Produced Figure 5 shows SEM-EDS micrographs of Al alloy sample prepared by electrolysis. There are precipitate particles with regular shape in Figure 5(a) and is the enlargement of a square precipitated particle in Figure 5(b). The precipitated particles of square with protruding lobes and the triangular shape are all AISc-Zr phase, which are quite similar to those observed in literature [14,15].

keV

Figure 5. SEM-EDS analysis of Al alloy sample produced in electrolysis: (a) Al-Sc-Zr alloys (sample NO.ZW-2), (b) AlSc-Zr particle, (c) EDS analysis on the particle in (b)

The chemical composItIOn of the precipitated particle can be further contirmed by the SEM-EDS results, as shown in Figure 5( c), in which the particle only has AI, Sc, Zr elements and the proportion of the sum of (Sc, Zr) atoms to Al atoms is 1:3, i.e., the same as AI3(Sc,Zr). In Figure 6, the SEM photographs illustrate the element mapping of AI 3(Sc,Zr) particle presented in Figure 5(c). It is obvious that the amount of Al within the particle is lower than that in the alloy matrix (see Figure 6(a), Al mapping). It is known that AI3Zr with L12 space lattice structure would have priority to generate in the molten aluminum alloys containing Sc and Zr [16]. And then Sc atom may diffuse to the particle of Al3Zr and replace part of Zr atom, forming AI 3(Sc,Zr) with the L12 space lattice structure. This is why the mapping of Sc and Zr element displays complementarity within the area of the precipitated particle (see Figure 6(b) and Figure 6(c).

Figure 6. Element mappings of AI 3(Sc,Zr) particle. (a) aluminum, (b) zirconium, (c) scandium

1313

Cyclic Voltammetry Figure 7 shows the cyclic voltammetry curve of cryolitic melt with 2 wt% SC203 and 0.1 wt% Zr02 at a molybdenum electrode. It indicates that the electro-reduction of Zr4 + proceeds a two-step process during electrolysis according to literature [17]. Tn the cathodic scan, two cathodic current peaks A and B are observed.

References I. M. Asta, V. Ozolins, "Structural, vibrational, and thermodynamic properties of AI-Sc alloys and intermetallic compounds," Physical Review B, 64 (2001), 094104/1-14. 2. M. Ferry, N. E. Hamilton, F. J. Humphreys, "Continuous and discontinuous grain coarsening in a fine-grained particlecontaining AI-Sc alloy," Acta Materialia, 53 (2005), 1097-1109. 3. K. Venkateswarlu, L.c. Pathak, A.K. Raya, Goutam Das, "Microstructure, tensile strength and wear behaviour of AI-Sc alloy," Materials Science and Engineering A, 383 (2004), 374380. 4. M. Schobel, P. Pongratz, H. P. Degischer, "Coherency loss of AI 3(Sc,Zr) precipitates by deformation of an AI-Zn-Mg alloy," Acta Materialia, 60 (2012), 4247-4254.

Temperature: 960 ·C

5. Y. Harada, D. C. dun and, "Microstructure of AI3Sc with ternary transition-metal additions," Materials Science and Engineering A, 329-331 (2002),686-695.

Scan rate: 100 mV/s

Reference electrode: Pt

6.1. N. Fridlyander, E. A. Tkachenko, V. V. Bersteney, "Effect of microstructure on the cracking resistance characteristics of AI-ZnMg-Cu-Zr wrought high-strength alloy," Material Science Forum, 396-402 (2002), 1347-1351.

Figure 7. Cyclic voltammetry curve of SC203 (2 wt%) and Zr02 (0.1 wt%) dissolved in Na3AIF6-4 wt%MgFz-2 wt%CaF 2

7. M. Vlach et aI, "Phase transformations in isochronally annealed mould-cast and cold-rolled AI-Sc-Zr-based alloy," Journal of Alloys and Compounds, 492 (2010), 143-148.

Zr4 + may begin to get two electrons when the scanning potential arrives at -0.7 V (peak A), and the cathodic peak A could be related to the reduction of Zr4 +/Zr+:

8. E. P. Kwon et al. "The Effect of an Addition of Sc and Zr on the Precipitation Behavior of AA6061 Alloy," Metals and materials international, 16 (2010), 701-707.

(I)

9. Q. Liu, J. Xue, J. Zhu, C. Guan, "Preparing AluminiumScandium inter-alloys during reduction process in KF-AIFrSc203 melts," Light metals, (2012), 685-689.

And Zr may come out later at -1.05 V. Cathodic peak B could be related to the reduction of Zr+/Zr: zr+ + 2e

Zr

10. C. Guan, J. Xue, Q. Liu, "Preparing Aluminum-Scandium Alloys Using Direct Hall Reduction Process," 3rd international Symposium on High-Temperature Metallurgical Processing, TMS, 15 May 2012.

(2)

Tn the reverse scan process, two corresponding oxidation peaks of Zr appeared at A' and B'. However, the cathodic peak corresponding to the reduction of Sc is hard to be found in the testing curve above. Tn the Sc-Zr phase diagram [18], it is showed that scandium and zirconium can be soluble each other arbitrarily. This could decrease the deposition potential of Sc, which would result in a move of peak of Sc3+ to near the position of zr+ peak.

II. M. Harata, K. Yasuda, H. Yakushiji, "Electrochemical Production of AI-Sc Alloy in CaC12-Sc203 Molten Salt," Journal ofAlloys and Compounds, 474 (2009), 124-130. 12. C. B. Fuller, J. L. Murray, D. N. Seidman, ''Temporal evolution of the nanostructure of AI(Sc, Zr) alloys:Part TChemical compositions of AI 3(Scl_ xZrx ) precipitates," Acta Materialia, 53 (2005), 5401 - 5413.

Conclusions

13. M. Song, Y. He, S. Fang, "Effects of Zr Content on the Yield Strength of an AI-Sc Alloy," Journal of Materials Engineering and Performance, 20 (2011),377-381.

1. AI-Sc-Zr alloy can be electrochemically prepared in sodium cryolite electrolyte at 960°C with the addition of SC203 and Zr02; in which Zr content in Al alloys can increase with increased Zr02 concentration in electrolyte and prolonged electrolysis time.

14. A. Tolley et aI, "Segregation in AI 3(Sc,Zr) precipitates in AISc-Zr alloys," Scripta Materialia, 52 (2005), 621 - 625.

2. SEM and EDS analysis show that AI 3(Sc,Zr) particles form in Al alloy prepared by electrolysis.

15. X. Dai et aI, "Precipitation behavior of AI 3(Sc,Zr) particle in Al-9.0Zn-2.5Mg-2.5Cu-0.15Zr-0.2Sc alloy," Transactions of materials and heat treatment, 33 (2012),61-66.

3. Voltammetry study shows that Zr4+ proceeds a two-step process in electrolysis, while no single peak appears for Sc deposition.

16. S. Hori et al. "Phase decomposition in splat quenched AI-6% Hf alloy," Journal of the Japan institute of Metals, 32 (1982), 408-412. 17. Z. Chen, Y. Li, S. Li, "Electrochemical behavior of zirconium in LiCI-KCI molten salt at Mo electrode," Journal of Alloys and Compounds, 509 (2011),5958-5961.

Acknowledgement General support from aluminum industry in China and University of Science and Technology Beijing is acknowledged.

18. A. Palenzona, S. Cirafici, "The Sc-Zr (Scandium-Zirconium) system," Journal of Phase Equilibria, 12 (1991),53-55.

1314

CBF Environmental & Anode Electrical Connections SESSION CHAIR

Marc Gagnon Aluminerie Alouette Sept-lIes, QC Canada

Light Metals 2013 Edited by: Barry Sadler TMS (The Minerals, Metals & Materials Society), 2013

FUME TREATMENT SYSTEMS BASED ON RTO TECHNOLOGY FOR CARBON BAKING FURNACES Matthias Hagen and Dr. Bernd Schricker L TB Luft-und Thermotechnik Bayreuth, 95497 Goldkronach, Germany Keywords: Emissions, Fume treatment, VOC, PAH, HF, SOx, RTO, benzene, B(a)P

Abstract

?

it

Increasing environmental demands, especially for emissions of the carcinogenic PAH (Polycyclic Aromatic Hydrocarbons) require new technologies for the treatment of fumes from carbon anode production plants. Thermal systems have been supplied to several paste plants as well as baking furnaces for the production of anodes, electrodes and cathodes. Due to the specific pollutants such as HF and the behaviour of the sticky condensates, the systems have to be designed properly to ensure a high availability. Pre-filters and preheating avoid clogging of the heat exchanger material of regenerative thermal oxidizers (RTO's). In addition, stringent health and safety issues as well as a specific fire protection system have to be considered during the design, construction and operation of a new plant. The paper shows the required process steps and the technical solution on the basis of plants for several baking furnaces in Poland and France.

'" .2

. c

:I: 0

40 35 30 25 20

c

15

1;

10

0

Figure 1: Distribution of PAH from Different Sources (L TB measurement data)

Situation of Emissions

In many countries, benz(a)pyrene, B(a)P is used to monitor PAH's instead of measuring all PAH16 compounds and it is currently listed separately in the draft for the IPPC non-ferrous metals from July 2009. So the critical pollutants of PAH16 (or B(a)P) and Benzene are important considerations for carbon manufacturing plants and have to be considered during the design of the fume treatment systems (FTS).

Independent of the technical systems used for the production of carbon products, such as electrodes, special graphite products or anodes, Polycyclic Aromatic Hydrocarbons (PAH's) are an important issue for the production manager. As PAH's are classified as carcinogenic, a major target is to reduce emissions and the impact on people as well as on the environment.

In the past, conventional dust collecting systems like electrostatic precipitators or bag house filters have been used. To enable a proper efficiency, these systems have to be equipped with a cooling system in order to condense some of the gaseous compounds. Figure 2 shows that not all PAH's can be condensed, and for example, naphthalene is more than 90% gaseous. Consequently, purification with conventional filter systems as a stand-alone solution is no longer possible in order to meet current emission limits.

Sources of emissions occur across the whole manufacturing plant for processes where pitch is used. The following areas can be identified in carbon plants: Liquid pitch storage and melting Paste plant or green carbon production Baking furnace As the baking furnace represents the majority of the emissions of a carbon manufacturing plant, this paper will only investigate the emissions from baking furnaces. The U.S. Environmental Protection Agency (EPA) has classiiied seven PAH compounds as probable human carcinogens: benz(a)pyrene, benz(b )fluoranthene, benz(k)fluoranthene, chrysene, dibenz(a,h)anthracene, and indeno(I,2,3-cd)pyrene [I].

Based on the above. a new method for the treatment of emissions is required. Thermal treatment has been recognized already as an efficient method to reduce volatile organic compound (VOC) emissions. Several Regenerative Thermal Oxidizers (RTO's) have been installed for liquid pitch storage and green paste plant emissions. The solution at EMAL has already been reported [2].

In order to compare emissions and to have a common basis, 16 PAH compounds which can be found in all emissions are grouped together and referred to as PAH 16. The distribution within this group has been measured by LTB and Figure 1 shows the distribution for four different emission sources. Most important is that different emission sources show a very similar distribution of the different PAH's. Hence, this could be seen as a strong basis for future selection of a fume treatment system.

To avoid the need to replace existing fume treatment plants of baking furnaces, some efforts have been made to improve the overall efficiency by optimizing the combustion inside the furnace. It has been reported that firing systems could be improved in order to reduce the PAH emissions [3], but a further thermal treatment is mandatory [4].

1317

i

,(%< 30"C)

'"particulate (% > 0,2

different vapor pressure behavior of each single PAH compound.

i,

100% 90%

-

80%

-

70%

-

-

If the requirements are more stringent (typically the case in Europe) and the efficiency has to be higher or soot occurs, a second step of treatment has to be considered. For baking furnaces already equipped with an electrostatic precipitator (ESP), this could be used or a specific pre-filter can be installed instead. The pre-filter is filled with ceramic media which are specially chosen to provide a higher turbulence and impact surface compared to those used in an RTO. This enables the collection and storage of particles in the pre-filter instead of the RTO heat exchanger. Figure 4 shows a pre-filter which was in operation for two days and which is already loaded with particles and condensates.

60% 50% 40% 30% 20% 10%

Figure 2: Physical State ofPAH16 at Common Fume Conditions (temperature 120°C, humidity 5 % vol.; LTB measurement data) Concept of a Thermal Fume Treatment

Based on the results of several ROxiTHERMTM plants [5], it has become apparent that normal RTO systems would fail due to the relatively high amount of condensed particles, i.e. tar. When considering the use and performance of an RTO system, the following items have to be considered: Particulate matter and pre-filtering Design of the RTO (condensates & redundancy) Hydrogen Fluoride (HF) emission due to butts recycling SOx emissions

Figure 4: Packing in a Pre-filter

Pre-Filter

Currently, several baking furnaces for the production of electrodes and anodes are equipped with RTO systems. On all these applications, the focus is on PAH control. The achievable efficiency measured at various single RTO systems and the average numbers (red line) are shown in Figure 3.

99

,., "

L

98

!:

97

'(3

96

w

95

Q)

iE

!:

0

94

5

93

:g II) Q)

C

92

Figure 5: Pre-fi Iter System of a FTS (during installation) As the storage volume for particles is limited, the pre-filter has to be cleaned regularly. The purification has to be done "offline", i.e. without normal use of the filter. For the design of a system, a second pre-filter needs to be considered. Figure 5 shows a pre-filter system with four filter units for a flow rate of 135,000 Nm 3 /h (80,000 scfm) where three units are always in operation in parallel, allowing one to be cleaned.

91

conventIonal

K IU

system L0 J

The purification works similar to the so-called "burn-out" of a RTO, where hydrocarbon deposits are gasified via injection of hot air, which leads to an increase of temperature up to more than 400°C (750°F).

These measurements have shown that the achievable destruction efficiency of a single RTO system is limited due to adsorption and desorption effects which are a result of the

l318

Table 1: B(a)P Efficiency of a Optimized 4-chamber RTO during Normal Operation and Burn-out mode

Condensates in RTO In spite of the installation of pre-filters, a considerable amount of condensates can still reach the RTO which is a big challenge for standard RTO systems. Due to an accumulation, especially at the bottom inlet of the ceramic heat exchanger media, the free area for the fumes is clogged and the heat transfer is reduced. The clogged channels of the ceramic media can be seen in Figure 6 as well as the cleaned ceramic media after a burn-out. Another important factor is the increased fire risk due to flammable deposits in the RTO. This risk can be reduced by operating a special cleaning mode, the burn-out.

Operation mode

BaP Input

BaP Output mg/m3

mg/m3 normal normal normal Burn-out Burn-out Burn-out

>

4.68 4.18 2.95 6.24 8.21

0.13 0.19 0.08 0.24 0.18

7.96

0.22

BaP Efficiency % 97.3 95.6 97.3 96.3 97.9 97.3

..

.c.c·:· ...· · .. ·c:.·.

Figure 6: Ceramic Media Before and After Burn-out During this burn-out, hot gas is drawn from the RTO combustion chamber down to the inlet of the ceramic media and all organic deposits, such as sticky hydrocarbons, are gasified. As the gasified deposits have a high calorific value, they will be used as fuel for the system, and will be directed into the oxidation chamber. For this, LTB has developed a special annular gap burner, Figure 7, through which the gases are injected directly into the combustion chamber and completely oxidized. This allows the burn-out to work without increased emissions.

i ...........

Figure 8: 4-chamber RTO with Emission-Free Burn-out Mode The 4-chamber RTO shown in Figure 8 operates according to the following functions, which change during normal operation: Chamber 1: Chamber 2: Chamber 3: Chamber 4:

fumes inlet purging to avoid peaks clean gas outlet burn-out/ purification

Furthermore, the 4-chamber RTO shows another benefit: Since most fume treatment plants have to be run with a very high availability, several installations are operated in parallel to enable maintenance and regular inspections without any reduction of fume treatment capabilities. For example, a baking furnace usually runs for several years without a shut-down, and a bypass is allowed only for a short time. Therefore, one would have to install three conventional 3-chamber RTO units in order to keep full redundancy. For these applications, LTB developed a method to run the fume treatment system with two RTO systems only, even during maintenance of one RTO. These RTOs can be operated in a special mode, which allows one unit to run with double the volume flow for a limited time, the socalled TWIN MODETM shown in Figure 9.

Figure 7: LTB Annular Gap Burner A normal 3-chamber RTO can only run as a 2-chamber RTO during burn-out mode which leads to a reduced cleaning efficiency due to the missing purging cycle. LTB therefore uses a fourth chamber in order to ensure a continuous 3-chamber mode even when one chamber is operated in burn-out mode (emission-free burn-out mode). Measurements at a plant in Germany showed that the B(a)P-emissions of a optimized 4chamber RTO do not increase significantly even during the burn-out mode of one chamber, Table I.

During this new mode of operation, the purging and burn-out will be avoided and two pairs of the four chambers will operate in parallel. As shown in Figure 9, two chambers run with the fume inlet and two with the clean gas outlet providing the double fume treatment capacity.

l319

reaction takes place, the material is transported downwards in a moving bed and discharged. The gases leave the plant cleaned. A number of installations in different industries show good HF abatement with reasonable consumption of limestone. Various Measurements have shown an average consumption of limestone of below 10 kg per kg HF, Table 2 (values based on the use ofBlaubeuren limestone). Table 2: Specific Consumption of Limestone for HF Absorption in a LTB packed bed filter Flowrate

HF inlet

HF outlet (measured)

Figure 9: TWIN MODETM System during double volume flow operation

m3/h S.T.P 25,000

mg/m3 S.T.P