Continuous Casting Edited by K. Ehrke and W. Schneider
Deutsche Gesellschaft für Materialkunde e.V.
Weinheim · New York · Chichester Brisbane · Singapore · Toronto
Dipl.-Ing. Kurt Ehrke ALUMINIUM Essen GmbH Sulterkamp 71 D-45356 Essen Germany
Prof. Dr. Wolfgang Schneider VAW Aluminium AG Forschung und Entwicklung Georg-von-Boeselager-Str. 25 D-53117 Bonn Germany
International Congress Continuous Casting held from 1315 November 2000 in Frankfurt /Main Organizer: DGM · Deutsche Gesellschaft für Materialkunde e.V. Program Committee Dipl.-Ing. Kurt Ehrke, ALUMINIUM Essen GmbH (Chairman) Dr. Hilmar R. Müller, Wieland Werke AG, Ulm Prof. Dr. Wolfgang Schneider, VAW Aluminium AG, Bonn Dipl.-Ing. Gunnar Halvorsen, Elkem Aluminium AG, Oslo (N) Dr. Dirk Rode, KM Europametal AG, Osnabrück
This book was carefully produced. Nevertheless, authors, editors and publisher do not warrant the information contained therein to be free of errors. Readers are advised to keep in mind that statements, data, illustrations, procedural details or other items may inadvertently be inaccurate.
Cover photo: Lüko GmbH Casting Unit for Extrusion Billets VAW aluminium AG AIRSOL VEIL® technology
Library of Congress Card No. applied for A catalogue record for this book is available from the British Library Deutsche Bibliothek Cataloguing-in-Publication Data A catalogue record for this publication is available from Die Deutsche Bibliothek ISBN 3-527-30283-2 © WILEY-VCH Verlag GmbH, D-69469 Weinheim (Federal Republic of Germany), 2000 Printed on acid-free paper. All rights reserved (including those of translation in other languages). No part of this book may be reproduced in any form by photoprinting, microfilm, or any other means nor transmitted or translated into machine language without written permission from the publishers. Registered names, trademarks, etc. used in this book, even when not specifically marked as such, are not to be considered unprotected by law. Composition: WGV Verlagsdienstleistungen GmbH, Weinheim Printing: betz-druck, Darmstadt Bookbinding: Schaumann, Darmstadt Printed in the Federal Republic of Germany
Preface The aim of the conference, organized by the DGM Continuous Casting Committeee, is to highlight the importance of continuous casting of aluminium, copper and magnesium to the international fabricating industry. The conference lectures, generated by the Call for Papers, cover technological advances in all sectors which are important for the manufacture of high quality continuously cast products. Besides melt treatment, casting processes and structure of continuously cast ingots, modelling of casting will be a major topic of the conference. Numerical modelling becomes more and more dominant as a research tool to improve casting processes and the resulting products. The advantages are reduced development times and development costs. The programme of the symposium reflects this with numerous papers dealing with modelling of nucleation, heat and fluid flow as well as stresses and structure. Another new approach of the conference are the supplier sessions. The organizing committee hopes that the conference programme encourages specialists of the non-ferrous industry worldwide to take part in this meeting. K. Ehrke Chairman of the Conference
Contents 25 Years of DGM Continuous Casting Research W. Schneider, VAW aluminium AG, Research and Development, Bonn (D)..............................1 Melt Treatment Hydrogen in Aluminum Containing Copper Alloy Melts – Solubility Measurement and Removal K. Neumann, B. Friedrich, K. Krone, IME Process Metallurgy and Metal Recycling, RWTH Aachen (D) J. Jestrabek, E. Nosch, Schwermetall Halbzeugwerk GmbH, Stolberg (D) .............................15 Fundamental Research About Liquid Metal Filtration B. Hübschen, J. Krüger, RWTH Aachen (D) N. Keegan, Pyrotek Engineering Materials Limited, Dudley (GB) W. Schneider, VAW Aluminium AG, Bonn (D) .........................................................................20 Impact of Grain Refiner Addition on Ceramic Foam Filter Performance N. Towsey, W. Schneider, H.-P. Krug, VAW aluminium AG, Research and Development, Bonn (D) A. Hardman , London & Scandinavian Metallurgical Co. Limited, Rotherham, South Yorkshire (GB) N. Keegan, Pyrotek Engineering Materials Ltd., Netherton, West Midlands (GB)..................26 Review of Dissolution Testing and Alloying Methods in the Casthouse G. Borge, Bostlan, S.A., Larragane, Mungia (E) P. Cooper, S. Thistlethwaite, LSM Co.Ltd., Rotherham (GB) ..................................................33 The Effect of Casting Parameters on the Metallurgical Quality of Twin Roll Cast Strip Y. Birol, Marmara Research Center, Gebze-Kocaeli (TR) G. Kara, A. Soner Akkurt, ASSAN Aluminum Works, Istanbul (TR) C. Romanowski, FATA Hunter Inc., Riverside (USA) ..............................................................40 Casting Technology and Processes – Aluminium Influence of Different Lubricants on the Friction between the Solidifying Shell and the Mould during the DC Casting of AlMgSi0.5 F. Dörnenburg, VAW aluminium AG, Research and Development, Bonn (D) S. Engler, Foundry Institute, Aachen University of Technology, Aachen (D)..........................47 Prediction of Boundary Conditions and Hot Spots during the Start-up Phase of an Extrusion Ingot Casting S. Benum, Hydro Aluminium R&D Materials Technology, Sunndalsøra (N) D. Mortensen, H. Fjær, Institute for Energy Technology, Kjeller (N) .....................................54 Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
VIII Improved Metal Distribution during DC-casting of Aluminum Alloy Sheet Ingots P. Tøndel, J. Hayes, I. Thorvaldsen, Elkem Aluminium, Mosjøen (N) G. Grealy, G. Tahitu, Corus Research Development & Technology, Ijmuiden (NL) E. Jensen, Elkem Research, Kristiansand (N) D. Brandner, Corus Aluminium Walzprodukte GmbH, Koblenz (D) .......................................61 Determination of Material Properties and Thermal Boundary Condition from Casting Trial on Alloy AA7075 J. Rabenberg, J. Storm, Corus Research Development & Technology, Ijmuiden (NL) I. Opstelten, Corus Research Development & Technology, now with TNO Bouw, Delft (NL) J.-M. Drezet, École Polytechnique Fédérale de Lausanne, Lausanne (CH)............................71 Single-roll Strip Casting of Aluminium Alloys E. Straatsma, W. Kool , L. Katgerman, Delft University of Technology, Laboratory for Materials, Delft (NL) ................................................................................................................77 Continuous Casting of Semisolid Al-Si-Mg Alloy T. Motegi, F. Tanabe, Chiba Institute of Technology...............................................................82 Casting Technology and Processes Yield and Quality Improvements for Semi-Continuously Cast Copper Alloys C.-M. Raihle, Outokumpu Process Automation, Västerås (S) ..................................................89 Continuous Casting Technology for Magnesium U. Holzkamp, H. Haferkamp, M. Niemeyer, University of Hanover, Institute of Materials Science, Hanover (D)................................................................................................................94 Local Distribution of the Heat Transfer in Water Spray Quenching F. Puschmann, E. Specht, J. Schmidt, University of Magdeburg, Institute of Fluid Dynamics and Thermodynamics, Magdeburg (D) ........................................................101 Grain Structure, Microstructure and Texture of Copper Ingots Produced during the Continuous Casting Process V. Plochikhine, V. Karkhin, H. Bergmann, Department of Metallic Materials, University of Bayreuth (D)......................................................................................................109 Technologies and Installation for Electrochemical Hardening of Wear Surfaces R. Boiciuc, V. Munteanu, G. Petrache, Uzinsider Engineering S.A., Galati (ROM).............115
IX Modelling – Heat and Fluid Flow; Nucleation Numerical Mass and Heat Flow Predictions in Aluminum DC Casting: A Comparison of Simulations with Melt Pool Measurements A. Buchholz, Corus Research Development &Technology, Ijmuiden (NL) B. Commet, Pechiney Centre de Recherches de Voreppe (F) G.-U. Grün, VAW aluminium AG, Research and Development, Bonn (D) D. Mortensen, Institute for Energy Technology, Kjeller (N) ..................................................123 Investigations of the Primary Cooling in Sheet Ingot Casting H. Fjær, D. Mortensen, Institute for Energy Technology, Kjeller (N) A. Buchholz, Corus Research Development & Technology, Ijmuiden (NL) B. Commet, Pechiney Centre de Recherches de Voreppe (F) J.-M. Drezet, Laboratoire de Métallurgie Physique, EPFL, Lausanne (CH) ........................131 Boiling Curve Approach for Thermal Boundary Conditions in DC Casting J. Zuidema jr., L. Katgerman, Netherlands Institute for Metals Research, Laboratory for Materials, Advanced Materials and Solidification Technology, Delft (NL) I. Opstelten, Corus Research Development & Technology, Ijmuiden (NL)............................138 Theoretical and Experimental Study of Vertical Continuous Casting of Copper M. Uoti, Helsinki University of Technology, Laboratory of Metallurgy, Helsinki (SF) M. Immonen, K. Härkki, Outokumpu Poricopper OY, Kuparitie, Pori (SF)..........................143 Modeling of Grain Refinement in Aluminum Alloys A. Greer, A. Tronche, University of Cambridge (GB) ............................................................149 Modeling of the Grain Refinement in Directionally Solidified Al -4.15 wt.% Mg Alloys using Cellular Automaton – Finite Element Approach M. Vandyoussefi , A. Greer, University of Cambridge (GB) ..................................................154 Modelling – Stress and Structure ContiSim™: Process and Material Modelling of Continuous Casting in Macro and Micro Scale J. Boehmer, Process Modelling and Informatics, Betzdorf/Sieg (D)......................................163 Crystal Growth Morphology during Continuous Casting C. Caesar, Munich (D)............................................................................................................169 3D-Modeling of Ingot Geometry Development of DC-Cast Aluminum Ingots during the Start-Up Phase W. Droste, G.-U. Grün, W. Schneider, VAW aluminium AG, Research and Development, Bonn (D) J.-M. Drezet, CALCOM SA, Parc Scientifique, Lausanne (CH) ............................................175
X The Influence of Casting Practice on Stresses and Strains in 6xxx Billets – A Statistical and Modelling Study B. Henriksen, S. Braathen, E. Jensen, Elkem ASA Reasearch, Kristiansand (N)...................184 Modelling of Macrosegregation in Continuous Casting of Aluminium T. Jalanti, M. Rappaz, École Polytechnique Fédérale de Lausanne, Laboratoire de Métallurgie Physique, Lausanne (CH) M. Swierkosz, M. Gremaud, Calcom SA, Lausanne (CH) ......................................................191 The Effect of the Differencing Scheme on the Numerical Diffusion in the Simulation of Macrosegregation B. Venneker, Netherlands Institute for Metals Research, Delft (NL) L. Katgerman, Delft University of Technology, Laboratory for Materials, Advanced Materials and Solidification Technology, Delft (NL) ............................................199 Application of a New Hot Tearing Analysis to Horizontal Direct Chill Cast Magnesium Alloy AZ91 J. Grandfield, Cooperative Reasearch Centre for Cast Metals Manufacturing, CSIRO Manufacturing Science & Technology, The University of Queensland (AUS) C. Davidson, J. Taylor, Department of Mining, Minerals and Materials Engineering, The University of Queensland (AUS)......................................................................................205 Micro- and Macrostructures Nucleation Studies of Grain Refiner Particles in Al-Alloys P. Schumacher, University of Oxford (GB) ............................................................................213 Effect of Solute Elements on the Grain Structures of Al-Ti-B and Al-Ti-C Grain-Refined Al Alloys A. Tronche, University of Cambridge (GB) A. Greer, University of Cambridge (GB) and Péchiney Centre de Recherches de Voreppe (F) .......................................................................................................................218 Grain Refinement Process in Aluminium Alloys Type AlZnMgZr T. Stuczyn´ski, M. Lech-Grega, Institute of Non-Ferrous Metals, Light-Metals Division, Skawina (PL) ..........................................................................................................224 Coupled Influence of Convection and Grain-refining on Macrosegregation of 1D Upwardly Solidified Al 4.5% Cu P. Jarry, Pechiney Centre de Recherches de Voreppe (F) H. Combeau, G. Lesoult, LSG2M, Ecole des Mines de Nancy (F).........................................233 Tensile Behaviour of DC-cast AA5182 in Solid and Semi-solid State W. van Haaften, W. Kool, L. Katgerman, Laboratory of Materials, Delft University of Technology, Delft (NL) ...........................................................................................................239
XI The Columnar to Equiaxed Transition in Horizontal Direct Chill Cast Magnesium Alloy AZ91 J. Grandfield, Cooperative Reasearch Centre for Cast Metals Manufacturing, CSIRO Manufacturing Science & Technologiy, The University of Queensland (AUS) C. Davidson, J. Taylor, Department of Mining, Minerals and Materials Engineering, The University of Queensland (AUS)......................................................................................245 Study of Heterogeneous Nucleation of α-Al on Grain Refiner Particles during Rapid Solidification P. Cizek, B. McKay, P. Schumacher, University of Oxford (GB)...........................................251 Effect of Instability of TiC Particles on the Grain-Refining Behavior of Al-Ti-C Inoculants in Aluminum Alloys M. Vandyoussefi, A. Greer, University of Cambridge (GB) ...................................................257 Grain Refiners for Thin Strip Twin Roll Casting R. Cook, London & Scandinavian Metallurgical Co. Limited, Rotherham (GB)...................263 Characterisation and Optimisation of Thixoforming Feedstock Material S. Engler, Gießereiinstitut, RWTH Aachen (D) D. Hartmann, EFU Gesellschaft für Ur-/Umformtechnik mbH, Simmerath (D) I. Niedick, Volkswagen AG, Braunschweig (D), (former EFU GmbH)..................................269 Experimental Study of Linear Shrinkage during Solidification of Binary and Commercial Aluminum Alloys D. Eskine , L. Katgerman, Netherlands Institute for Metals Research, Delft (NL) ................276 The Influence of the Cooling Rate on the Type of the Intermetallic Phases in the Aluminium Alloys of the 3XXX (AlMnMgSi) Group T. Stuczyn´ski, M. Lech-Grega, Institute of Non-Ferrous Metals, Light Metals Division, Skawina (PL) ..........................................................................................................................282 Suppliers Session – Aluminium Horizontal Direct Chilled (HDC) Casting Technology for Aluminium F. Niedermair, Hertwich Engineering GmbH, Braunau (A) ..................................................293 Automatic “Bleed Out” Detection and Plug Off in VDC Billet Casting M. Lück, Wagstaff Inc., Spokane WA (USA)...........................................................................300 The AIRSOL VEIL® Technology Package for Aluminium Billet Casting G. Bulian, M. Langen, VAW aluminium AG, Bonn (D) ..........................................................302 The Manufacturing, Design and Use of Combo Bag Distributors in Sheet Ingot Casting S. Tremblay, Pyrotek High-temperature Industrial Products Inc., Chicoutimi (CAN) R. Green, Pyrotek Engineering Materials Ltd., Netherton (GB) ............................................310
XII Recent Quality and Efficiency Improvements Through Advances in In-Line Refining Technology V. Dopp, S. Simmons, Pyrotek, Inc., Tarrytown, New York (USA).........................................316 Suppliers Session – Copper Horizontal Continuous Casting of Copper Alloy Billets M. Brey, SMS Meer GmbH, Demag Technica, Veitshöchheim (D)........................................325 The Outokumpu UPCAST® System L. Eklin, Outokumpu Castform Oy, Pori (SF) ........................................................................333 Author Index .........................................................................................................................341 Subject Index.........................................................................................................................343
25 Years of DGM Continuous Casting Research Wolfgang Schneider VAW aluminium AG, Research and Development, 53117 Bonn, Germany
1
Introduction
Continuous casting experts from the non-ferrous metal industry in Germany, Netherlands, Austria, Norway and Switzerland have taken part in the Continuous Casting Committee of the Deutsche Gesellschaft für Materialkunde DGM since its foundation 1972. During the formation of the committee the different technologies used in DC. casting resp. continuous casting of semis feedstock were taken into consideration. Working groups were therefore formed to deal with the vertical DC. casting of Al and Cu as well as the horizontal casting of Al, Cu and Zn. Over the years, goals were redefined and changes were made to the organisation structure when the industrial requirements made this necessary. In former times the vertical DC casting experts were associated all together in one working group, but due to the Copper DC casting experts being underrepresented, the priority of the work centred on Aluminium DC casting In 1990 an own working group specifically for the DC. casting of Cu was founded. A further important change was the foundation of the Working Group Spray Forming in 1993. It must be mentioned however that not only an expansion of the activities of the Committee took place in the past but also activities had to be stopped and working groups had to be disbanded. Among these, for example, is the working group which dealt many years with the strip casting of Al. The closing of the last strip casting unit in Germany led to this step because the working group saw little reason for a continuation of its work. The actual organisation structure of the Committee Continuous Casting can be seen in Figure 1. The current existing working groups are to be seen in this figure. The activities carried out in these working groups will however not be considered here. One working group will be described in more detail which was not mentioned previously but was an important part of the Continuous Casting Committee for many years.. This is the Working Group Research, as seen in Figure 1. In the following beside the structure and the tasks of this working group the research activities of the past will be described. By means of exemplary results, projects which have been carried out within the framework of the activities of the Working Group Research will be presented. Finally, the future outlook of the working group will be given.
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
2 DGM Committee Continuous Casting
Working Group
Working Group
Working Group
Vertical Casting D.C.
Vertical Casting D.C.
Horizontal Casting C.C.
Aluminium
Copper
Copper / Zinc
Working Group Spray Forming
Working Group Research
Figure 1: Structure of the DGM Continuous Casting Committee
2
Research Working Group: Objectives, Tasks and Structure
The foundation of the Working Group Research took place in the year 1975. Since then, the working group is in nearly unchanged structure part of the Committee Continuous Casting. A major objective of the foundation of the Working Group Research was to interest the universities in continuous casting research and to get it established there over the long term by means of the concerted execution of research projects. This has been successful for nearly two decades. The working group was, and is still today, made in such a way that, besides the members from universities, also members of the different working groups of the Committee also take part. These are the heads of the working groups whose task it is among others, to make suggestions about research projects discussed before in their working groups. The project ideas are discussed with the university representatives concerned and a research proposal is prepared for the submission to relevant funding organisations. The working group normally meets once a year. In these meetings the progress of the actual projects is reported and new project suggestions are discussed. In the beginning the working group had to ensure that adequate equipment was available at the universities for the continuous casting research and it should give support in the procurement of funding for agreed research projects. With regard to the latter it was and will be important in the future that the industry partners were and are concerned in the preparation of the research proposals. At this point, however, it has to be mentioned that the execution and the care for research projects were not alone reserved to the Working Group Research but that there have been carried out also research projects in the other working groups of the Committee. Another important aspect of the work of the working group in the past was, to awaken the interest of the university students in continuous casting. Numerous graduates from universities who carried out the research work within the scope of their doctoral thesis subsequently took up a position in the continuous casting industry on finishing their work. In the beginning the working group´s activities focused on investigations dealing with material science of DC casting. Over the years the activities were expanded with investigations into melt quality and process technology. Correspondingly the university institutes as members of the working group expanded too. In Figure 2 the structure of the
3 Working Group Research is shown and the university institutes are named which were resp. still are long-term members of the working group. Special mention has to be made here of the Gießerei-Institut of the RWTH Aachen where under the management of Professor Engler and with the assistance of the Working Group Research in the mid of the Seventies the first DC casting unit at a university was installed. Unfortunately the unit was dismantled recently and thus is no longer available for future research work. In Figure 3 the number of research projects completed at the universities since 1979 is shown. It can be seen that the number of projects has decreased since 1995, at present no research projects are being carried out within the scope of the working group. The reasons for this will be discussed later. Working Group
Research
Foundry Institute
Heads of Committee Working Groups
RWTH Aachen Prof. S. Engler
Inst. of Non Ferrous Metallurgy RWTH Aachen Prof. J. Krüger
Inst. of Metal Science TU Berlin Prof. W. Reif
Inst. of Energy Technology TU Clausthal Prof. R. Jeschar
Inst. of Metal Research
MPJ Stuttgart Prof. B. Predel, Dr. E. Fromm
Inst. of Material Sciene TU München Prof. H. M. Tensi
Number of Projects
Figure 2: Structure of the research working group
8
Working Group RESEARCH
7
DGM Continuous Casting Committee
6 5 4 3 2 1
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
Year
Figure 3: Number of completed research projects per year since 1975
1997
1999
4
3
Research Working Group: Research Priorities
The list of past research projects of the Working Group Research can be divided into the following subjects: • Melt Quality • Ingot Solidification • Process Technology In the following, an overview of the research projects carried out in these subjects will be given and exemplary results from chosen works presented. Naturally a selection must be made at this point since not all projects can be dealt with. Of course this does not mean that the projects not considered here made no important contribution to the understanding of DC casting. 3.1
Melt Quality (MPI Stuttgart, RWTH Aachen)
As is well known non-ferrous metal melts show complex interactions with the surrounding atmosphere. The centre of interest of the research work carried out was therefore on diluted hydrogen in Al and Cu melts and its measuring. Moreover the oxidation of the liquid aluminium as well as the properties of oxide layers on the melt were investigated. This knowledge is of importance to the handling of melts in the cast house. In further research work the thermodynamic fundamentals of Li removal from primary aluminium as well as investigations into the removal of Na and Li from Al melts were carried out. Within the scope of the research work on the hydrogen solubility in Al and Cu melts, a measuring method was developed which allows the continuous measurement of the hydrogen content by the means of H2 partial pressure measurement. The measurement principles of this method are shown schematically in Figure 4. In this method the probe consists of a graphite disc which is connected to a pressure meter by a gas tight ceramic pipe. After immersing the probe in the melt the probe is quickly evacuated allowing the hydrogen to diffuse into the probe until the pressure in the probe and the hydrogen partial pressure in the melt are equal. The actual hydrogen content is then calculated with simultaneous temperature measurement with the help of Sievert´s equation. An essential advantage of this method in comparison to other methods is that by dosing of the probe with hydrogen the measurement time can be shortened substantially, so that a continuous hydrogen measurement is practicable. The above described method is commercially available today as measurement equipment under the trademark CHAPEL. Within the scope of the investigations on the oxidation behaviour of Al melts the strength of oxide layers as well as the initial stage of the oxidation were investigated. For this a measuring method for the measurement of the strength of the oxide layer was developed, which is shown in a schematic depiction in Figure 5. With this measuring method the maximum torque is determined that the oxide layer can put up a rotating circular stamp in a fixed ring before it tears. In extensive tests the strength of oxide layers on pure aluminium melts as well as on AlMg and AlSi melts with different Mg and Si contents was measured. It could be seen that with increasing age of the oxide layer and with increasing temperature the strength of the oxide layer increases significantly. Moreover, it could be measured, among other things that an enrichment of defined elements increases the strength of the oxide layer. To these elements belong for example Mg, Ca, Na and Li. Their effect decreases with the holding time of the melt.
5 vacuum pump data storage
computer measuring probe
valve 2
data acquisition pressure measuring instrument
temp. measuring device
valve 1 dosing valve
H2
ceramic pipe
graphite disc
thermocouple aluminium melt
Figure 4: Principles of the CHAPEL hydrogen measurement method
The investigation of the initial state of the oxidation of Al melts was carried out at pressures below 10-6 bar. In these investigations it was determined that after skimming of the melt surface, the re-covering time until the formation of a protecting oxide layer with a thickness of about 1 µm was in the time scale of 10-3 to 10-2 sec.
Figure 5: Apparatus for measuring of the strength of oxide skins
3.2
Ingot Solidification (RWTH Aachen, TU Berlin, TU München)
The as-cast structure of the DC cast ingots has as well known considerable influence on their processing and the quality of the subsequently produced semis. DC cast ingots show structural defects which are related to the casting and solidification process. The research
6 work carried out was therefore aimed at investigating the solidification processes in the DC casting of non-ferrous metals. The main focus of interest was, among others the solidification of the sub surface of the DC cast ingots. In several research projects the formation of surface segregations and the stability of solidifying shell zones were investigated. A further priority was the melt treatment for the achievement of a fine equiaxed as-cast grain structure. Here, basic investigations into the understanding of the grain refining mechanism of Al with Titanium and Boron were carried out, new effective grain refiners for Al and Cu alloys were also developed. In additional projects the starting phase of the DC casting of Al rolling ingots and the influence of fluid flow on the as-cast structure in vertical DC casting of Al rolling ingots and of CuSn alloy billets were investigated. In the following, extraordinary results from some of the above mentioned investigations will be presented. The investigations on surface segregation of DC cast ingots with specially developed test moulds showed that the essential transport mechanism for the formation of segregations is the metallostatic pressure of the melt in front of the shell zone, i.e. this is main responsible for the structure defect surface segregation. Beside this, it could also be seen that with increasing thickness and finer as cast structure of the shell zone a significant decrease in the degree of the surface segregations can be achieved. water controller
electronic controller testing procedure lifting system chill
extensometer
shell load-cell
pneumatic closing system
hydraulic cylinder
hydraulic pump
Figure 6: Test equipment for measuring the mechanical properties of solidifying shell zones
In series of interesting investigations the stability of shell zones of Al and Cu alloys solidifying in the mould were investigated. For these investigations a test equipment was developed so that by tensile test the mechanical properties of a solidifying shell zone in dependency of the solid portion in the solidification range could be measured. The developed measuring equipment is schematically shown in Figure 6. Essential items of this equipment comprise the water-cooled copper mould: from this mould the shell zone grows into the melt which is located in an insulating box. A measuring device projects through the insulating form into the hollow space and into the cast shell. This device is connected to a tensile measuring machine. The most important results of these extensive investigations can be summarised as follows. The mechanical properties of solidifying shell zones are on a low level as shown by the examples of an Al alloy and Cu alloy respectively in Figure 7. With increasing solid portion of the shell zone during its duration in the air gap region the tensile strength increases and the elongation of fracture decreases. This is seen in Figure 8.
7 CuSn30
2,5
AlCu4
2,5
547 - 551 °C 850 - 929 °C
2,0
Tensile Strength in N/mm2
2
Tensile Strength in N/mm
2,0
882 - 927 °C 845 - 937 °C 879 - 940 °C
1,5
888 - 937 °C 860 - 938 °C
1,0
Temperature in Shell Zone
854 - 942 °C 870 - 940 °C
0,5
Temperature in Shell Zone
1,5 576 - 583 °C
1,0
581 - 591 °C
597 - 603 °C
0,5
881 - 932 °C
608 - 616 °C
905 - 937 °C
0
0
0,5
1,0
608 - 614 °C
0
1,5
623 - 629 °C
0
0,2
0,4
Elongation in %
0,6
0,8
1,0
Elongation in %
Figure 7: Mechanical properties of solidifying shell zones
99
2,5
AlMgSi0,5 94 91
0,5
2,0
0,4
1,5
0,3
1,0
0,2
0,5
0,1
0
0
600
610 620 Temperature in °C
630
Elongation in %
Tensile Strength in N/mm
2
Tensile Strength Elongation
0
Figure 8: Mechanical properties of shell zones in solidification range
For defined alloys a grain refinement can lead to an increase in the elongation of the solidifying shell zone. As already mentioned the grain refinement is an important melt treatment measure for achieving a fine equiaxed as-cast structure of the DC cast ingots. Among others the deformation behaviour of the billets is improved by an equiaxed grain structure. For the DC casting people, however, it is of particular importance that with the grain refinement the crack formation can be avoided. For Al alloys with the TiB2 containing AlTiB master alloys effective grain refiners are available. Projects carried out in co-operation with the Working
8 Group Research made a substantial contribution to the understanding of the grain refining mechanism of Al with TiB2. For the first time, reports were made on the importance and differing effect of the different alloying elements. Moreover TiB2 could be proved in the grain centre as well as Ti enrichments could be analysed at the borides for the first time. In addition a new grain refiner for Al was developed on the basis of TiC. The same holds true for the grain refinement of Cu alloys. Here the working mechanism of Zirconium as grain refiner could be clarified and the achieved knowledge was used for the development of a Zirconium containing grain refiner master alloy. The effectiveness of the newly developed grain refiners is shown in exemplary manner in Figure 9. T= 700°C th = 5min
700
Average Grain Diameter in µm
Average GrainDiameter in µm
CuSn - Alloy
Al99,7
800
AlTi5B0,2
600 AlTi5B1
500 400 300 200 100
AlTi5C0,25
0
1
2
3
4
E - Cu + m%Zr GT: 1474 K CuSn4 + m%Zr GT: 1523 K CuSn8 + m%Zr GT: 1473 K
1600
1200 Holding Time 1,5min
800
400
5
Grain Refiner Addition Rate in kg/t
0
0
0,02 0,04 0,06 0,08 0,10 Zr Concentration in m%
Figure 9: Grain refinement of aluminium and copper alloys
It has been mentioned several times that within the scope of the research projects carried out to some extent new test techniques were developed to facilitate work on the set task. To these belongs a method which has been developed for the measurement of the butt curl of Al rolling ingots during the start up phase. In this method steel wires are cast into the narrow sides of the ingot butt. These are connected with at the starter block mounted inductive linear transducers so that the progress of the butt curl can be recorded during the start-up phase. With the recorded data it is then possible to evaluate the curling speed and to determine the maximum butt curl. The above described measuring method proved to be very effective and reliable and is nowadays used for butt curl measurements world wide. 3.3
Process Technology (TU Clausthal, RWTH Aachen)
Besides the metallurgical aspects of continuous casting the casting technology is also of importance for the quality of the cast products. Due to the different products different technologies in vertical and horizontal casting are used. In the centre of the research activities of the Working Group Research concerning the process technology were the direct cooling in DC casting and the processes occurring in the mould used for the horizontal DC casting of Cu alloys. Additionally a project was carried out to investigate the working mechanism of lubrication in DC casting of A alloys. Direct cooling in DC casting by impinging of water onto the hot ingot surface is a complex process. According to the metal being cast different cooling techniques are used. In the scope of long-term research the different cooling techniques and cooling conditions were simulated in laboratory-scale tests and, on the basis of the subsequent test results, calculation models
9 were elaborated. In the projects carried out the influence of the most important parameters of spray water cooling as used in the DC casting of Copper and the film cooling as is used in the DC. casting of Aluminium were investigated. The composition of the cooling water also formed part of the investigation. During the cooling of hot metal surfaces with liquids, different stages of cooling are gone through. These are obvious in Figure 10. Of these the stable film boiling stage is unwelcome because it leads to a reduction in the heat extraction and as a result to a slower overall solidification of the ingot. In the investigations on spray and film cooling stable film boiling and the Leidenfrost temperature were of particular interest. The Leidenfrost temperature is the temperature of a hot surface at which the evaporation time of the liquid in contact with the metal surface is the longest, and the extracted heat amount per time interval in dependence on the surface temperature is at a minimum. The investigations into spray cooling showed that the heat transfer in the stable film boiling stage can only be influenced by the water flow rate. The Leidenfrost temperature is moved to higher temperatures with increasing water flow rate while the heat transfer below the Leidenfrost temperature additionally depends on the thermophysical material data and the surface roughness of the ingot to be cooled. The investigations into film cooling showed similar results as those reported on above. It could be measured too that, with increasing exit speed of the cooling water at the gate of a slotted nozzle, the heat extraction is increased during the stable film boiling stage and the vapour film thickness is decreased. That means the vapour film breaks down earlier and the Leidenfrost temperature is moved to higher values. At this point it must be mentioned that stable film boiling only takes place in DC casting of steel while in DC casting of Cu the partial film boiling stage is also of importance. In DC casting of Al, however, the partial or stable film boiling can only be watched during the start-up phase, especially of rolling ingots, while in the stationary casting phase nucleate boiling is the dominant cooling mechanism because of the actual surface temperatures.
q in W/m
2
10
B
6
C
10 5
1. free convection 2. nucleate boiling 3. partial film boiling 4. stable film boiling
A 10
4
2
1 10
3
1
10
3
1
w
4 10
sa
2
10
3
in K
Figure 10: Heat flux density as a function of temperature difference between surface and water
In continuing investigations into the influence of the cooling water composition on the cooling of hot surfaces it could be detected that the influence of the salt content (e.g. NaCl, KCl) is interconnected with the gas content (O2, CO2) of the water. Due to the results of tests and theoretical considerations it could be concluded that with decreasing salt content the gas content increases. This results in a more stable vapour film and leads to a decrease of the
10 Leidenfrost temperature. Because of the aforementioned results a continuation of the investigations into the influence of lubricants in cooling water seems to be necessary. Finally, investigations into the horizontal DC casting of Cu alloys will be considered shortly. Within the scope of these investigations the solidification behaviour of different Cu alloys was examined. By means of temperature measurements in the mould the influence of different casting parameters on the heat transfer conditions was determined. Furthermore, the influence of the withdrawal parameters of the ingot on the friction mechanisms acting between shell zone and mould was determined by making force measurements. On the basis of the achieved knowledge, solidification models have been developed with which the as-cast structure quality of horizontal DC cast Cu ingots can be optimised.
4
Research Working Group: Outlook
The working Group Research is currently in a difficult situation. In Figure 3 it has been shown how the number of research projects has dropped significantly in the last years. For three years no more publicly funded projects have been carried out at the member universities under the umbrella of the working group. The few activities which are still running are bilateral projects with the industry and therefore not accessible to the public. The decrease in number of the projects is mainly related to the fact that it is more and more difficult to find project financing: in comparison to former times, significant reduced financial resources are available for the funding organisations. This has the consequence that markedly fewer research projects can be funded or longer waiting periods have to be accepted before the start of funding of the approved projects. This often has the consequence that applications are either withdrawn or not made at all because years can pass before work can begin. A further problem seems to be that the importance of continuous casting is no longer recognised. This means that suitable projects are not given the necessary priority and therefor are very often not considered for public funding. To overcome the aforementioned problems it was tried some time ago to found a private funding organisation for DC casting research with industry members. The aim was to fund research projects at the universities with the membership fees. Unfortunately the foundation of the society did not take place because the interest of the industry was not sufficient enough to get the necessary seven foundation members together. The difficult financial situation concerning project funding also decreases the interest of the universities in continuous casting research. A further problem is that a change of university members due to age considerations takes place frequently and it is hard to interest new members in continuous casting research under the circumstances described above. For a survival of the continuous casting research and of the work of the Working Group Research of the Committee Continuous Casting new efforts are required. One possible solution could be a further expansion of the Working Group Research at European level. By further internationalisation of the working group, additional funding programs for the financing of research projects could be claimed. Furthermore the importance of continuous casting would be promoted more and the funding of projects by funding organisations would be facilitated. In addition, this could mean more direct funding of research projects by the industry at the universities and the idea of the private funding organisation of continuous casting research could be taken up again.
11
5
Research Working Group: Selected Publications
Melt Quality [1] W. Kahl and E. Fromm: Examination of the Strength of Oxide Skins on Aluminium Melts, Met. Trans. 16B (1985) S. 47-51 [2] E. Fromm: Bestimmung der Wasserstoffkonzentration in Al-Schmelzen durch eine kontinuierliche Messung des H2-Gleichgewichtsdruckes, Aluminium 65 (1989), S. 12401243 [3] X.-G. Chen and S. Engler: Measuring Hydrogen Content in Molten Aluminium Alloys using the CHAPEL Technique, Cast Metals 6 (1993) 2, S. 99-108 [4] F. Patak: Untersuchungen zur Natrium- und Lithiumentfernung aus Hüttenaluminium, Dissertation. RWTH Aachen 1983 Ingot Solidification [5] R. Ellerbrock und S. Engler: Oberflächenseigerungen von Stranggußlegierungen, Metall 37 (1983), S 784-788 [6] L. Ohm und S. Engler: Festigkeitseigenschaften erstarrender Randschalen aus AlLegierungen, Gießereiforschung 42 (1990) 3, S.131-147 und 4, S. 149-162 [7] W. Droste und S. Engler: Vorgänge beim Angießen von Al-Walzbarren im Stranggießverfahren, 8. Internationale Leichtmetalltagung, Wien, 1987 [8] C.H. Dickhaus und S. Engler: Mechanische Eigenschaften erstarrender Randschalen aus Al- und Cu-Legierungen, in Stranggießen, DGM-Informationsgesellschaft GmbH 1995, S 55-66 [9] W. Schneider und W. Reif: Untersuchungen zur Deutung der Vorgänge bei der Kornfeinung von Aluminium mit AlTiB-Vorlegierungen, Gießereiforschung 32 (1980) S. 53 [10] A. Banerji and W. Reif: Development of AlTiC Grain Refiners Containing TiC, Metall. Trans. 17A (1986), S 2127 [11] R. Mannheim und W. Reif: Kornfeinung von CuSn-Legierungen mit Zirkon, Bor und Eisen sowie CuAl-Legierungen mit Titan, Bor und Zinn., in Erstarrung metallischer Schmelzen, Deutsche Gesellschaft für Materialkunde 1981, S. 109-140 Process Technology [12] H.R. Müller und R. Jeschar: Wärmeübergang bei der Spritzwasserkühlung von Nichteisenmetallen, Zeitschr. f. Metallkunde 74 (1983), S. 257-264 [13] C. Köhler, E. Specht and R. Jeschar: Heat Transfer with Film Quenching of Vapourizing Liquids, Steel Research 61 (1990) 11, S. 553-559 [14] H. Kraushaar, H. Griebel und R. Jeschar: Einfluß der Kühlwasserqualität auf den Abkühlvorgang heißer Oberflächen, in Stranggießen, DGM-Informationsgesellschaft GmbH, 1995, S 231-240 [15] D. Hartmann und S. Engler: Erstarrungsverhalten von Cu-Legierungen beim horizontalen Stranggießen, Metall 46 (1992) H.2, S. 139-144 und H.4, S. 333-340
Melt Treatment
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
Hydrogen in Aluminum Containing Copper Alloy Melts – Solubility, Measurement and Removal Karsten Neumann, Bernd Friedrich, Klaus Krone IME Process Metallurgy and Metal Recycling, RWTH Aachen, Aachen, Germany
Jürgen Jestrabek, Elmar Nosch Schwermetall Halbzeugwerk GmbH, Stolberg, Germany
1
Introduction
Aluminum containing copper alloys show a significant hydrogen pick up during melting which is caused by reduction of humidity from various sources like atmosphere, scrap and melt covering agents. Reaction products are aluminum oxide and hydrogen, which is easily taken up by the melt. In combination with the solubility change during solidification, gas porosity may arise in cast products if critical limits of hydrogen concentration in the melt are exceeded. Currently, major quantities of CuAl5Zn5Sn1 (Nordic Gold) strip are produced for the fabrication of Euro currency coins. Some production lots show material defects caused by gas pores. The current project identifies sources for increased hydrogen contents and investigates the possibility of a quality improvement by removing increased hydrogen contents from CuAl5Zn5Sn1 melts.
2
Hydrogen Solubility
In order to predict expected hydrogen contents in the melt, the hydrogen solubility in copper melts and the influence of the main alloying elements on the solubility have to be considered. In copper alloys, the alloying elements aluminum, zinc and tin all lower the hydrogen solubility in both solid and liquid state [1,2]. However, no consistent thermodynamic data set is available to describe the effects of these elements in dependence of concentration and temperature. A previously published investigation [2] contains data to estimate solubility for one temperature in each solid and liquid state. The solubility is calculated by adding up the influences of the alloying elements. (Table 1). This simplification is limited to low contents of alloying elements. Table 1. Influence of alloying element i on hydrogen solubility in mass-ppm H per mass-% i [1] Alloying Element i 700 °C 1150 °C Aluminum -0,06 ppm / [% Al] -0,52 ppm / [% Al] Tin -0,01 ppm / [% Sn] -0,22 ppm / [% Sn] Zinc -0,01 ppm / [% Zn] -0,22 ppm / [% Zn]
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
16 Using this data, the hydrogen contents in copper and CuAl5Zn5Sn1 at 700 °C and 1150 °C in equilibrium with water vapor have been calculated (Table 2). Partial pressures of 1013 hPa and of 25 hPa as present in usual atmospheres have been considered. Table 2. Estimated hydrogen solubilities for pure copper and CuAl5Zn5Sn1 in mass-ppm Material Partial pressure of water 700 °C 1150 °C 0,56 ppm 6,4 ppm p H O = 1013 hPa Copper, oxygen free 0,09 ppm 1,0 ppm p H O = 25 hPa 0,20 ppm 2,48 ppm p H O = 1013 hPa CuAl5Zn5Sn1 0,03 ppm 0,39 ppm p H O = 25 hPa 2
2
2
2
From binary alloy solution data, a solubility curve for hydrogen in CuAl5Zn5Sn1 has been estimated. The effect of the melting interval has been included (Figure 1).
Figure 1: Equilibrium solubility of hydrogen in CuAl5Zn5Sn1 in equilibrium with 25 hPa total water pressure
3
Methods for Hydrogen Determination
Hydrogen can be determined in copper alloy melts either by in-situ measurement or by sampling and subsequent analysis of the samples. The major advantage of the in-situ measurement is the elimination of hydrogen losses during casting and solidification of the samples. Two principles for in-situ-determination are currently available. One method already in use in the steel industry works by contacting a well defined volume of carrier gas with the melt; this results in an equilibrium hydrogen content in the carrier gas corresponding to the hydrogen content of the melt which can be measured by thermal conductivity. This method requires expensive disposable probes for each measurement. The second method measures the electromotive force between a reference material and the melt with a proton-conducting ceramic serving as the electrolyte. However, this method has not yet been introduced into industrial practice and it is still uncertain whether the electrolyte is resistant to the alloying elements contained in copper alloy melts. In the present work, for technical and economical reasons it was decided to use the classical two step method with sampling/solidification and subsequent analysis of samples.
17 3.1
Sampling Method
A major requirement for reproducible hydrogen determination by sampling is the rapid solidification and further cooling of samples to avoid hydrogen losses during sampling. The samples can be cast into moulds or be taken using disposable sampling probes of different shapes which are immersed into the melt. The advantages of casting samples are low sampling costs and rapid solidification caused by the heat capacity of the mould. A major disadvantage is the possible influence of different mould temperatures on the solidification speed and thus on the hydrogen contents determined. Disposable samplers may offer a higher reproducibility concerning thermal conditions and by avoiding the casting step. However, due to the insufficient surface quality of the samples and the slower solidification inside the thermally insulating probes, the cast samples typically deliver better results. Therefore, samples cast into a steel mould have been used in the current project (Figure 2).
70 mm
Sample ∅ 6 mm ca. 4 g
10 mm
Figure 2: Sample shape used
Water has been used for quenching the samples after solidification. Due to the rapid cooling (> 500 K/min), a hydrogen pick up by decomposition of water can be neglected. Liquid nitrogen has also been tested, but led to slower cooling rates due to an isolating gas film forming on the sample surface (Leydenfrost phenomenon), which caused significant hydrogen losses during cooling (Figure 3).
Figure 3: Effect of water and nitrogen quenching on hydrogen content determined in CuAl5Zn5Sn1
3.2
Analysis Procedure
The samples are cut into pieces 2-5 g of weight and are degreased using organic solvents. The hydrogen content of the samples is determined by carrier gas hot extraction, using a LECO RH-402 analyzer with an extraction time of approx. 5 minutes at 800 °C. The equipment is
18 calibrated daily by gas dosing and blank measurement. Hydrogen determination by melt extraction has also been tested, but has not been used as standard procedure because of the increased zinc evaporation during the extraction which leads to contamination of the analysis equipment and is less reproducible.
4
Influences on Hydrogen Contents
As the production batches differed substantially in hydrogen content, investigations have been made to determine possible correlations between environmental or process parameters and hydrogen contents of melts. Atmospheric water vapor and humidity of scraps are considered to be the main source for hydrogen pick up. According to the reaction gas 1 + 13 Al = H+ 61 Al2 O3 ( H,Al:species dissolved in the melt ) 2 H 2O the decomposition of water vapor leads to increased hydrogen contents in the melt. In the present study, the hydrogen contents determined in CuAl5Zn5Sn1 melts show a significant correlation with the atmospheric humidity at cast time (Figure 4).
Figure 4: Correlation between atmospheric partial pressure of water vapor and hydrogen contents in CuAl5Zn5Sn1 melts
Lubrication agents contained in chips and other scrap materials may be another possible source of hydrogen. However, a significant correlation between the amount of humid scraps and the hydrogen contents has not been determined yet. Further investigations in this respect are in progress.
5
Production Scale Testwork
Many of the material defects occurring in CuAl5Zn5Sn1 coin strip production have to be attributed to increased hydrogen contents of the melts. Because the main source for hydrogen
19 is the atmosphere, hydrogen pick up is more or less inevitable in normal melting practice. Thus, an additional melt treatment step for hydrogen removal is required. In general, hydrogen can be removed either by a vacuum treatment or by inert gas purging, the latter being the only feasible alternative for economical reasons. Gas Purging with nitrogen using an impeller injection system (Foseco) was chosen as the best suitable method. The purging operation is carried out in a 40 t short coil, crucible type induction furnace (diameter: 1,80 m; impeller immersion depth: 1,50 m). Currently, in a 30 minute treatment with 40 l/min of Nitrogen, a decrease in hydrogen content of 0,15-0,2 ppm is achieved. The hydrogen content in CuAl5Zn5Sn1 melts can be reduced below a critical value, and the quality of the rolled product can be significantly improved. Meanwhile, all production lots are treated by this gas purging operation.
6
Conclusions
Different possibilities for determination of hydrogen in zinc and aluminum containing copper alloy melts have been evaluated. A sampling procedure has been developed which allows reproducible determination of hydrogen in CuAl5Zn5Sn1 (Nordic Gold) melts by carrier gas hot extraction using a LECO RH-402 analyzer. Atmospheric water vapor has been identified as a source of increased hydrogen contents found in production melts of CuAl5Zn5Sn1. Due to this fact, a melt treatment step for hydrogen removal is advisable. Production scale tests show that impeller gas purging of CuAl5Zn5Sn1 melts in a 40 t induction furnace is an effective way to reduce hydrogen content to an acceptable level. Meanwhile, all production lots are treated with the gas purging operation developed. Further optimization is still required to enhance the reproducibility and stability of the process.
7
References
[1] R.O. Thomas, S. Harper, J.E. Bowers, Gaseous and Gas-forming Elements in Copper and Copper Alloys, International Copper Research Association, Inc., New York 1983 [2] E. Fromm, E. Gebhardt, Gase und Kohlenstoff in Metallen, Springer-Verlag, Berlin, Heidelberg, New York 1976 [3] K. Neumann, B. Friedrich, K. Krone, Wasserstoffgehalte in aluminium- und zinkhaltigen Kupferlegierungen, in NiM 2000 (Ed. D. Hirschfeld), Schriftenreihe der GDMB, Clausthal 2000
Fundamental Research About Liquid Metal Filtration Bettina Hübschen1, Joachim G. Krüger1, Neil J. Keegan2, Wolfgang Schneider3 1
RWTH Aachen, Aachen, Germany Pyrotek Engineering Materials Limited, Dudley, UK 3 VAW Aluminium AG, Bonn, Germany 2
1
Introduction
Increased quality demands for aluminum have led to the fact that in today’s casthouses filtration of molten aluminum has become a standard operation. For filtration the use of Ceramic Foam Filters (CFF) is a common method to remove inclusions. Deep bed filtration is considered as a dominant filtration mode in Ceramic Foam Filters. Inclusion capture in deep bed filters is the result of two sequential events: transport of the particle to the filter wall and attachment of the particle at the wall [1]. For both events the fluid flow in the channels of the filter is a very important parameter. As inclusions are generally smaller than the pore sizes of the filters used, they are deposited on the pore walls of these filters. So already detached particles can be carried away with the liquid and washed out of the filter during hydrodynamic perturbations, for example if flow velocity rapidly changes or if sudden vibrations occur [2]. The aim of this work was to investigate the flow behavior inside a CFF as a function of flow rate, pore size, pore shape and flow direction. For these investigations two different water models were used.
2
Fundamentals
In Ceramic Foam Filters the liquid has to find its way through a tortuous path of pores connected by channels. Since inside the filter the pore size changes, also the velocity of the melt changes and there are region of turbulent and of laminar flow. In regions of laminar flow, a deposition of particles can occur while in turbulent flow regions very small particles can agglomerate and then be attached in the laminar flow regions. However rapid increases in flow velocity due to perturbations can lead to the fact that previously laminar flow regions will become turbulent and already detached particles will be released as a consequence. Within the pores of the CFF different flow conditions and mixing behavior can occur, depending on flow velocity. The CFF can be regarded as a continuous reactor and the different volume fractions can be determined to characterize the flow behavior inside the filter. One important characterization of the fluid flow is the average residence time which is given by equation 1 [3]. V t= (1) v& where t : average residence time, V: volume of fluid in the vessel, v& : volumetric rate of fluid flow.
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
21 Usually the residence time in reactors differs from the calculated residence time and a variety of different residence times can be found. This means that some fluid elements spend a longer and others a shorter period of time in the system. This distribution of residence times is an important characteristic and describes the performance of a reactor. Tracer measurements are a tool to investigate the flow behavior. There are several methods for introducing tracer material into a system of which pulse input of tracer was used in this work. This involves the feeding of a quantity of tracer over a short time period into the system. The tracer material must not interact with the fluid in the reactor or the reactor itself and the amount of tracer has to be negligible in comparison with the amount of fluid present in the system. In addition the time period over which the tracer is introduced to the system has to be very small compared to the calculated residence time. The concentration of the tracer in the outlet stream is measured and plotted against time. The results can be plotted in dimensionless and therefore more general form by using the variables (C-Diagrams) [3]: c C= (2) Q V where Q = quantity of tracer injected, V = Volume of fluid in the vessel, C = dimensionless exit concentration. And t (3) Θ= t where: Θ = dimensionless time, t = actual time, t = average residence time. The area under each C curve must be unity since all the tracer introduced to the system must eventually leave the system. ∞
∫ CdΘ ≡ 1
(4)
0
According to the distribution of residence times it is possible to define different reactor types of which the C Diagrams have a typical design. This is to be seen in figure 1.
Figure 1: Basic reactor types: left: plug flow, middle: backmix flow, right: presence of dead volume [3]
In the case of plug flow the tracer elements introduced to the system do not mix at all while they pass the reactor and arrive at the outlet exactly at Θ = 1. So there is no spread of residence time. (figure 1,left). In a backmix flow reactor the tracer is dispersed immediately and uniformly throughout the system. This means the tracer concentration in the outlet stream is equal to the concentration inside the reactor. Thus the C diagram shows a decrease in tracer
22 concentration starting from unity during the test run. This means that a fraction of the tracer stays inside the system for a time much longer than expected while another fraction passes the system much quicker (figure 1, middle). The presence of dead volume regions is indicated by a maximum in the C diagram at a time smaller than the average residence time and C > 1 (figure 1, right). The volume of the reactor seems to be much smaller than it actually is. Real reactors usually are a mixture of plug flow, backmix and dead volume. The C diagram for such a mixed model is shown in figure 2. The volume fractions can be determined from the diagram.
Figure 2: Determination of the volume fractions in a mixed model [3]
Figure 3: Full Scale filter box model
FWhile in most cases dead volume decreases the performance of a rector, for filtration a certain amount of dead volume is essential for the deposition of particles.
3
Experimental Setup
For the testwork two different model types were used. The first one was a full scale filter box model. Tracer tests on real CFF were made to investigate the change of flow behavior with the flow rate and filter pore size. The second water model type used was a specially designed single channel model to simulate the flow in one channel of a CFF. Three different single channel models were tested to investigate the influence of pore shape and flow direction on the flow behavior. The setup of the models used for the investigations is shown in figures 3 and 4. For all tests sulphuric acid was used as tracer and also KMnO4 was added for visualization of the flow. A small amount of tracer was added at the inlet of the filter/model and an electrode at the outlet measured the electric conductivity which gives an immediate signal for any change in concentration. The signal was recorded by a computer and thus the concentration-time-curve could be plotted. A valve in the outlet stream allowed different flow rates to be adjusted.
23
Figure 4: Experimental setup and design of single channel models
4
Results
4.1
Single Channel Models
The C-diagrams for each model were evaluated and the different volume fractions determined. Thus the influence of flow rate could be presented in a flow rate-volume fraction diagram. The first single channel model tested consisted of a sequence of spherical pores (model 1). The results of the different volume fractions measured are shown in figure 5. 100
100
Vp
90
Vd
Vm
Vp
90
80
80
70
70
60
60
50
50
40
40
30
30
20
20
10
10
Vd
Vm
0
0 0
0,2
0,4
0,6
flow velocity, mm/s
Figure 5: Model with spherical pores
0,8
1
1,2
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
flowvelocity, mm/s
Figure 6: Model with elongated pores
The amount of dead volume was very high in this model. With rising flow rate however the amount of dead volume decreased and plug flow volume increased. Due to the fact that the pore shape of this model did not correspond to the pore shape of CFF’s, a second model with elongated pores was tested (model 2). The results are shown in figure 6. In this model the amount of dead volume was generally smaller than in the spherical pores model. At the same time the variation of the results was higher.
24 The third model tested finally considered the change in flow direction the fluid experiences during its passage through the CFF. So a specially designed tortuosity model was constructed (model 3) and the results of the measurement are to be seen in figure 7. 100
100 90
Vp
Vd
Vm
80
80
70 60
60
50 40
40
30 20
20
10 0 0
0,2
0,4
0,6
0,8
1
1,2
flowvelocity, mm/s
0 10
15
20
25
Vd
Figure 7: Volume fractions as function of flow rate
30
35
40
FlowRate, gal/min Vm
Vp
Figure 8: Volume Fractions in a 80 ppi CFF
for tortuosity model
While in the model with spherical pores the amount of dead volume was much higher than the amount of mixing volume, now both parts are similar up to a certain flow rate. Then mixing volume increases rapidly and dead volume decreases at the same time. 4.2
Full Scale Model
In analogy to the single channel model tests, the volume fraction distribution in real CFF’s was measured. Figure 8 shows the results for a 80 ppi filter. In this CFF the amount of dead volume was smaller than in the single channel models. The amount of dead volume is smaller than the amount of mixing volume even for very small flow rates. From a certain flow rate on, the amount of mixing volume decreases rapidly and dead volume increases at the same time. This indicates highly turbulent conditions inside the filter. This also shows a very good correlation to the results of the single channel models. 4.3
Tests with Inclusions
Another test series with the single channel model involved the addition of Al2O3 particles to the model and the observation of their flow behavior as a function of flow rate. Results of the elongated pores model for small and high flow rates are shown in figures 9 and 10. While in figure 9 the particles flow down the model without any turbulences, in figure 10 the particles are circling in one pore for several seconds. In this case no dead volume was available to make a deposition of particles possible.
25
Figure 9: Particle flow in the elongated pores model for low flow rates
5 • • • • • • •
6
Figure 10: Particle flow in the elongated pores model for high flow rates
Conclusions The single channel models used showed an increase in mixing volume when flow velocity increased At the same time dead volume - which is essential for the deposition of particles decreased. Also the shape of the pores was found to be very important. The amount of dead volume was higher if the pores were spherical. If the liquid changes its flow direction, the amount of dead volume is smaller and decreases more rapidly with flow rate. In the real CFF the amount of dead volume was generally smaller than in the single channel models In the CFF the amount of dead volume decreased when flow velocity increased. If there are turbulent conditions inside the pores and no dead volume is available the deposition of particles becomes very unlikely.
References
[1] Eckert, C. E.; Miller, R. E., Molten Metal Filtration: Fundamentals and Models, Light Metals 1984, 1281-1304 [2] Desmoulin, J.-P., Reliability of molten metal filtration, Light Metals 1992, 1093-1099 [3] Szekely, J.; Themelis, N., Rate Phenomena in Process Metallurgy; Wiley-Interscience, 1971
Impact of Grain Refiner Addition on Ceramic Foam Filter Performance Nicholas Towsey1, Wolfgang Schneider1, Hans-Peter Krug1, Angela Hardman2, Neil J. Keegan3 1
VAW aluminium AG, Research and Development, Bonn, Germany London & Scandinavian Metallurgical Co.Limited, Rotherham, South Yorkshire, UK 3 Pyrotek Engineering Materials Ltd., Netherton, West Midlands, UK 2
1
Abstract
An extensive program of work has been carried out to evaluate the efficiency of ceramic foam filters (CFF’s) under carefully controlled conditions. The first phase of this work, reported at previous international meetings, showed that ceramic foam filters have the capacity for high filtration efficiency and consistent, reliable performance. The next phase of the program was to study their performance under conditions closer to those in production. Work was carried out to establish the influence of grain refiner (Al-3%Ti-1%B) additions made before the filter. The evaluation program was again conducted using AA 1050 alloy and metal quality was, as before, determined using LiMCA and PoDFA. Spent filters were also analyzed. In order to better understand the impact that a grain refiner addition has on filter performance, trials were also undertaken using specially produced commercial purity aluminum and Al-0.7%Ti rods.
2
Introduction
The performance of ceramic foam filters (CFF’s) under simulated production conditions has been studied extensively for an AA1050 alloy as reported previously1-3. In these studies, grain refiner was deliberately omitted in order that a baseline understanding of filter performance could be established. Under such conditions, LiMCA results showed that ceramic foam filters could give mean filtration efficiencies comparable in range to those of competitive filter systems such as bed filters and rigid media filters. The mechanism of filtration appeared to be associated with the formation of `bridges´ of inclusions across the `windows´ of the ceramic foam pore structure in 50ppi and finer filters. In the work presented here, the impact of a grain refiner addition (in rod form) on melt quality and ceramic foam filter performance has been studied for 3:1 TiBAl (Al-3%Ti-1%B) grain refiner and 50 ppi filters.
3
Experimental Procedure
Trials were carried out at the specially dedicated production scale R&D unit at VAW’s Rheinwerk plant. An AA 1050 alloy, batched using reduction line metal, was cast into ingots
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
27 by the direct chill process at a flowrate of 10 tonne/hr, whilst metal quality measurements were made in the launder. A schematic of the experimental layout employed for this work is shown in Figure 1. Grain refiner rod was fed via a guide tube into the melt as far as possible upstream of the filter box. The time available for the rod to dissolve before the metal entered the filter was approximately 1.5 minutes. LiMCA was the main technology used for assessing inclusion concentrations. PoDFA was used as a back up, with the added benefit of providing particle identification capabilities. Spent filters were also assessed for each trial. The majority of this work was carried out with 3:1 TiBAl rod as normally used in production. A number of additional trials were devised to study the impact of grain refiner on melt cleanliness in more detail. The possible `mechanical´ effects of a rod addition, such as vibration and oxide pull-in at the point of entry, were addressed by feeding a rod of commercial purity Al into the melt. A binary Al-Ti rod containing 0.7% Ti (approximately the same amount of `free´ Ti as that found in a 3:1 TiBAl grain refiner) was used to study the dissolution of TiAl3 platelets along the length of the launder in the absence of TiB2 particles. All the rod material used in this program was evaluated metallographically. In most cases the rod alloys were fed at a rate of 1kg/tonne, slightly higher than that for routine production to deliberately intensify any effects that might occur. For most trials the rod was only entered into the melt partway through the cast after about 20-30 minutes of stable LiMCA readings, to clearly highlight the effect of the rod addition on LiMCA counts. To simulate `real´ casting conditions, additional trials were conducted where the rod was fed from the start of the cast. Furnace LiMCA/PoDFA Rod feed
±4.5m
LiMCA/ PoDFA
Mold
Ceramic Foam Filter LiMCA/ PoDFA
Figure 1: Pilot plant schematic
The `loading´ on the filter, here defined as the inclusion content of the AA1050 metal flowing from the furnace, was varied by a stirring (high load) or settling (low load) practice. The impact of sustained high loading throughout casting was investigated by stirring the furnace during casting. Finally, trials were conducted without a CFF where LiMCA was used to evaluate the effects of TiAl3 dissolution rate. Here three LiMCA units along the length of the launder system were used to monitor the changes in particle concentration with increased distance from the rod insertion point. For all categories of trials, replicate casts were made to confirm reproducibility of the results.
28
4
Results and Discussion
4.1
Impact of a 3:1 TiBAl Rod on the Efficiency of a CFF
Figure 2 shows the impact of a 3:1 TiBAl addition on the performance of a CFF when the incoming inclusion loading is high. It can be seen that before the grain refiner rod was introduced (after 28mins) the high level of filter efficiency found in the previous trial series for the 50ppi CFF’s was confirmed. The CFF has once again removed the vast majority of the incoming inclusion loading. After the 3:1 TiBAl rod is introduced a steady rise in the inclusion value can clearly be seen exiting the filter. In order to get a more realistic appraisal of this effect the grain refiner was also fed from the start of some casts (Figure 3). 50ppi CFF, 3:1 T iBAl 1kg/T , stirred before cast 14
12
Before filter
Impurity Level N15 (k/kg)
10
After filter 8
6
4
ROD IN 2
0 0
10
20
30
40
50
60
70
Ca sting Time (min.)
Figure 2: 3:1TiBAl rod – fed partway through cast – “high” inclusion load 50ppi CFF, 3:1 T iBAl in at start of cast, stirred 14
12
Impurity Level N15 (k/kg)
10
8
6
Before filter After filter 4
ROD IN 2
0 0
10
20
30
40
50
60
70
80
Casting Time (min.)
Figure 3: 3:1TiBAl rod – fed from the start of casting – “high” inclusion load
Figures 3 and 4 indicate that the overall efficiency for this cast and the efficiency across the inclusion size distribution range are decreased compared to casts where grain refiner was not used1-3. These efficiency levels are now more consistent with those reported previously under production conditions4. More significantly the downstream cleanliness levels for the cast in Figure 3 were around 2 – 3k/kg as opposed to the 0.2 – 0.3k/kg level without grain refiner. The observed effect was found to occur mainly in the 15 – 35 µm size range.
29 To appreciate the magnitude and implication of this post filter effect, it is necessary to consider the pre filter effect of adding grain refiner rod. This is best seen on a well settled melt where the effect of adding the rod results in an increase in N15 of 0.5 – 1.5 k/kg – Figure 5. Comparing Figure 2 and Figure 5, it is believed reasonable to assume that the pre filter effect when the loading is `high´ is of a similar order of magnitude to when it is `low´. The post filter effect when the loading is `high´ (4-6k/kg) is thus `disproportional´ to the pre filter effect (0.5-1.5k/kg). When the loading is low as in Figure 5, the post filter effect is much milder. This very fact suggests that it is not the agglomeration of <5µm borides alone (and their subsequent release) that has caused the effect. The particles exiting the filter are, therefore, assumed to be agglomerates of inclusion species arising from the furnace metal interacting with particles from the grain refiner. 120
100
reduction rate %
80 Aver age efficiency 60
40
20
0 N15
1520
2025
2530
3035
3540
4045
4550
5055
5560
6070
7080
8090
90100
100120
120150
150300
P article siz e d istribu tio n [µm]
Figure 4: Inclusion removal efficiency for a cast with 3:1TiBAl rod shown in Figure 3 50ppi CFF, 3:1 T iBAl 0.6-1 kg/T , 60 minute settle
14
Impurity Level N15 [k/kg]
12
10
ROD IN @
0.6kg/T
8
after filter
6
before filter 46min, 1kg/T
4
2
0 0
10
20
30
40
50
60
70
80
Casting T ime (min.)
Figure 5: 3:1TiBAl rod – fed partway through cast – “low” inclusion load
Table 1 summarizes the effect of 3:1 TiBAl rod on the N15 inclusion levels for all the 50ppi CFF’s trialled.
30 Table 1: Effect of 3:1 TiBAl on N15 Inclusion Level Melt Condition 3:1 TiBAl Added Before Filter (1kg/tonne) Increase in Before Filter N15 Increase in After Filter N15 Inclusion Level (k/kg) Inclusion Level (k/kg) Settled (clean) 0.5 - 1.5 0.25 - 0.5 Stirred (dirty) 0.5 - 1.5 2.00 – 6.0 In order to better understand the observed effects of the 3:1 TiBAl rod addition on the before and after LiMCA curves the next phase of the work investigated whether these results could be attributed to some or all of the following : • The possible `mechanical´ effects of rod addition introducing inclusions into the melt by disturbing its surface. • Slower than expected aluminide dissolution. • CFF inclusion loading, as defined previously. • Boride agglomeration in the launder along with other inclusions present in the melt. 4.2
Impact of Mechanical Disturbance
The LiMCA results of trials when a rod of commercial purity aluminum was fed into the metal clearly showed that the impact of mechanical disturbance or oxide pull-in, due to the rod feeder system used in this study, was negligible and could thus be discounted. 4.3
Impact of Partially Dissolved TiAl3 Particles
This was investigated by conducting trials using a specially produced 0.7% TiAl rod (i.e. no TiB2 phase, similar free Ti (TiAl3) to the 3:1 TiBAl rod). Note, it was assumed that the absence of borides had no effect on the TiAl3 dissolution rate. 50 ppi CFF, 0.7%Ti rod, stirred before casting
14
Impurity Level N15 [k/kg ]
12
after filter
10
before filter 8
6
rod up to max. = 2.5kg/T
4
rod in 1kg/T 2
0 0
10
20
30
40
50
60
70
80
90
Casting T ime (min.)
Figure 6: 0.7%TiAl rod – fed partway through cast – “high” inclusion load
Figure 6 shows the LiMCA curves for a 50ppi CFF with dirty metal (stirred melt) where the 0.7% TiAl rod has been introduced after 30 minutes of casting. The same effect was observed for a settled melt. Significantly, no effect of the addition is evident before or after the filter.
31 This suggests that the presence of TiAl3 does not contribute to the before or after filter counts and that the effects being measured are not being partly caused by undissolved TiAl3. Further trials feeding 3:1 TiBAl into the system without a CFF showed that the dissolution of TiAl3 did not have a significant influence on the results. During these trials three LiMCA units were situated along the launder covering over 10m of its length. The N15 value could be seen to have lifted slightly after the rod was added (after 20 minutes) but a constant value was recorded for all three LiMCA units. 4.4
Impact of Inclusion Loading on the Efficiency of a CFF
Figure 7 shows the LiMCA traces for a trial where the loading on the CFF was kept at a high level throughout by stirring continuously during the cast. Grain refiner was omitted from this trial to check if loading alone resulted in release effects. It can be seen that despite a very high incoming inclusion loading (>20k/kg) no release effects at all are evident. In fact, despite there being severe metal level disturbances due to the vigor of the stirring, the 50ppi CFF displayed a very high efficiency and a stable and consistently low post filter LiMCA value (0.25k/kg). It could be concluded therefore, that inclusion loading alone at the levels investigated did not reduce the filter efficiency. This is accepting that at even higher loading levels filter saturation and diminished efficiency may occur. Figure 8 looks at the problem in another way, with the rod entered at the start, as per normal practice, but this time with stirring occurring later in the cast. Here, the same response as before can be noted. When a higher loading is introduced from the furnace in the presence of the grain refiner, a sharp decline in the filter efficiency occurs. 50 ppi CFF, NO rod, stirred during cast
18 16
Impurity Level N15 [k/kg]
14 12 10
after filter
8
STIR
6
before filter
STIR
4 2 0 0
10
20
30
40
50
60
70
80
90
Cast ing T ime (min.)
Figure 7: No grain refiner – stirred throughout cast – “high” inclusion load
In summary, it is postulated that the introduction of Ti & B containing grain refiner material alters the behavior of the ceramic foam filter in trapping and/or retaining particles thus causing them to have a diminished efficiency compared to those found in the absence of grain refiner. This was only found to be significant when the incoming inclusion loading is high. If good furnace practices are followed and the inclusion loading is low (settled melts) there appears to be only a minimal impact of the grain refiner on the filter’s performance. Metallographic assessment of spent filters suggested that the `bridges´ of inclusions across the filter cell junctions seen in the absence of grain refiner1-3, do not appear when grain refiner
32 is employed. It is believed that the interaction of the TiB2 particles and the inclusions in the metal or filter alter this mechanism of filtration and is responsible for the significant increase in inclusion counts at the filter outlet when the grain refiner was introduced later in the cast. `Bridges´ across the `window´ regions (at least for the particle types in the study) may be a form of cake filtration and appear to be associated with high filtration efficiencies. 50 ppi, LSM 3:1 TiBAl rod 1kg/T, ’pre-settled’ 90min.
14
Impurity Level N15 [k/kg]
12
after cff before cff
10 8
5min.air stir started
6 4
Rod in 1kg/T
2 0 0
10
20
30
40
50
60
70
80
90
Casting T ime (min.)
Figure 8: 3:1 TiBAl – fed from start of casting plus stirred partway through cast –“low” inclusion load
5
Conclusions
1. Ceramic foam filters have the capacity for high efficiencies in the absence of grain refiner, even under severe disturbance conditions and with sustained high loading throughout the cast. 2. At high inclusion loading the introduction of a 3:1 TiBAl grain refining rod leads to a reduced filtration efficiency. When the inclusion loading is low there appears to be a minimal impact of the grain refiner on the filter’s performance.
6
Acknowledgments
Sincere thanks are due to VAW Rheinwerk personnel and D.Gründler & N. Ozturk of VAW’s R&D division, without whose dedication the success of this program would not have been possible.
7
References
[1] N.J.Keegan, W.Schneider, H.P.Krug, Light Metals 1999, pp 1031 - 1041 [2] N.J.Keegan, W.Schneider, H.P.Krug, 6th Australasian Asian Pacific Course & Conference, Aluminium Cast House Technology : Theory & Practice (Ed.: M.Nilmani), TMS 1999, 159-174. [3] N.J.Keegan, W.Schneider, H.P.Krug, Light Metals 1997, 973-982. [4] C.Dupuis, G.Beland, J.P.Martin, Proceedings of the 32nd Annual Conference of Metallurgists, Quebec, Canada, CIM, 1993, 349-358.
Review of Dissolution Testing and Alloying Methods in the Casthouse Gregorio Borge1, Paul S. Cooper2 and Stuart R. Thistlethwaite2 1
Bostlan, S.A., Larragane, 1 E-48100 Mungia (Spain).
[email protected] LSM Co.Ltd., Fullerton Road, Rotherham, South Yorkshire S601DL (Great Britain),
[email protected];
[email protected]
2
1
Introduction
The properties of Al alloys are largely dependent on the correct addition of the alloying elements before the casting process. There are a number of ways of performing these alloying additions such as pure metal, master alloys, powder injection and a variety of compacted powders additives (tablets, mini tablets and briquettes). The choice of which alloying addition to use for each element is complex [Thistlethwaite 1992]. A number of competing factors have to be taken into account and their relative importance may vary from plant to plant and product to product. Some of the key criteria include metal temperature available, virgin:scrap ratio, furnace type and layout, addition and stirring practices, alloy change frequency and end product quality. The cost of alloying is not always easily defined. There are not only raw materials costs, but also processing, yield, quality and overhead considerations, which need to be taken into account when selecting the most appropriate alloying technique. LSM and Bostlan´s experience as worldwide suppliers of different products for the aluminium alloying industry is that in recent years consumption of compacted additives has noticeably increased. Some casthouses have stopped their injection production lines, and new facilities for this addition practice are rarely set up, mainly due to capital costs and further the strict quality control requirements of the raw materials because of safety risks when handling powders. Compared to master alloys additives, compacted powder additives are easy to handle; cold metal quantity to be added to the furnace is not very high (since the lowest concentration of the alloying metal in the compact is 75%); accurate additions for compositional adjustments can be performed if necessary; and stocking costs are reduced. In the mid 90s some studies on dissolution of compacted powders were published [Young 1993; Campbell 1994; Fisher, 1994; Perry 1994; Shafyei 1995]. Many of these works are focused on laboratory studies, so it can be said that the general mechanism and the behaviour of the compacted additives in small furnaces is known (exothermic heating of the compact, intermetallic compounds and swelling of the compact). During subsequent years, the literature concerning dissolution and recovery of alloying metals from compacted powders has significantly decreased. The most recent work covers deeper studies on the intermetallic compounds influence for the explanations of the dissolution mechanisms [Bristow 1999; Lee 2000]. All this background literature is useful for understanding the behaviour of the compacted additives, but it could be said that in general no recommendations for an industrial practice have been given. There are several issues the casthouse is interested in including dissolution Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
34 rates and final recoveries in industrial scale furnaces; production of skims/drosses and correct skimming practices for each additive; correct stirring. The challenge is to know how laboratory tests relate to industrial practice.
2
Products
Pressing mixtures of metallic powders (Mn, Fe, Ti, Cu, Cr, Ni, Pb, etc) with aluminium, a flux, or a mixture of both components produces compacted powder additives for the aluminium alloying industry. Alloying metal contents range from 75% to 85%. The most common compacted additives are tablets and mini tablets. Both are cylindrical. Tablets are nominally 90 mm diameter; height and weight depend on the alloying element, but usual figures range from 1250 to 1333 g and approximately 45 mm height. For mini tablets, 40 to 45 mm diameters are available, with 50 to 200 g weights and different heights as well. Tablets and mini tablets are produced in hydraulic presses, using special steel punches and dies. To maintain acceptable tool life requires the use of a lubricant in the formulation of the product. A comparable product is a briquette with the pressure produced between two compacting rolls with “pillow shaped” indentations in the rolls to form the tablet shape.
3
Techniques
The performance and behaviour of compacted additives in molten aluminium furnaces is usually studied by the TP-2 test as published by The Aluminum Association [Aluminum Association 1990], although many laboratories adapt the procedures or equipment to their own facilities. The TP-2 test includes the use of crucible furnaces, temperature ranges around 732±10ºC, sampling every minute during the test’s first ten minutes, and frequent stirring practices. Following the test, dissolution rates and final recoveries of the alloying metal can be studied, as well as skims produced and reactivity phenomena (bubbles, flames, fumes). Scientific literature [Perry 1994; Shafyei, 1995] mentions also microscopic techniques for the study of the intermetallic compounds, whose formation is the first step before the dissolution. The connection of these study techniques, and their corresponding application, to industrial practice can be sometimes confusing or not clearly seen, mainly due to problems arising from the different scale when working in the casthouse. New techniques or working methodologies applied to the study of the compacted additives can help to understand the dissolution process. 3.1
The Microscopic Behaviour (the Monitoring of what Happens in the Furnace)
3.1.1 Classical Monitoring Using a steel cone as in the Aluminum Association TP-1 grain refiner test 10 kg of Al is heated in a small resistance furnace. A whole or part tablet is placed on its edge in the bottom of the steel cone mould. The tablet and mould are preheated and then lowered into the bath and allowed to fill to the level of the notch. The mould is held in the bath for the time required before removing and quenching. The experiment is repeated with different hold times if necessary.
35 The cast cones are sectioned vertically, bisecting the circular face of the mini tablet. The cut faces of the samples are ground flat to reveal the structure and any undissolved remnants of the tablet. Figure 1 shows some scanned images of the cast samples for 85% Mncontaining 200 grams mini tablets.
Figure 1: Steel cone mould test for 200 g 85%Mn mini tablets. Mini tablets were extracted after 60 seconds, 120 seconds, and 240 seconds after addition
Visual information given by this technique is direct. The methodology has not been widely mentioned in the literature [Bristow 1999], but comparisons between different materials can be performed easily and at relatively low cost by this technique, especially if dissolution rates in the first few minutes are important. Thus, fast checking of the behaviour of the material is possible. As disadvantages, no micrographic studies are possible, and scaling problems could arise when applying results to an industrial furnace. As an example, there is no possibility of using complete standard tablets [only a portion]. 3.1.2 X-Ray Monitoring Recently, the Department of Materials of the University of Birmingham has developed a technique that allows a direct view through the dissolution process based on X-Ray radiations. Using this technique, a continuous monitoring of the compacted additive can be recorded on videotape. For the test a mini tablet can be added into a sand mould 125 mm wide and 250 mm deep containing 7 kg aluminium. This technique confirms directly the results obtained from others, especially those concerning the swelling and breaking down of the compacted additive. Flames in the first seconds are due to the presence of solid lubricants in the formulation of the additives for proper compaction. Further swelling phenomena are due to the formation of intermetallic compounds between the alloying element powder and the aluminium. The additive is finally broken due to the melting of the aluminium/flux within the mini tablet. This stage does not mean that the alloying metal (for example Mn) has been recovered at this time, but that the compacted structure has been broken down. The X-ray technique is costly, although direct and comparative results can be quickly obtained. However the dimensions of the sand mould are not adequate for standard tablets or for adding more than one mini tablet as the alloying level would be too high, and swelling phenomena could be uncontrolled. In addition temperature control is not possible during the experiments.
36 3.1.3 Swollen Compacted Additives The techniques described are only useful for direct monitoring of one mini tablet due to the size limitations. Real additions are never like this: many compacted additives are added together usually in the same part of the furnace, so liquid aluminium may not enter into the mini tablet so easily. Swelling phenomena and further metal recovery may thus be delayed. Swollen compacted additives (one mini tablet, one standard tablet, or some mini tablets) can be obtained from industrial scale test furnaces with a more realistic approach to the customer’s situation. As an example, Figure 2 shows some results for 75% Mn 100 grams mini tablets added at 730ºC and extracted from the furnace at the times shown. The furnace used was a rotary oxycombustion 400 kg facility and a 15 cm diameter by 8 cm height holed ladle was used for sinking and extracting the samples. Table 1 below summarizes data from the tablet samples. Table 1: Tablet samples Time (s) Temp (ºC) W (g) 0 N/A 100 30 728-725 103 45 730-728 137 60 731-729 59 70 732-730 23
h (mm) 17.55 24.95 25.00 N/A N/A
Ø (mm) 40.05 49.90 55.10 N/A N/A
ρ (gcm-3) 4.6 2.4 2.3 N/A N/A
The extracted and swollen samples can be cut and polished for examining on an optical microscope. Figure 3 shows an example for the piece extracted at 45 seconds. The intermetallic compounds can be seen. Advantages of this technique include a direct monitoring of the real process in a furnace. Samples extracted can be weighed, and directly analysed. For example, it can be seen that at 30 seconds the swelling phenomena has started but no aluminium has entered into the mini tablet yet. If related to previous techniques (X-Ray), it can be concluded that even if the mini tablet is destroyed at 70 seconds, dissolution is not complete. Conclusions from the micrographic analysis can also be obtained; in this case the samples are closer to what would be obtained from an industrial furnace. Finally, this 400 kg furnace allows also the addition and retention of swollen standard tablets or groups of mini tablets. The main disadvantage of this technique is the cost. 3.2
The Macroscopic Behaviour
In terms of the use of compacted additives, issues of importance in the casthouse include high recovery in a short time, the production of skims and/or dross due to the operation, and the reactivity (flames, fumes, bubbles) of the material added to the furnace. Knowledge of the dissolution mechanism given by the microscopic techniques helps the manufacturer to produce materials with different characteristics for achieving better results. However these results have to be accomplished in an aluminium furnace. The Aluminum Association’s TP-2 test does not specify any furnace size for the dissolution and recovery test. Crucible furnaces of any size could be used. Laboratory costs imply that small (10 to 40 kg) furnaces are generally used. The results of these tests are subject to doubt because they are on a much smaller scale:
37
Mn75% - 730ºC
0’’
30’’
45’’
60’’
70’’
Figure 2: Swollen samples extracted from the 400 kg furnace
Figure 3: Microscopy of the sample extracted at 45 seconds (x50)
•
Standard tablets cannot be directly studied, since they add too much material for the final alloying level. • Stirring practices are different to those used in the casthouse. The TP-2 test proposes a highly effective stirring system for every minute of the test for the first 10 minutes. • Dross production and reactivity have to be estimated in a very small surface. Skimming practices are thus also different from those used in the casthouse. Issues of relevance to practical dissolution of compacted additive powders cannot be explained just by experiments carried out in small crucible furnaces. With the size of the experiment being such an important factor, a methodology was developed for a 400 kg furnace in order to perform continuous and reproducible dissolution tests. The furnace is shown in Figure 4. The main aim of using an almost industrial furnace is to obtain results with no scaling problems. The content of this kind of furnace should be stirred for research work, since homogenisation of the melting bath is necessary for maintaining the temperature and for adequate and repetitive sampling. Stirring is performed before every sampling process, using a rake not hitting the tablets (or mini tablets) added. Samples are usually taken every five
38 minutes for 40 minutes (for Mn, Fe or Cr tests, for example) or even for 90 minutes (for Ti). For performing a proper comparison of the results, materials are added without any packaging, whereas industrial practice is usually to add plastic or foil wrapping, and/or cardboard boxes or paper sacks. Temperature control is performed with a thermocouple sunk in the bath; it is usually accepted to have a range of ±10ºC for each experiment, but usually a ±4ºC after material addition can be achieved.
Figure 4: A rotary 400 kg furnace for aluminium dissolution tests
Some advantages of this facility for reproducing the industrial practices are: • Realistic stirring practices. A ceramic rake is used in this case, which is longitudinally used for homogenisation. This can be taken as a very similar practice to that of many customers. • Distribution studies can be performed. Since differences arise from adding all the material in the same point or in evenly distributed points, a furnace with an adequate surface as this can be used for this kind of research. • Realistic skimming practices. Experiments can be also performed following the customer’s practices for skimming: either before or after addition, or with addition of drossing fluxes if required. • Furnace capacity ensures the researcher or the customer that any final alloying level of any aluminium series alloy required can be achieved. Addition of standard complete tablets is not a problem. The main issue with this kind of experiments is the cost, which is affected by the large aluminium quantity used, even if aluminium can be recovered after the experiment. This technique usually yields dissolution curves closer to real situations in the casthouse, but the control of the experimental factors is more difficult and costly than with a smaller (and more easily controlled) furnace.
4
Conclusions
The increase in consumption of compacted powders for alloying aluminium in the casthouse has focused recent research developments in this field. This work has presented different working methodologies and new-in-the-field applications for ascertaining and proving the behaviour of tablets and mini tablets in aluminium furnaces. It could be said that a supplier
39 controlling these diverse techniques can give a more complete answer to many problems of the customer/producer. In this sense, the most evident application in order to satisfy/answer a customer’s dissolution problem is the large capacity furnace. This furnace combines an almost industrial facility with the possibility of performing designed and highly controlled work: working conditions are closer to those of the casthouse, and macroscopic research work can be performed to obtain results concerning dissolution rates, final recoveries, dross production, and reactivity. Dissolution rate results given by this furnace are usually lower than those given by the typically used laboratory scale crucible furnaces. In most cases this should not be taken as an ineffective method or product, but as a different approach to the problem. On the other hand, the microscopic monitoring of the behaviour of the compacted additives has been improved with new technologies and different methodologies. Fast and low cost comparative analysis of the dissolution rate can be performed sinking a steel cone mould containing a mini tablet or a tablet portion into a prepared crucible furnace. A direct insight of the swelling and breaking down of a mini tablet can be performed using a sand mould and an X-Ray technique. Fast and comparative (but costly) direct results can be obtained. Finally, different types of real swollen samples obtained from a furnace without scaling problems can be classically studied with the microscope. All of these research results are useful in order to propose new developments/products by the supplier, and to understand specific customer related problems under many different situations.
5
References
[1] Aluminum Association, The, Standard Test Procedure for Measuring the Dissolution of Aluminum Hardeners, The Aluminum Association, 1990 [2] Bristow, D.J., Lockwood, S., Woodcock, T.G., and Cook, R., 128th TMS Annual Meeting and Exhibition, 1999. [3] Campbell, G.T., Bridges, R.E. and Niedzinski, M., Light Metals, 1093-1097, 1994. [4] Fisher, P., Cooper, P.S., and Thistelthwaite, S.R., Dissolution Mechanisms in Aluminium alloy additives. 123rd TMS Annual Meeting and Exhibition, 1994. [5] Lee, Y.E., and Houser, S.L., Dissolution mechanism for high melting point transition elements in aluminium melt, 129th TMS Annual Meeting and Exhibition, 2000 [6] Perry, W.H., Aluminium recovery from ‘all metallic’ hardener briquettes, Light Metals, 841-848, 1994. [7] Shafyei, A., and Guthrie, R.I.L., Dissolution mechanism of compact briquettes of high melting point additives stirred in liquid aluminium, Light Metals, 831-839, 1995. [8] Thistlethwaite, S.R., Review of alternative methods for alloying aluminium, Light Metals, 1005-1011, 1992. [9] Young, D.K., Setzer, W.C. and Boone, G.W., New concept in alloying aluminum, Light Metals, 745-751, 1993.
The Effect of Casting Parameters on the Metallurgical Quality of Twin Roll Cast Strip Yucel Birol1, Gökhan Kara2, A. Soner Akkurt2, Chris Romanowski3 1
Marmara Research Center, P.O. Box 21, 41470 Gebze-Kocaeli, Turkey ASSAN Aluminum Works, Tuzla, Istanbul 81700, Turkey 3 FATA Hunter Inc., Riverside, California 92507, USA 2
1
Abstract
31 different samples covering a range of casting parameters for the AA8006 alloy, were cast on industrial scale with 1725mm and 2184 mm wide Speed Casters. This paper describes the general trends correlating casting parameters to the metallurgical quality of the cast strip.
2
Introduction
When compared to the traditional hot mill process, the relatively low capital cost of twin roll casters, in combination with their lower energy and manning costs, have made twin roll casting an increasingly popular method of producing a wide range of aluminum flat rolled products [1]. The recent trend has been to reduce the gauge at which these casters operate to <3mm [2,3]. To fully utilize the potential of their thin gauge and wide strip caster investment, Assan Aluminum, in cooperation with FATA Hunter and Marmara Research Center undertook a long-term, industrial-scale, collaborative development program to characterize the twin roll casting process. Some initial results describing the effect of casting parameters on the twin roll cast strip microstructure were reported recently [4]. Additional casting trials were carried out in this study with the AA 8006 alloy to investigate further the effect of casting parameters on the metallurgical quality of twin roll cast strip. This paper presents the initial macro characterization of selected samples.
3
Experimental Procedures
A series of casting trials were conducted at ASSAN on 1725 mm and 2184 mm wide Speed Casters. The trials were carried out with AA 8006 and in the gauge range from 6.0 mm to 3.0 mm. The cast widths used in the trials varied between 1074 and 2000 mm. Samples were taken, in each trial, after running the caster under a particular set of casting parameters until steady state conditions were established. Specimens were sectioned for metallurgical analysis from the center of as-cast sheets. Each sample was finished with collodial silica and etched with 0.5%HF solution for microstructural investigations. Optical and stereo microscopy techniques were employed to examine the micro- and the macrostructure of the as-cast samples. Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
41
4
Results and Discussion
Several types of segregation patterns were identified under different casting conditions. Solute-rich channels running along the casting direction at or near the center plane were most frequent. The channel segregation was not always restricted to the center plane and was occasionally aligned at some angle to it. Another type of segregation has not produced channels at all. A dispersion of relatively small, equiaxed regions of solute-rich material around the center plane have formed in this type of segregation.
Figure 1: The longitudinal cross sections of two strips cast at approximately 3mm at low (a) and high casting speeds (b), and that of another strip cast at 6mm and nearly the same speed with that in Figure 1a (c)
Among the several casting parameters investigated in the present work, the casting speed and the casting gauge were most influential on the segregation intensity (Fig. 1). The intensity of segregation increased with increasing casting speed and/or with increasing casting gauge. There was clearly a limit for the casting speed at each casting gauge upto which no evidence of segregation could be found. Segregation has always become more prominent with increasing casting speed above this limit. It is also worth noting that the thickness of the strip which could be cast without segregation decreased with increasing casting speed (Fig. 2). 1200 with CLS
6
8011 with CLS 8011 without CLS 8006 with CLS
THICKNESS (mm)
8006 without CLS
5
4
3
2 100
120
140
160
180
200
220
LINE SPEED (cm/min)
240
260
280
Figure 2: Segregation limit diagram on a binary coordinate system with casting speed versus casting gauge
Each point in Fig. 2 represents a different set of casting conditions and an arbitrary boundary can be drawn to identify those conditions required to cast a segregation-free strip. The seperating force, which was not a controlled variable in the casting trials which were performed under the gap-control mode, appeared to be an overall parameter describing the segregation intensity rather some segregation, as one would expect. accurately. Lower
42 seperating forces, which represent casting conditions leading to less fraction solid entering the roll gap, were always found to be associated with at least
a)
b)
Figure 3: Effect of casting width on segregation behavior: a) 1080mm and b) 2000mm
The casting width was also found to have a big impact on the segregation behavior. Segregation became increasingly more prominent with increasing strip width, provided that the other casting parameters were held constant (Fig. 3). This can be attributed to the increase in the amount of heat to be removed with a constant heat sink (i.e. same size rolls) which in turn, implies an increase in the solidification time and possibly to the difficulty in uniformly distributing the metal flow in the tip when casting wider strips. The amount of segregation was found to be affected also by the tip set-back values. Larger set-back values are more likely to allow solidification to be complete before the rolling action starts thus limiting the channel formation. This is evidenced by an increase in seperating force with increasing set-back. Larger tip setback values apparently reduce the intensity of segregation and thus help to improve the quality of the strip. The concentration of the graphite sprayed onto the casting rolls as a parting agent seems to affect the segregation intensity to a lesser extent. The amount of segregation increased with increasing graphite concentration possibly owing to unfavorable heat transfer conditions. Grain refining practice, on the other hand, was found to have a big impact on the segregation behavior. Strips which were not grain refined seemed to experience substantial segregation which was spread wide on both sides of the center plane with channels arranged in a “V” pattern, extending from the centerline towards the surface at some angle to the casting direction (Fig. 4). On the other hand, the segregation was confined to the center plane in grain refined strips and its imntensity was reduced. Imrovement in chemical homogeneity with grain refinement was reported by other investigators as well [5]. It is fair to conclude then that the grain refinement is critical not only to control the grain size but also to control the intensity of segregation.
a)
b)
Figure 4: Effect of grain refiners on two samples; with a) and without b) grain rafinement
43 All strips, regardless of the casting gauge, revealed a very fine structure at the surface and a relatively coarser one in the core of the strip. While the gradual coarsening of the structure from the surface to the center results from the cooling rate gradient encountered in the strip casting process and is thus typical of all strip-cast alloys, it is regarded as a metallugical defect in the case of thin strips which reveal a very characteristic surface zone owing to the very high solidification rates encountered during strip casting. This has become increasingly more prominent with decreasing casting gauge and the surface of 3mm strips were featureless even after heavy macroetching. The matrix in this surface layer was found to be heavily supersaturated with the alloying elements, particularly with those which have very low diffusivities, thus making a subsequent thermal exposure a very critical step. This zone was found to recrystallize to yield a rather coarse structure when annealed at sufficiently high temperatures.
a)
b)
Figure 5: Effect of graphite concentration on depth of surface chill zone; a) 0,7% and b) 1,75% graphite concentrations
The graphite concentration in the parting agent, which had only a minor role on the segregation behavior, was found to have a big impact on the depth of the surface chill zone, and thus seems to be critical regarding the quality of the cast strip. The depth of the chill zone which undergoes substantial grain coarsening upon thermal exposure is reduced substantially when the graphite concentration in the parting agent is doubled (Fig. 5).
5
Conclusions
Among the several casting parameters investigated in the present work, the casting speed and the casting gauge appeared to be most influential on the segregation behavior of the strips. The intensity of segregation has always increased with increasing casting speeds for a given casting gauge. The thickness of the strip which could be cast without segregation decreased with increasing casting speeds. Likewise, the casting speed which could be tolerated to produce a strip free of any segregation increased with decreasing casting gauge. Segregation became increasingly more prominent with increasing strip width, provided that the other casting parameters were held constant. Graphite concentration of the parting agent seems to affect the segregation intensity to a lesser extent. the grain refinement is critical not only to control the grain size but also to control the intensity of segregation.
44
6
Acknowledgements
It is a great pleasure to thank Osman Cakir of Marmara Research Center and Dr. Murat Dündar and Seda Ertan of ASSAN Aluminium and Mr. Saun Hamer for their contributions in the experimental part of this work.
7
References
[1] Technischer und wirtschaftlicher Vergleich von kontinuierlichen Gless-und Walzverfahren zur Herstellungvon kaltgewalten Aluminiumbandern Informationsblatter, Deutsche Gesellschaft für Materialkunde e.v., Adenaueralle 21, D-6370 Oberursel, Germany. [2] R.Beals, B.Taraglio, B.Carey and C.Romanowski, Light Metals 1997, R.Huglen, ed.; The Minerals, Metals & Materials Society, 1997, p.757-763. [3] A.B. Espedal and R. Roder, “Prospects of Thin Gauge High-Speed Strip Casting Technology”, Light metals 94, U. Mannweiler, TMS, 1994, p. 1197-1203. [4] 6(UWDQ0'QGDU<%LURO.6DUÕR÷OX(2]GHQ$6$NNXUW*<ÕOGÕ]ED\UDN6+DPHU and C.Romanowski, Light Metals 2000, R.Huglen, ed.; The Minerals, Metals & Materials Society, 2000, p.757-763. [5] P.S. Cooper and P. Fisher, “Grain Refining Strip cast Aluminum”, Light Metals 1982, J.E. Andersen, ed.; AIME, New York, N.Y., 1981, p.
Casting Technology and Processes – Aluminium
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
Influence of Different Lubricants on the Friction between the Solidifying Shell and the Mould during the DC Casting of AlMgSi0.5 Frank Dörnenburg 1, Siegfried Engler2 1 2
VAW aluminium AG, Research and Development, P.O. Box 2468, 53014 Bonn, Germany Foundry Institute, Aachen University of Technology, Intzestr. 5, 52072 Aachen, Germany
1
Abstract
Good, uniform billet quality is the most important requirement in the production of billets and ingots. The lubricant used can be an influential factor in meeting this requirement. One role performed by lubricant is to reduce friction between the solidifying shell and the mould wall. In this paper, an experimental set-up which was used is presented to measure this friction. Different lubricants and different amounts of lubricant were applied during the DC casting of AA 6063 (AlMgSi0.5). The measurements correlated with the roughness of the cast billets. It could be observed that with a rougher surface, higher friction values were measured. Using a new lubricant [1], a very smooth surface could be achieved although the friction was higher than with a conventional lubricant. This was the result of a large contact area between the surface of the billet and the mould wall. Finally, a new model which describes the processes in the gap between the solidifying shell and mould wall is presented.
2
Introduction
The production of semi-continuously cast billets and ingots without surface defects is becoming more and more important. One quality criteria is the roughness of the surface. The surface should be as smooth as possible. The type of mould used has a great influence on the quality of the billets and ingots. Using electromagnetic moulds, a very smooth surface can be obtained because the solidifying shell is not in contact with the mould wall [2]. Due to costs, however, the electromagnetic process (EMC) is used very seldom. For all other types of moulds, one key to getting a smooth surface is to reduce the friction between the ingot surface and the mould wall. When casting slabs, it is usual to apply a grease on the mould wall before each casting. Billets are mostly cast using moulds which have a continuous lubrication system. Some advantages are mentioned in [3]. This paper presents the results of friction measurements taken during the DC casting of AlMgSi0.5. After presenting the testing equipment, the friction curves are presented and described in detail. Finally, a modified model for the processes occurring in the gap and in the contact zone between the solidifying shell and the mould wall is shown.
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
48
3
Experimental Set-up
The experiments were carried out at the Foundry Institute, Aachen University of Technology. It was possible to cast DC billets with a maximum length of 2m. The hot-top mould at the Foundry Institute has a diameter of 156 mm (equivalent to 6"). With the type of casting equipment used, different kinds and amounts of lubricants can be applied continuously. Three different force sensors with strain gauges were used for the friction measurements. Figure 1 shows of the position of the force sensors.
Mould 1
2 120°
Launder
120° 120°
3 Force Sensor Figure 1: Schematic diagram of the position of the three force sensors relative to the mould
The following figure (Figure 2) illustrates the position of one force sensor. The forces were measured separately at each force sensor [4]. A
D
A:
Screw
B
B:
Mould
C
C:
Disc
D:
Force sensor
E:
Casting table
E Figure 2: Position of one force sensor on the casting table
All experiments were carried out with the alloy AA 6063 (AlMgSi0.5). The starting casting speed was 100 mm/ min and the final casting speed was 150 mm/min. The temperature of the melt was 740°C.
49 To be able to compare the different units employed by researchers and authors to measure the amount of lubricant used for different casting speeds and mould sizes [5], the following procedure was undertaken: the amount of lubricant [ml/min] was divided by the circumference of the mould and the casting speed so that the final unit is [ml/m²]. This unit is called specific lubricant volume.
4
Results and Discussion
All the following diagrams show the friction versus time. In all experiments, the friction increased during the experiment. The reason for this is that the axis of the mould (and of the casting table) is not perpendicular to the axis of the hydraulic descending device. The billet did not descend parallel to the axis of symmetry of the mould. During the casting process, the following forces were measured by the force sensors: • part of the weight of the launder. One side of the launder was connected to the mould, the other part was positioned on a pillar. • the weight of the mould • the weight of the melt inside the mould • the weight of the water inside the mould • the weight of the lubricant inside the mould • the buoyancy of the melt in the hot-top • the friction forces between the solidifying shell and the mould wall 80 70 Bleed out
60
Force in N
50
67 ml/m2 165 ml/m2 217 ml/m2 Rape oil
40 30 20 10 0 -10
0
50
100
150
200 250 Time in s
300
350
400
450
Figure 3: Friction measurements for different amounts of rape oil
The weight of the melt inside the mould can be neglected because only the friction forces between the melt and the mould wall are measured by the sensors. Furthermore, the weight of the launder, the mould, the water and the lubricant and the buoyancy of the melt in the hot-top are constant during the experiment. These forces were subtracted. Due to the time taken between starting casting and reaching the final casting speed, a definition had to be made to be able to compare the different experiments. It was therefore defined that t=0s when the final speed starts. At this moment the forces are defined as 0 N.
50 Figure 3 illustrates the friction forces for different amounts of specific lubricant volumes. The increase in the forces at 150s was caused by a bleed-out. By stopping the process and then re-starting it, the initial forces can be observed again. The bleed-out was obtained with a very low specific lubricant volume. By increasing the amount from 67 ml/m² to 165 ml/m² then to 217 ml/m², the friction decreases. 80 17 ml/m 2 24 ml/m 2 28 ml/m 2 36 ml/m 2 48 ml/m 2 Castor oil
70 60
Force in N
50 40 30 20 10 0 -10
0
50
100
150
200
250
300
350
400
450
Time in s
Figure 4: Friction measurements for different amounts of castor oil
The following figure (Figure 4) shows the same curves for castor oil. As mentioned in [5], when using castor oil, the minimum amount of lubricant is lower than with rape oil. The reason for this is the higher viscosity of castor oil. Two bleed-outs, with a specific lubricant volume of 17mm/m², can be observed. The high values of the experiment using 28ml/m² were caused by a small particle getting between the hot-top and the mould wall which resulted in a surface defect shown in Figure 5. The other curves are more or less identical and reaches values of 10N. Obviously castor oil is more able to reduce the friction. This lower friction forces can be correlated with Figure 6 and Figure 7. The surface of the billet using castor oil is smoother.
Figure 5: Surface defect caused by a small particle between hot-top and mould wall
51
Figure 6: Surface finish using 165 ml/m² of rape oil
Figure 7: Surface finish using 24ml/m² of castor oil. The smoother surface can also be seen with the lower friction forces
Force in N
Additional experiments were carried out using new lubricants: glycerine, a mixture of rape oil with fatty alcohol and a rape oil/ water mixture. Figure 8 illustrates the forces for these three new lubricants. The highest measured values of all experiments were found with glycerine. This correlates with the very rough surface shown in Figure 9. In contrast to this, the surface of the rare oil/ water billet is very smooth, comparable to surfaces obtained with AIRSOL VEIL or Airslip moulds [3], [6] and [7], Figure 10.
150 140 130 120 110 100 90 80 70 60 50 40 30 20 10 0 -10
696 ml/m2 glycerin with 12% Water 418 ml/m2 Rape oil / fatty alcohol 209 ml/m2 Rape oil/water
0
50
100
150
200
250
300
350
Time in s
Figure 8: Friction forces for three different "new" lubricants
400
450
52
Figure 9: Worst surface finish of all the experiments using glycerine.
5
Figure 10: Best surface finish of all the experiments using a rape oil/ water mixture
Conclusions
Friction forces between solidifying shell and mould wall
The friction forces between the solidifying shell of the billet and the mould wall can be correlated with the roughness of the billet surface. A very rough surface results in high friction forces, e.g. with glycerine. If the surface is very smooth – with a rape oil/ water mixture – the forces are marginally higher than those for pure rape oil or castor oil which give a rougher surface, Figure 11. This is caused by a large contact area between the surface of the billet and the mould wall. In case of the rape or caster oil surface, only the peaks of the surface are in contact with the mould wall and therefore the friction is less. If the surface becomes even rougher, due to a poorer separation or lower efficiency of the lubricant, more peaks are in contact with the mould wall and the forces increase more and more.
Lubricant does not work properly => very rough surface (e.g. glycerine)
Very smooth surface => Large contact area between shell and mould wall (e.g. when using a mixture of rape oil and water Roughness of the surface
Figure 11: Schematic depiction of the relationship between friction and roughness of the surface
A model of the processes between the mould wall and the solidifying shell was presented in [8] for the first time According to this work, some more features can be added to this model as seen in Figure 12. A layer of vaporized and cracked lubricant is to be found between the new lubricant and the solidifying shell. Naturally, the position and thickness of the layer is not
53 constant. There is a mixture of new, vaporized, cracked lubricant and vaporized and condensed water on the surface of the billet and the mould. All these elements have an influence on the heat transfer coefficient and consequently on the billet surface.
Figure 12: Schematic diagram of the processes in the gap showing of the relationship between friction forces and roughness of the surface
6 [1] [2] [3] [4] [5] [6] [7] [8]
References F. Dörnenburg, German Patent DE 197 43 689 C2 1999 D.G. Goodrich, J.L. Dassel, R.M. Shogren, Journal of Metals 34 1982, 45-49 F. Dörnenburg, S. Engler, Light Metal 1998, 1169-1173 Weiss, Dissertation, RWTH Aachen 1994 F. Dörnenburg, Dissertation, RWTH Aachen 1998 W. Schneider, M. Langen, German Patent DE 42 12 531 C1 1992 F.E. Wagstaff, W.G. Wagstaff, R.J. Collins, United States Patent 4 598 763, 1985 E. Lossack (Editor) Symposium Stranggießen, Deutsche Gesellschaft für Materialkunde e.V. (DGM) 1978, 104-109
Prediction of Boundary Conditions and Hot Spots during the Start-up Phase of an Extrusion Ingot Casting Steinar Benum1, Dag Mortensen2 and Hallvard Fjær2 1 2
Hydro Aluminium R&D Materials Technology, N-6600 Sunndalsøra, Norway Institute for Energy Technology, N-2027 Kjeller, Norway
1
Abstract
During the start-up phase of an extrusion ingot casting hot tears or shrinkage porosity may form from the centre cone of the starting block. If this occurs for alloys susceptible to hot tearing, the hot tear may extend through the length of the casting. It was suspected that a starting block with a relatively large cone in the centre led to a hot spot at the top of the cone. Temperature measurements in the ingot and the starting block during the start up phase and transient calculations of the start-up phase of the process indicated that the temperature in the upper part of the cone was above the solidus temperature for the investigated alloy. Calculations involving the coupled evolution of the temperature-, fluid flow- and stress-field during casting (ALSIM/ALSPEN) were also performed. Here, the influence of the thermally induced deformations on the heat transfer at the ingot surfaces were included in the boundary conditions of the heat and fluid flow model. The results from modelling and the experimental measurements were compared. A clear correlation is found between the formation of hot tearing and the calculated stresses and temperature gradients. Furthermore, the simulations indicate that reducing the cone height will reduce the hot tearing probability.
2
Introduction
A number of publications have dealt with the problem of cracks occurring due to the start up. One of the parameters that affects formation of cracks is insufficient and varying heat transfer between the starting block and the ingot. This lead to an increased sump depth during the start up phase. Jensen and Schneider [1,2] recommended to use a cone in the centre of the starting block to avoid this. Today, most starting blocks are designed with such a cone. During the start-up phase of an extrusion ingot casting hot tears or shrinkage porosity may form above the centre cone of the starting block. This hot tear may extend through the length of the casting. It was suspected that a hot spot at the top of the cone or a reduced heat transfer in this region could support the formation of a starting crack. A series of experiments with varying starting conditions and measurements of temperature in the starting block and the ingot were launched to identify the problem. These experiments were accompanied with numerical simulations.
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
55
3
Experimental
Five casting experiments with 228 mm diameter extrusion ingots were performed. The casting material was an AA6060 alloy with a liquidus temperature of 655°C and a solidus temperature of 555°C. The casting equipment used was a standard Hydro gas cushion system. During the start-up phase the casting speed was varied as following. Trial no. 1: constant speed of 130 mm/min. Trial no. 2:started with 130 mm/min and then lowered to 80 mm/min after 50 mm. The speed was increased to 110 mm/min after 150 mm. Trial no. 3: constant speed of 110 mm/min. Trial no. 4: started at 80 mm/min and then increased to 110 mm/min from 0 to 150 mm casting length. Trial no. 5: started with 80 mm/min and increased to 110 mm/min from 50 to 150 mm. The casting speed ramping is illustrated in Figure 1. 140
Casting speed [mm/min]
Metal
Hot top
130 120
Trial no. 1 Trial no. 2 Trial no. 3 Trial no. 4 Trial no. 5
110 100 90 80
T6
T5
T4 T1
T2
T3
Starting block
70 0
0.05
0.1
0.15
0.2
0.25
0.3
Casting length [m]
Figure 1: Casting speed ramping for the different starting trials
Figure 2: Placement of thermocouples in steel starting block and aluminium ingot. Thermocouples T1-5 is placed in the starting block 1 mm from the surface and T6 is placed in the extrusion ingot 2 mm above the surface.
During the trials the temperature variations were measured by thermocouples placed in the positions indicated in Figure 2. The metal temperature going into the mould were measured by an additional thermocouple placed just above the inlet to the mould. The metal was grain refined with AlTi5B1 rod. 2 kg/ton in the start-up phase and less than 1 kg/ton in the stationary period. The metal temperature at the inlet of the mould was aimed at being from 700 to 720°C.
4
Model Description
ALSIM [3] is a finite element method model for the development of the time dependent heat and fluid flows during DC casting. ALSPEN [4] is the stress model, where the compatibility equations, the momentum equations and the constitutive equations are solved by a finite element technique. The material is described as an elastic-viscoplastic material. The solution domain is the part of the ingot that is considered to be solid, i.e. where the temperature is predicted to be below a given coherency temperature Tc. These models have been coupled, as explained in [5], and the influence of calculated displacements and pressure on the thermal boundary conditions between the ingot and the bottom block are included in the
56 ALSIM/ALSPEN model. Where the local temperature of the ingot is higher than Tc, a heat transfer coefficient depending on the ingot surface temperature is applied for the bottom of the ingot. Where the temperature has become lower than Tc, and ALSPEN has computed the displacements, the heat transfer coefficient becomes dependent on both the local gap distance and the estimated contact pressure. Where the calculated gap has a magnitude of more than 0.2 mm, an air gap thermal boundary condition is applied. Where the gap distance is calculated to be less than 0.2mm, the heat transfer coefficient is assumed to depend on the local surface temperature of the ingot and the normal pressure pn. With ALSPEN, vertical forces, counterbalancing the total weight of the ingot, are distributed underneath the ingot. Emulating some elastic response from the starting block, the local contact pressure pn is estimated from the calculated gap distance. Although not believed to be accurate, this formulation enables us to incorporate some effect of temperature and pressure dependency in the contact heat transfer. Although there are no general accurate criteria for hot-tearing available, we assume that the formation of hot tears is correlated to the stress state and the viscoplastic straining during the final stage of solidification. One proposed hot tearing parameter (HTP) from [6] has been based on an assumption that hot tears will form if the viscoplastic strains obtained in a critical temperature interval become too large. The HTP is defined as a scaled ratio of the viscoplastic strain rate to the cooling rate, HTP = 10 4 × ε& vp T& , and is a differential formulation of this assumption. However, investigations of semi-solid materials [7] have revealed that the mechanical properties and the magnitude of the ultimate strain are quite different in compression compared to dilatation. Therefore, we here also put weight on the mean stress values from our simulations when considering the risk of hot tears.
5
Observations and Measurements
5.1
Hot Cracking Tendency
The length of starting cracks were measured by an ultrasonic device and the resulting average lengths are given in Table 1. Two trials resulted in starting cracks that closed when entering the steady state conditions. The first trial resulted in a crack going through the whole ingot length, which indicates that the steady state casting speed was too high to achieve a closure of the starting crack. The last trials were performed without cracking. Table 1: Average measured starting crack lengths Trial no. 1 2 3 4 5 Crack length [mm] Entire ingot 58 214 0 0 5.2
Temperature Measurements
The inlet melt temperature was measured and the average values were 710°C for trials no. 13, 715°C for trial no. 4 and 705°C for trial no. 5. The measured melt temperature as a function of time was used as input to the ALSIM calculations. The measured temperatures in the starting block (T1-5) and in the centre of the ingot (T6) are given for trial no.1-4 in Figure 3.
57 Trial no. 1
Trial no. 2
700
700 600
500
T1 T2 T3 T4 T5
400 300 200
Temperature [°C]
Temperature [°C]
600
100
T1 T2 T3 T4 T5 T6
500 400 300 200 100
0
0 0
50
100
150
200
0
50
Time [sec]
Trial no. 3
150
200
Trial no. 4
700
700 600 T1 T2 T3 T4 T5 T6
500 400 300 200 100
Temperature [°C]
600 Temperature [°C]
100 Time [sec]
T1 T2 T3 T4 T5 T6
500 400 300 200 100
0
0 0
50
100 Time [sec]
150
200
0
50
100
150
200
Time [sec]
Figure 3: Temperatures measured in the starting block (T1-5) and in the centre of the ingot (T6) for trial no.1-4
The temperature measurements for trial no. 5 do only slightly deviate from the measured temperatures in trial no. 4 and are therefore not shown here. The measurements were started when the filling of the moulds started. During the filling phase all temperatures in the starting block increases rapidly with a similar rate. As the casting starts and the relative position of the water impingement area moves, the rate starts to decrease as a function of distance from the centre of the cone. In T1 the temperature reaches a maximum after around 25 seconds (190250°C). Note that the temperature exceeds the solidus temperature in position T5 and for trial no. 1 and 3 also in position T4 (peak temperatures of 558 and 567, respectively). The peak temperatures in position T5 are 652, 646, 639, 635 and 636°C for trial no.1-5, respectively. After about 70 seconds the temperature exceeds the solidus temperature for the alloy and stays higher until about 100 seconds. In trial no. 1 the peak temperature even reaches the liquidus temperature of the alloy. For position T6 the temperature increases rapidly up to around 660 as the liquid metal heats the thermocouple and then decreases a 10-20°C before it decreases rapidly due to the effect of the water cooling after approximately 80 seconds. At this stage the temperature in the ingot becomes lower than the temperature which is measured in position T5 in the starting block.
6
Simulations
6.1
Temperature
Due to geometry limitations it has been necessary to use a modified geometry for the ALSIM simulations. The end of the cone on the starting block is normally placed above the bottom of the hot top, see Figure 1.
58 700
Temperature [°C]
600 500
T1 T2 T3 T4 T5 T6
400 300 200 100 0 0
50
100
150
200
Time [sec]
Figure 4: Temperatures from simulation with ALSIM (trial no. 4) 2.5
no.1
2
Mean stress [MPa]
Mean stress [MPa]
2.5
no.3
1.5
no.4 1 80
no.2 100
2
1.5
no.1-10 no.1-21
no.5 120
140
160
180
1 80
200
Time [sec]
100 120 140 Time [sec]
3 Trial no. 1 Trial no. 2
Trial no.3 Trial no.4
Trial no.5
3
no.3 no.1
1.5
0.75
HTP [1/110000K]
HTP [1/110000K]
2.25
no.4 no.2
100
1.5 0.75
no.5
0 80
2.25
120
140 Time [sec]
160
180
200
0 80
no.1-10 no.1-21 100 120 140 Time [sec]
Figure 5: Mean stress and HTP in the centre of the ingot at 600°C. To the right results from variations on trial 1 with lower cone heights are shown
In the model a horizontal expanding row of elements are located below the hot-top, which distinguishes between an Euler domain on the top and a mixed Euler-Lagrange domain below, as explained in [3]. The uppermost part of the solid cone travelling downwards through this area would require a dynamic deformable mesh (including re-meshing) between the fixed hottop and the top of the moving bottom block, which is not a part of the model. In the calculations, the cone has therefore been ended just below the hot top.
59 The resulting temperatures from trial no. 4 are shown in Figure 4. The differences between the measurements and the calculations are due to, among other things, the lack of the hot stream impingement on the top of the cone during the initial filling ("mould filling" are not included in the calculation). The reduced cone size in the calculation decreases the distance to the colder parts of the bottom block, and hence results in a lower peak temperature. Plots of the liquidus and solidus isotherms before they leave the cone shows that the sump depth reaches a maximum and that the isotherms nearly becomes vertical to the cone surface. For calculations with a reduced cone size the temperatures in position T5 and T4 becomes substantially lower. 6.2
Stress and HTP
Figure 5 gives the variation in maximum mean stress in the centre of the ingot as a function of time for all cases. If the peak mean stress at the temperature 600°C (corresponding to a solid fraction equal to 0.95) is used as an indicator, the trials may be ranked in the sequence trial no. 1,3, 2, 4 and 5 with regard to probability for cracking. Remark that this ordering of the trials is in agreement with the length of the cracks for trials 1, 2 and 3. Trials 4 and 5 did not crack, but 5 should have the lowest cracking tendency since a lower (or equal) casting speed where applied for this trial compared to trial 4, see Figure 1. If the HTP parameter at 600°C is used, trial no. 2 is ranked after 4. The overall agreement on cracking tendency with the HTP parameter is not as good as with the mean stress values. Results from two calculations with a cone height reduced with 10 and 21 mm, respectively, and with the trial no. 1 parameters, are also shown in Figure 5. The peak values for those cases are substantially lower than in the cases with the standard cone. As stationary conditions is approached trials no. 2-5 converge. Trial no. 1, with a higher casting speed, remains on a higher level.
7
Discussion
The hypothesis of this work was that insufficient heat transfer between the ingot to the cone could yield a hot spot just above the cone. In fact, the temperature in the cone becomes too high for the cone to act as a cooling medium during the critical phase of the solidification. That is, when the sump leaves the cone. From simulations it is found that the isotherms nearly becomes vertical on top of the cone. The cone is not found to affect the temperature in the liquid metal above the cone significantly. Here, the temperature stays at a stable value until the liquidus and soldius isotherms start to approach the top of the cone. This means that most of the heat transport is occurring in the ingot itself and that at a critical stage of the start up phase the cone is actually heating the ingot and liquid metal. Conclusively, the starting block should be re-designed or the steel cone should be replaced by a copper or aluminium cone with a higher heat conductivity. This is earlier found to give better results [1,2]. There is a good correlation of the simulations on stress and HTP compared with the observed temperatures and cracking behaviour. It is therefore believed that the HTP and especially the mean stress parameters may be used for optimising the starting block both with regard to design, material and starting conditions. The simulations performed with reduced cones yielded lower peak HTP and mean stress values than the cases with a full cone. Actually, the mean stress value for trial no. 1 with a reduced cone reaches a level between 2
60 and 4 with a full cone. Hence, in this case a reduction of the cone may reduce the starting crack problems.
8
Conclusions
From experimental trials and simulations the following conclusions are made: 1. The coupled ALSIM/ALSPEN model has been applied on 5 different casting trials. The calculated mean stress values at the centre of the ingot (at a fixed temperature in the mushy zone) shows that the stress levels for each trial are ranked in the same order as the observed crack lengths. 2. Due to a heating of the cone during the filling and holding period of the casting, the upper part of the cone acts as a heating source and yield a hot spot in the ingot above the cone. A reduction in the cone height or change of material may counteract this effect.
9
References
[1] E.K. Jensen and W. Schneider, Ligth Metals 1990, (Ed. C. M. Bickert), TMS, 1990, p.931 [2] E.K. Jensen and W. Schneider, Ligth Metals 1990, (Ed. C. M. Bickert), TMS, 1990, p.937 [3] D. Mortensen, “A mathematical model of the heat and fluid flows in direct-chill casting of aluminium sheet ingots and billets”, Metall. and Materials Trans., 30B:119-133, 1999. [4] H. G. Fjær and A. Mo, “ALSPEN - A Mathematical Model for Thermal Stresses in Direct Chill Casting of Aluminium Billets”, Metall. Trans., 21B (1990), 1049-1061. [5] H. G. Fjær, D. Mortensen, A. Håkonsen, E. A. Sørheim. Coupled Stress, Thermal and Fluid Flow Modelling of the Start-up Phase of Aluminium Sheet Ingot Casting. Light Metals, TMS, Warrendale, PA, 1999, 743-748. [6] H. G. Fjær, E. K. Jensen, A. Mo, in Proceedings of the 5th International Aluminum Extrusion Technology Seminar, The Aluminum Association, 1992, Vol. 1, pp. 113-120. [7] C. L. Martin, D. Favier and M. Suèry, Int. J. Plasticity, 1999, 15, pp. 981-1008.
Improved Metal Distribution during DC-casting of Aluminum Alloy Sheet Ingots Per Arne Tøndel Elkem Aluminium, Mosjøen, Norway
Gary Grealy CORUS RD&T, IJmuiden, The Netherlands
John Henry Hayes Elkem Aluminium, Mosjøen, Norway
Gabriel Tahitu CORUS RD&T, IJmuiden, The Netherlands
Einar Kristian Jensen Elkem Research, Kristiansand, Norway
Inge Jan Thorvaldsen Elkem Aluminium, Mosjøen, Norway
Dietmar Brandner CORUS Aluminium Walzprodukte GmbH, Koblenz, Germany
1
Introduction
Metal distribution is a very important factor in the production of high quality rolling ingots. It is well-known in the aluminum industry that poor metal distribution creates surface defects, such as oxide patches, pits and dimples, formation of casting cracks, an increase of subsurface oxides and reduced metal cleanliness. The “ideal” metal distributor will be one that directs the flow of metal from the vertical down spout into the mold in a quiescent manner while ensuring that temperature homogeneity of the liquid metal is maintained throughout the mold. In an exercise to identify the design criteria of an “ideal” distributor, an evaluation of a number of in-mold metal distribution systems for the vertical aluminum DC-casting process has been carried out. The principle focus of the present work is an evaluation of the combo bag distributor. The present work describes an investigation of two combo bags of different design which behave quite differently with respect to final ingot quality. The small classical combo bag widely used in the aluminum industry with approximate dimensions (LxWxH) 330x100x120 mm is referred to as the “standard bag”. The experimental bag, approx. 75% larger, is referred to as the “large bag”. As a reference, a casting without a distributor was also investigated. In order to understand how the distributor bag operates, the experimental approach of the present work includes full-scale water modeling, numerical DC-casting modeling and fullscale casting trials. The water modeling provides flow pattern visualization by ink shots into the water flowing vertically through the spout. Flow velocities inside and outside of the bag have been determined by ultrasonic analyses. The water model used in this work does not include the effect of thermal gradients in the liquid sump, i.e. the driving force for buoyancy. Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
62 The ALSIM casting model provides simulation of the 3D situation during vertical DCcasting, for a specified mold design, alloy composition and casting parameters including cooling conditions. ALSIM calculates the 3D temperature distribution in the liquid and solid metal, the corresponding temperature profiles in the mold as well as the flow patterns in the liquid metal, inside and outside of the distributor bag. The effect of buoyancy due to the thermal/gravity gradients in the sump is included. Based on the results of the water modeling and the ALSIM DC-casting model, full-scale trials were conducted in the cast house at Elkem Aluminium Mosjøen in Norway to evaluate the production performance of the two bags. Experimental ingots were cast in a dimension of 600x1685 mm with an alloy specification corresponding to AA5182. During the casting trials, the thermal profile of the liquid metal in the mold was measured. The influence of bag design on oxide formation inside the distributor has been determined. An ingot cracking database has also been established identifying the relationship between bag design and ingot cracking tendencies.
2
Water Modeling
All experiments were carried out in a full-scale water model constructed of plexi glass supported in a steel frame. The water model incorporated a simplified sump profile, the dimensions of which were calculated mathematically using a simple 3D thermal model, and can be seen in Figure 1. Flow characterization was achieved by: ink injection through the wall of the down spout and by Ultrasonic Doppler Velocity Measurements (UDVM), the principle of which is illustrated in Figure 2.
vlocal vmeasured
Figure 1: Full-scale water model
Ultrasonic probe
Figure 2: Principle of UDVM
Two operating regimes for the distributors can be identified, these being the mold fill or start up phase and steady state casting. 2.1
Start-up Phase
Bags are compared qualitatively in the start-up phase, with both bags displaying similar behavior. The fill phase is seen to be very turbulent with a high degree of “splashing”, the mixing of air and liquid. The process is illustrated in Figure 3. It is believed that the
63 “splashing” phenomenon that occurs during the start-up phase is the source of many of the inclusions present within the cast product. Sp o u t Dis trib u to r b a g
M etal
Figure 3: Mixing of air and liquid during mold fill
2.2
Steady State Phase
The fluid flow pattern during the steady state phase can be separated into two parts, these being the flow inside and outside of the bag. 2.2.1 Flow Pattern Inside the Distributor The general flow pattern inside of both bags is the same. Figure 4 illustrates a sequence of ink shot photographs taken at 0.5 s intervals.
Figure 4: General flow pattern inside standard combo bag
However, fluid flow velocities in the standard bag are much higher than those in the larger bag, due to the smaller size/volume. This results in the flow per m2 being higher, and the residence time of the liquid in the standard combo bag being lower, about 6 seconds as opposed to 14 seconds for the larger bag. The higher flow velocities in the standard bag result in a greater chance of oxide generation due to more disturbance of the metal surface, exposing the liquid metal to the atmosphere and resulting in additional oxidation. The higher flow velocities in the standard bag also increase the risk of oxide detachment from the inner net, entrainment of oxides from the liquid surface, and higher driving forces acting on the oxides to force them through the inner net. 2.2.2 Flow Pattern Outside of the Distributor The fluid flow patterns from the outside of the two distributors are illustrated in Figures 5a and 5b. The flow patterns from both bags behave in a similar manner with respect to the direction being predominately horizontal towards the narrow face of the model. However, there are three main differences between the distributors. The first being that it is very difficult to obtain symmetrical flow patterns with the standard distributor [1]. The second is that the velocities with the standard distributor were significantly higher than with the larger
64 distributor bag. Thirdly, as indicated in Figure 5b there is some downward direction to the flow, which will result in a reduced tendency for oxide entrainment from the surface of the molten metal.
Figure 5a: Flow pattern from standard distributor
2.3
Figure 5b: Flow pattern from large distributor
Velocity Measurements
Three aspects of the fluid flow in the region of the inner net are seen to be very important in relation to the presence of inclusions in the cast ingot. These are the turbulent velocity fluctuations near the inner net, which will help to detach oxides, and the maximum velocity through the inner net, which determines the forces that drive oxide inclusions from the inside of the bag to the outside. Sub meniscus velocities inside and outside the bag are important in respect to oxide entrainment from the liquid surface. In Figure 6 the horizontal velocities at the exit of the distributors are plotted. Position C0 corresponds to the center of the exit area. Positive values represent a flow moving away from the distributor, negative values represent a flow moving toward the distributor. Velocity measurements were taken 50 mm from the exit of the distributors and 70 mm below the surface of the water. Although neither bag produces a truly symmetrical flow, the large bag produces a more consistent flow velocity across the width of the distributor. Center of distributor at bag exit
80
100
60
80 ene rgy 60 (m m/ s) 40
40 20 0 -20
120
MS Turbulent Kinetic
orizontal velocity (mm/s)
100
C0 L150
L100
L50
Left of bag center
R50 R100 R150 Right of bag center
20
0 0
-40
0.1
0.2
0.3
0.4
0.5
y/L
Standard bag
Large bag
Figure 6: Flow velocities at bag exit
Standard bag
Large bag
Figure 7: Fluid flow turbulence
Figure 7 displays the Root Mean Square of the turbulent kinetic energy (3D) along the distributor length. The RMS of the turbulent kinetic energy is a measure for the velocity fluctuations. The fluctuations near the inner net of the standard combo bag are much higher than the fluctuations near the net of the large combo bag.
65
3
ALSIM Casting Model Results
Details of the ALSIM casting model are given elsewhere [2]. The material properties were calculated using the microstructure model ALSTRUC [3]. Calculated liquidus and solidus temperatures are 635 °C and 489 °C, respectively. Figure 8a shows the calculated metal flow in the liquid sump 1 and some isotherms in the vertical symmetry plane parallel to the wide side of the ingot when using the standard combo bag. It is assumed that convection is relevant in the liquid in a temperature range from the casting temperature (700 °C) down to a chosen temperature 630 °C, corresponding to a calculated liquid fraction = 0.8. At even lower temperatures, it is assumed that the flow is quickly dampened so as to be negligible. The usual dampening terms in the flow equations are adjusted accordingly. The velocity of the metal coming out of the spout is of the order of 500 mm/s, quickly decreasing to the order of 10 mm/s in the bulk of the liquid sump.
Figure 8a: Standard bag with buoyancy
Figure 8b: No buoyancy
The effect of buoyancy can be seen by comparing Figures 8a with 8b. It is clear that buoyancy leads to a “lifting-up” of the isotherms and flow directions in the right part of the diagram, in particular at the level of the distributor bag. The density increase of the liquid when lowering the temperature from 700 to 630 °C is calculated by ALSTRUC to be ~ 1.3 %. In the center of Figure 8a at the position of the liquidus temperature, there appears to be a splitting of the vertical velocities leading to a downward stream in the direction of the solidification front. As illustrated in Figure 8b with no buoyancy, this effect is not seen. Closer to the ingot short side, the liquid stream follows the solidification front. The velocities in a horizontal plane are shown in Figures 9a and 9b at the level of the lower metal exit from the short side of the combo bag. The metal stream is divided in two main streams spreading out towards the ingot rolling sides respectively. This effect is far less marked without buoyancy. In the latter case, the meniscus zone and the area between the combo bag and the ingot rolling side is also clearly colder. The outer vortex tendencies, Figure 9b, are displaced to the left, Figure 9a. 1
Please note that it was necessary to use two scales for the flow velocities: for flow within the bag and nearby surroundings the arrow-velocity scale is 10 or 25 times higher than elsewhere.
66
Figure 9a: Standard bag with buoyancy
Figure 9b: No buoyancy
Comparing the calculated temperatures in the liquid with grid measurements, 15 mm below the metal level, deviations within a range of 0-6 °C are found for the standard combo bag. Fine-tuning of the input conditions to the calculations would probably decrease this difference even further. Without the buoyancy effect, the deviations would have been markedly larger. 2
Figure 10a: Large bag
Figure 10b: Standard bag
By increasing the size of the distributor bag, there is considerably less intense flow coming out of the large bag than from the standard bag, as illustrated by Figures 10 and 11. This was confirmed in the water modeling. In Figure 11, we see the calculated horizontal velocities of the liquid metal leaving the distributor bag at a distance of 50 mm from the bag. It is shown that the velocity curves have the same profile and maximum values in the direction of the ingot short side as in the water experiments, Figure 6. The results are, however, not directly 2
When comparing calculations with measurements, it should be stressed that in real casting there will always be a degree of non-symmetrical metal flow and temperature due to the slight variations in distributor geometry and mounting of the bag. This is in contrast to the actual calculations assuming perfect symmetry across the midplanes of the rectangular ingot.
67 comparable due to different experimental conditions for the water experiments relative to the Alsim calculations.
Figure 11: Horizontal velocities (m/s)
Some consequences of not having a distributor bag are illustrated in Figure 12. The heavy downward flow from the spout leads to a deepening of the sump and a heavy upward flow at some distance from the spout. In the symmetry plane parallel to the ingot short side there is as well a tendency to a sidewise widening of the lower part of the sump.
Figure 12a: Casting without distributor bag. Plane parallel to the ingot rolling side
4
Figure 12b: Casting without distributor bag. Plane parallel to the ingot short side
Full-scale Casting Experiments
A number of different distributor bags have been tested and characterized at industrial fullscale conditions. This chapter presents data from two of the experimental bags, the performance of which has been measured for several quality criteria:
68 4.1
Ingot Surface
The combo bag plays an important role with respect to as-cast ingot surface. Surface defects like oxide patches, pits, dimples and formation of casting cracks are greatly influenced by the design and geometry of the distributor bag. Some of the defects are illustrated by Figure 13. Oxide releases from the standard bag were observed in a V-shaped pattern on the rolling side along the ingot length. The large bag investigated in this work produces a significantly better ingot surface than the standard bag. The number of surface defects is drastically reduced, and the corresponding crack frequency is improved.
Figure 13: Photograph of ingot surface defects. Standard bag
Casting cracks are nucleated on heterogeneous particles like oxide inclusions. The crack database shows that casting cracks usually are nucleated on the rolling side of the ingot, beside or close to the exit of the combo bag. 4.2
Temperature Data
A correctly designed distributor provides a homogeneous and symmetrical metal distribution that, in turn, gives an even temperature distribution reducing the thermal stresses in the solidifying ingot. Temperatures were measured using a 24-thermocouple grid. The thermocouples (TC) were coated with a thin layer of boron-nitride in order to avoid reactions with the metal. TC tips were located 15 mm below the liquid metal surface. Figures 14a and 14b present the average temperatures for each position during steady state casting for the standard and the large bags respectively. For the standard bag, the thermal difference between the exit of the bag and the short side of the mold is 25°C. With the large bag the difference is 11°C. The temperature distribution pattern is more symmetrical with the large bag than the standard bag as illustrated by the 660°C isotherm. The temperature difference for ingots cast without a combo bag is 12°C, which is close to the value for the large bag.
69
Figure 14a: Isotherms 15 mm below the melt surface, based on temperature measurements. Standard bag
4.3
Figure 14b: Isotherms 15 mm below the melt surface, based on temperature measurements. Large bag
Distributor Bag
Based on water and ALSIM modeling results, as described in chapter 2 and 3, we know that the flow velocity and turbulence inside the bags differs considerably. In order to evaluate the differences on inclusion generation, an experiment was conducted in which the bags were weighted before and after each cast. Figure 15 presents the average weight of residuals inside the bags after casting versus the ingot weight. 1000 Std. bag
900
Large bag
Average weigth [g] of residuals
800 700 600 500 400 300 200 100 0 10
12
14
16
18
20
22
24
Ingot weigth [tons]
Figure 15: Weight of residuals in the distributor bag versus ingot weight
The weight of residuals increases in both cases. There appears to be no clear difference between the standard and large bag with respect to the amount of residuals. However, visual observations made during the casting experiments indicate that the standard bag generates and releases more oxides than the larger bag. Large oxide cakes were generated at the exit of the standard bag. This result was not unexpected due to the high velocities in the standard bag and the coarser weave of the inner net. The residuals inside the standard bag are heavily oxidized compared with the residuals in the large bag.
70
5
Conclusions
Two distributor bags have been tested and characterized by different techniques. In order to understand how the bag operates, the present work includes full-scale water flow modeling, numerical DC-casting modeling and full-scale casting trials. The methods used to monitor bag performance are very suitable and complimentary means to investigate developments of metal distribution systems. The water flow modeling provides flow pattern visualization (ink shots) and flow velocities. However, it can not provide buoyancy effects since the temperature in the liquid sump is constant. The numerical ALSIM casting model provides 3D thermal and fluid flow simulation of vertical DC-casting, including buoyancy effects. Full-scale trials have been conducted to evaluate the production performance of the two bags, which strongly supports the modeling results. The main difference between the two investigated bags is that the large bag provides improved process stability during DC-casting. Fluid flow and temperature distribution patterns are more symmetrical and homogeneous. The increased dimensions combined with the finer mesh of the inner net reduce the tendency for the release of inclusions into the ingot from the larger bag. Ingot quality has been improved in terms of reducing the surface defects and ingot cracking. Customer feedback shows that the large distributor bag has resulted in a marked reduction of sub-surface oxides and correspondingly less scalping.
6
Acknowledgement
The authors are grateful to the Corus and Elkem management for the permission to publish this paper. One of the authors (EKJ) in particular would like to acknowledge much appreciated advice and assistance from Mr. D.Mortensen [4] concerning the ALSIM flow calculations.
7 [1] [2] [3] [4]
References D. Xu, W K Jones Jr, J W Evans, D P Cook, Light Metals 1998, p1045-1050. D. Mortensen: Met. & Mat Trans.B, vol. 30B, Feb.99, p119-133. L. Dons et al, Met. & Mat. Trans, Vol. 30A, Aug.99, p2135 – 2146. Mortensen, Institute for Energy Technology, N-2007, Kjeller, Norway.
Determination of Material Properties and Thermal Boundary Condition from Casting Trial on Alloy AA7075 J.M. Rabenberg1, I.J. Opstelten2, J.C. Storm1 , J-M. Drezet3 1
Corus Research Development & Technology, IJmuiden, Netherlands Corus RD&T, now with TNO Bouw, Delft, Netherlands 3 École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland 2
1
Introduction
The work described here is part of the Brite/Euram project known as EMPACT [1]. Several European partners (both from industry and university) have joined in this project to develop tools for improving the DC casting of aluminum ingots. It centers on the development of mathematical models, which describe the micro- and macro segregation; the fluid flow, and the thermomechanical behavior of the DC cast ingots. An accurate description of the thermal boundary conditions is of paramount importance for the correct simulation of the casting process. Moreover, this boundary condition in practice also serves as an important control parameter. Data from thermocouples inserted during the cast have been used to obtain accurate information about the thermal history for several alloys (AA1050, 3004, 3104, 5182 and 6063 in the frame of EMPACT). Inverse modeling techniques are used to infer the thermal boundary conditions that have occurred during that cast from the measured temperatures. This technique has now also been applied to alloy AA7075, which is very susceptible to cracking. During the casting experiment cracks were formed but the measurements were successfully completed allowing the thermal boundary conditions to be deduced.
2
Method of Determination of the Thermal Boundary Condition
2.1
Experimental Devices
Two devices were employed: • The “L-rod” [2] with thermocouples concentrated in the shell-zone of the ingot. • A “ladder” with thermocouples distributed over half the ingot width. The temperature dependence of the thermal conductivity was acquired from the temperature data obtained with the “ladder” experiment using the inverse-modeling module of Calc®MOS [3]. The thermal boundary condition during the steady state regime of casting was subsequently obtained from the “L-rod” and “ladder” combined results applying the same technique.
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
72 2.2
Data Processing
Cooling conditions are considered stationary after the start up phase. Therefore, the time history of the recorded temperatures, T(x,t), can be converted with the casting speed to a function of distance. Thus, a 2-D temperature field, T(x,y), is obtained. The heat flux, q(y), at the surface causing this 2-D temperature field can be deduced using the stationary inverse modeling module of the FEM code Calc®MOS[2,4]. The inverse modeling routine basically consists of three steps: • calculation of the temperature field in the ingot with an estimated q(y) relation as input, in which y is the vertical distance to the impingement point of the water film, • comparison of the measured temperatures with the calculated temperatures at corresponding locations, • adjustment of the q(y) relation. This cycle is repeated until the (sum of the squared) differences found in step 2 are minimized to an acceptable level. By combining the resulting q(y) relation and the temperature profile at the surface, Tsurf, along the height of the ingot, the q(Tsurf) relation can be obtained. The inverse modeling routine can also be applied to deduce thermal properties such as the thermal conductivity from measured temperatures. In that case the measured temperatures of the outermost thermocouples are used as boundary conditions. The thermocouples in between serve as a comparison for the computed thermal field resulting from a specific set of thermal properties.
3
Results
3.1
Casting Data
Alloy: Casting Speed: Cooling Water: Cast Length: 3.2
AA 7075 40 mm/min. 25 m3/hr. 442 cm.
Ingot dimension: Metal level: Casting Temperature:
440 mm x 1470 mm 80 mm above mold exit. 693 °C ± 5 °C.
Temperature Data
The temperatures measured by the ladder thermocouples at the thermocouple location (x) as a function of time can be converted to a function of distance (y) with the casting speed. The location of the liquidus and the solidus isotherm can then be interpolated in the 2-D spatial grid (x-y) as illustrated in figure 1. The good agreement between the position of the isotherms as derived from the first and second row of thermocouples on the ladder yields confidence in the accuracy of the measurements and, furthermore, confirms that the casting process is in steady state. A further presentation in figure 2 shows the interpolated temperature field in the stationary upper part of the ingot as measured by the first row of five thermocouples on the ladder. It should be noted that the field of measurement and by consequence the derived sump shape is located in a plane at about 480 mm from the ingot centerline, i.e. only 260 mm from the narrow side. Therefore, these temperatures can only be used in the validation of 3-D models. It is reasonable to assume that the actual sump is deeper in the center of the ingot.
73
0
Distance from wide face [m] 0.05 0.1 0.15 0.2 0.25
0 ROW 1: T = 615 °C
Casting depth [m]
-0.05
ROW 1: T = 478 °C ROW 2: T = 615 °C
-0.1
ROW 2: T = 478 °C
-0.15 -0.2 -0.25 -0.3
Figure 1: Interpolated solidus and liquidus isotherms resulting from measurements using two rows of thermocouples
Figure 2: Interpolated temperature field
3.3
Thermal Conductivity
With the data processing routine described above first the thermal conductivity was determined as a function of temperature. The excellent agreement between measured and calculated temperatures obtained for the final set of thermal conductivity values is illustrated in figure 3.
74 700
T0 (imposed) Tcalc1 Tcalc2 Tcalc3 T4 (imposed) Tmeas1 Tmeas2 Tmeas3
Temperature [°C]
600 500 400 300 200 100 0 -0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
500
600
Distance from start secondary cooling [m]
Figure 3: Measured and calculated temperature profiles
Thermal conductivity [W/mK]
200 180 160 140 120 100 80 60
Cp=2.35e6 J/m^3K Cp=2.8e6 J/m^3K Alstruc
40 20 0 0
100
200
300
400
700
Temperature [°C]
Figure 4: The thermal conductivity as a function of temperature
The acquired thermal conductivity as a function of temperature is shown in figure 4 and compared with values obtained from alloy composition and casting conditions by the ALSTRUC model [5]. These values compare favorably. It is interesting to note that the inverse modeling routine actually predicts an increase of the thermal conductivity above 625 °C. This originates from forced convection (fluid flow) in the liquid, which is not accounted for in the present model. Replacing the purely thermal conductivity with an apparent conductivity is commonly used to compensate this. The apparent conductivity that the inverse modeling routine suggested is approximately 950 W/mK at 650 °C, i.e. about 10x the physical value. 3.4
Thermal Boundary Condition
With the thermal conductivity acquired in the previous section the thermal boundary condition can be obtained using thermocouple data from the L-rod and the ladder experiment. The heat flux distribution as a function of the distance from the onset of secondary cooling (water impingement point) resulting from the inverse modeling routine described above is shown in figure 5.
75 At first the heat flux reaches a value of 1.3 MW/m2 in the primary cooling region (the mold) over a length of about 1-1.5 cm. Next the heat flux falls to near zero in the air-gap region. At water impingement the heat flux reaches a value of 4-4.5 MW/m2 in a region of 0.75-1 cm in length, and then rapidly declines in the downstream region. 5.E+06
4.E+06
700 q", AA7075 cast T_ surf, AA7075 cast
600
400 300
2.E+06
T_surf [°C]
q" [W/m2]
500 3.E+06
200 1.E+06 100 0.E+00 -0.4
0 -0.3
-0.2
-0.1
0
0.1
Distance from impingement point [m]
Figure 5: Resulting heat flux and surface temperature
As a result of the EMPACT project an extensive database of heat flux values as a function of surface temperature has been obtained with a laboratory set-up dependent on different operational conditions such as the amount of cooling water, water temperature and alloy type [6]. From the data obtained with the AA1050 alloy a general formulation was obtained for the thermal boundary condition in the impingement region and the down streaming region. This formulation will be referred to as the EMPACT Empirical Model (EEM) [7]. The formulation for the down streaming region is similar to the formulation derived by Weckman and Niessen [8]. It is interesting to see if the EEM that was derived from laboratory data obtained with an AA1050 block is capable to predict the temperature data from the thermocouples inserted during the AA7075 cast. The results shown in figure 6 confirm that the EEM can be applied in this case with good confidence. 700
Temperature [°C]
600 500 400 300
T_ inv0 T_ inv1 T_ inv2 T_ inv3, imposed Tmeas1 Tmeas2 Lrod, surface T_ EEM0 T_ EEM1 T_ EEM2
200 100 0 -0.4
-0.3
-0.2
-0.1
0
Distance from impingement point [m]
Figure 6: Comparison of measured and predicted temperatures
0.1
76
4 • • • •
5
Conclusions Thermocouples have been successfully introduced during casting of an AA7075 ingot. The thermal conductivity has been derived from the thermocouple data. Likewise, thermal boundary conditions were obtained for this cast. The EEM is successful in predicting the heat transfer in the AA7075 case.
Acknowledgements
The funding of this work by the European Community under the Brite/Euram Fifth Framework is gratefully acknowledged. The authors express special thanks to Einar K. Jensen for many fruitful discussions, and for performing the simulations with ALSTRUC. The co-operation with Jan Zuidema jr. from TU Delft in deriving the EEM is also gratefully acknowledged.
6
References
[1] Brite/Euram project BE95-1112, Contract No. BRPR-CT-95-011. [2] J-M. Drezet, G-U. Grün and M.Gremaud: “Determination of Thermal boundary conditions of the DC casting process using inverse stationary methods”, Light Metals 2000, TMS, ed. R.D. Peterson, Nashville US, pp. 585-590. [3] Ph. Thèvoz, M.Rappaz and J.L. Desbiolles in Light Metals 1990, Ed. Ch.M. Bickert (TMS Warrendale Pa, USA, 1990), p. 975. [4] M. Rappaz, J.-L. Desbiolles, J.-M. Drezet, Ch.-A. Gandin, A. Jacot & Ph. Thévoz, “Application of inverse methods to the estimation of boundary conditions and properties”, Modeling of Casting, Welding and Advanced Soldification Processes (TMS Publ., Warrendale, USA, 1995), pp. 449-457. [5] ALSTRUC results by courtesy of ELKEM. [6] I.J. Opstelten & J.M. Rabenberg, “Determination of the experimental boundary conditions during aluminum DC casting from experimental data using inverse modeling”, Light Metals 1999, pp. 729-35. [7] J. Zuidema jr., I.J. Opstelten, L. Katgerman, “Boiling Curve Approach for Thermal Boundary Conditions in DC Casting”, to be presented this conference. [8] D.C. Weckman and P. Niessen, “A numerical simulation of the D.C. continuous casting process including nucleate boiling heat transfer”, Metallurgical Transactions B, 13B (1982), pp. 593-602
Single-roll Strip Casting of Aluminium Alloys E.N. Straatsma, W.H. Kool and L. Katgerman Delft University of Technology, Laboratory for Materials, Delft, NL
1
Abstract
The single-roll strip casting process is simulated with FLOW-3D® to determine the influence of key process parameters on the solidified aluminium strip. The input variable involving the heat transfer coefficient is obtained from experiments. The results of the simulations were verified by casting experiments on the single-roll strip caster. The process parameters involve hydrostatic pressure, wheel velocity, casting temperature and gap distance. Also we have investigated the influence of alloy composition and casting temperature on the as-cast microstructure of the strip. The composition of the AlMnMg-alloy is varied with minor additions of Fe, Cr, and Si. These additions affect the morphology of the intermetallics and the casting temperature affects the size of the intermetallics. The size of the intermetallics becomes smaller with a higher casting temperature. The equiaxed grain size is also affected by casting temperature. The length of the columnar zone for the various alloys is similar when casting takes place at liquidus temperature and varies at a higher casting temperature.
2
Introduction
Strip casting [1-5] is a continuous casting process to produce near net shape sheet with gauges of 1 to 25 mm. Strip casting results in high cooling rates, which offers substantial metallurgical advantages including an increased solid solution level and a refined microstructure. When strip is formed the interfacial thermal conductance is one of the most important controlling parameters of the strip casting process. The conduction between the solidifying strip and the cold substrate of the wheel, is usually quantified by an interfacial heat transfer coefficient h. The value of h has a major impact on the metallurgical properties of the solidifying strip, such as coarsening of the microstructure. The known values for h in the single-roll strip casting process are 0.8 kWm-2K-1 [5] and 20 kWm-2K-1 [6] for comparable velocity and thickness. In case of twin-roll casting different contact zones are taken into account and therefore a higher initial value is found of 35 kWm-2K-1[7]. The microstructure determines for a large part the properties of the final sheet [8] and therefore it is important to know the structures that occur and how they can be influenced. The as-cast structure of solidified aluminium strip is characterised by three zones; the first zone consists of fine crystals at the wheel side, the second zone of columnar crystals, and the third zone of free solidified equiaxed crystals. The influence of casting conditions on the length of the columnar zone and the grain size of the equiaxed zone has been simulated [9] for planar flow casting conditions. It was found that the length of the columnar zone increases with Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
78 increasing casting temperature and decreasing heat transfer coefficient. The grain size in the equiaxed zone is not affected by casting temperature and increases with increasing ribbon thickness and with decreasing heat transfer coefficient [9]. In actual experiments it was found that the grain size decreases with increasing casting temperature [10] and that additions of manganese have no grain refining effect [3]. This decrease of the equiaxed grain size seems to contradict the results on simulation [9]. However, in the actual experiments the heat transfer coefficient may become larger at higher casting temperatures or wheel temperatures [11, 12], which lowers the grain size. The intermetallics found in an AlMnMg alloy are Al6(FeMn), Al15(FeMn)3Si2, and Mg2Si. They become smaller with increasing cooling rate [13]. Superheating of the melt refines the size of the Al6Mn particles and this is attributed to nucleation of Al6Mn by Al2O3 formed in a reaction of various oxides with aluminium. In this paper the heat transfer coefficient is determined and this coefficient is used to simulate the strip casting process. From the simulations the major aspects are determined that influences the strip thickness. Further, the as-cast microstructure of aluminium strip.is determined as a function of casting temperature and minor additions of alloying elements.
3
Experimental
3.1
The Heat Transfer Coefficient
In our laboratory we produce 1 mm thick aluminium strip with a single-roll strip caster. A more detailed description can be found elsewhere [4].The temperature inside the wheel is measured by a thermocouple as a function of time. The thermocouple is located at 1 mm beneath the surface of the substrate. The heat transfer coefficient is obtained by recording the temperature inside the wheel during casting and matching the measured temperatures to the calculated ones, obtained from a one-dimensional analytical approach. For the analytical description, the casting process is simplified by considering a mould with an infinite length in which a semi-infinite amount of metal, initially at its pouring temperature (Tp), solidifies. The assumption has been made that the curvature of the wheel can be ignored because the radius of the wheel is much larger than the thickness of the strip. The wheel will be considered as semi-infinite and by assuming there is no slip between solidified strip and wheel, possible convection can be neglected. A detailed description of the used equations can be found in [5]. The temperatures inside the wheel are computed with h as a variable. The temperature profiles thus obtained are compared to the temperatures measured inside the wheel. In this way it is possible to conclude for which value of h the two profiles match. 3.2
Strip Thickness
The strip cast process is simulated with the commercially available CFD software package FLOW-3D®. The assumptions are that the fluid is incompressible and the transport is 2D. The heat transfer coefficient is based on the results of the former experiment. To determine the influence of the heat transfer coefficient on the strip thickness this parameter is varied from 10 to 50 kWm-2K-1. Further, the surface tension is constant and the viscosity depends on temperature. The pressure is varied from 400-1150 Pa, velocity of the wheel is varied from
79 0.43-2.50 m/s, the casting temperature is varied from 943-993 K and the gap between wheel and delivery system is varied from 1-3 mm. 3.3
Microstructure
Aluminium alloys are poured on the static copper wheel of the laboratory strip cast unit. For all castings a thickness of 10 mm was maintained. The wheel substrate has a uniform temperature of 20°C and the melt is poured with a superheat of 0°C or 70°C. Three compositions are used based on the AA3004 alloy, AlMnMg with 0.24 % Si, one with 0.74 % Fe and one with 0.07 % Cr (mass percentages).
4
Results and Discussion
From the temperature measurements a h value of about 1 kWm-2K-1 is found as can be seen from figure 1. We can verify the heat transfer coefficient by calculating the strip thickness [14]. Since we are casting a strip with a thickness of about 1 mm and taking into account a maximum cooling length of 450 mm (circumference of half the wheel) a strip thickness is then obtained of 0.33 mm. This indicates that there has to be another contact zone where the liquid has optimal contact and gives thus a higher coefficient. We assume a heat transfer coefficient of 35 kWm-2K-1 for the optimal contact zone [7].
temperature (°C)
140 120 100 80 60 40 20 0 0
1
2
3
time (s)
4
5
6
Figure 1: Experimental () and calculated temperatures with h= 2 kW/m2K (- - - -), 1 kW/m2K (–––) and 0.5 kW/m2K (000). Location: 1 mm beneath the wheel surface
The simulation results of the influence of a variable heat transfer coefficient on the length of the solidification zone shows that the length decreases from 100 mm to 20 mm with a heat transfer coefficient that increases from 10 to 50 kWm-2K-1. Also it is found that a variation of the heat transfer coefficient does not influence the strip thickness. Further, it turns out that the strip thickness is most affected by the hydrostatic pressure, the roll speed and the distance between wheel and delivery system [15]. Figure 2 shows the influence of the casting temperature on the microstructure. For all alloys, the transition of columnar to equiaxed crystals is clearly visible for castings with no superheat (Fig. 2a). The length of the columnar crystals ranges from 1.6 mm to 2.2 mm. The alloy that contains chromium gives the largest columnar length.
80
a) b) Figure 2: Typical as-cast macrostructure. Wheel temperature: 20 °C. (a) superheat: 0 °C; (b) superheat: 70 °C
The average grain size in the equiaxed zone is about 89 µm, which is approximately constant for the various alloys. Fig. 2b shows a typical macrostructure from an alloy that is cast with a superheated melt (70 °C) on the wheel at room temperature. The transition of columnar to equiaxed crystals is not as clear as with a superheat of 0 °C. The length of the columnar crystals varies for the different alloys and ranges from 2.3 mm to 6 mm. For all alloys the grain size in the equiaxed zone is coarser than with a superheat of 0 °C and ranges from 290 µm to 880 µm. The chromium-containing alloy gives in all cases the largest values. The effect of superheat on the columnar length is confirmed by literature [9]. The increased grain size in the equiaxed zone found with higher superheat is not supported in literature by simulations [9] or by actual experiments [12]. A possible explanation might be the remelting of grains caused by the released latent heat of other growing grains [9], which was not accounted for in [9]. The observation reported in literature [3] that the grain size is not much affected by minor additions of alloying elements is valid in case of no superheat. The coarsening of the grains is caused by the influence of chromium [3] and this is strongly visible in case of superheat. The largest differences are found for the morphology of the intermetallic Al15(FeMn)3Si2. Figure 3 shows the different morphologies found for the Al15(FeMn)3Si2 intermetallic. Chinese script (Fig. 3a) is the most common morphology and is clearly visible in AlMnMg+Si and AlMnMg+Cr alloys at a wheel temperature of 20 °C and a superheat of 0 °C. The morphology does not change for AlMnMg+Si with a superheat of 70 °C, whereas the other alloy, AlMnMg+Cr, shows a slight change of the morphology into needle-shape. The second type of morphology is a coarse plate shape such as the one shown in Fig. 3b; it is present in the AlMnMg+Fe alloy for both conditions.
a) b) Figure 3: Morphologies of the intermetallic Al15(FeMn)3Si2 of three alloys. (a) AlMnMg+Si or AlMnMg+Cr, (b) AlMnMg+Fe. Superheat: 0 °C, wheel temperature: 20 °C
81
5
Conclusions
During solidification of the strip on the wheel there are two contact zones. The strip has initially optimal contact. That contact time is short and a high heat transfer coefficient is present. After this first zone a second zone of non-optimal contact is present. In this zone the heat transfer coefficient is approximately 1 kWm-2K-1. The strip thickness is not influenced by the casting temperature and the heat transfer coefficient, but the contact time is. The thickness is mainly influenced by pressure, wheel velocity and the distance between wheel and delivery system. The columnar length and the grain size are affected by the casting temperature and become larger with increased casting temperature. Remelting of the growing grains might cause the increased grain size. A small addition of Si, Fe or Cr results in different morphologies of the intermetallic phase.
6
References
[1] K. Tada, A. Ohno, Aluminium, 1993, 69, 1092, [2] E.F. Emley, International Metals Reviews, 1974, 21, 99, [3] M. Cortes, Light Metals 1995, (Ed.: J. Evans), The Minerals, Metals & Materials Society, Warrendale, PA, USA, 1995, 1161, [4] E.N. Straatsma, W.H. Kool, L. Katgerman in Light Metals (Ed.: C.E. Eckert), The Minerals, Metals & Materials Society, Warrendale, PA, USA, 1999, 919, [5] E.N. Straatsma, W.H. Kool, L. Katgerman, Materials for Transportation Technology, Euromat 99 (Ed.: P.-J. Winkler), VCH, Weinheim, D, 1999, 1, 58, [6] G. –X. Wang and E. F. Matthys, Melt Spinning, Strip Casting and Slab Casting, (Eds.: E. F. Matthys and W. G. Truckner), The Minerals, Metals & Materials Society, Warrendale, PA, USA, 1995, 205, [7] M.J. Bagshaw, J.D. Hunt, R.M. Jordan, Cast Metals, 1988, 1, 16, [8] G. Moritz and F. Ostermann, J. Inst. Metals, 1972, 100, 301, [9] L. Granasy, A. Ludwig, Melt Spinning and Strip Casting - Research and Implementation, (Ed.: E. .F. Matthys), 1992, 53, [10] E. Vogt, G. Frommeyer, J.E. Wittig, Mat. Sci. Eng., 1988, 98, 295, [11] L. Granasy and A. Ludwig, Mat. Sci. Forum, 1991, 77, 211, [12] S.C. Huang and H.C. Fiedler, Mat. Sci. Eng., 1981, 51, 39, [13] L. Bäckerud, Solidification Characteristics of Aluminium Alloys, Skan-Aluminium, Oslo, N, 1986, 1, 101, [14] D.R. Poirier and G.H. Geiger, Transport Phenomena in Materials Processing, The Minerals, Metals & Materials Society, Warrendale, PA, USA, 1994, 343, [15] E.N. Straatsma, W.H. Kool, L. Katgerman, to be published.
Continuous Casting of Semisolid Al-Si-Mg Alloy Tetsuichi Motegi and Fumi Tanabe Chiba Institute of Technology, Department of Metallurgical Engineering
1
Abstract
One of the most interesting cast technologies is the semisolid casting process and its products. Ingots made by the semisolid casting are utilized to the thixocasting process. The advantages of this process are the low casting temperature, good fluidity of the semisolid alloys, and homogeneity of the solidified structures. In general, the semisolid alloy ingots are produced by either electromagnetic stirring or mechanical stirring of the molten alloys. However, both require huge equipment and the costs are high. We have invented a new casting machine for semisolid alloy. It possesses an inclined cooling plate, which is used to make numerous crystal seeds and to disperse in the molten alloy before casting. Molten alloy with crystal seeds flow into a tundish and the seeds grow granular crystals within it. Thus the semisolid alloy is cast by a DC continuous casting machine. In this investigation, a molten Al-Si-Mg casting alloy was poured down a cooling plate to form abundant crystal seeds in the molten alloy. It was then cast into the horizontally continuous casting machine.
2
Introduction
Recently, utilization of aluminum alloy products tends to expand rapidly in various industries due to light weight, thin form, small size, high qualities, low cost and recycability. In order to produce aluminum alloy materials, continuous casting of semisolid alloy has been an attractive method. It is known that semisolid alloys containing high solid fraction exhibit high fluidity. At present various continuous casting methods of semisolid alloys are known in commercial processing. Thus, the rheo casting and thixocasting processes are very useful for producing homogenously solidified structures in castings. Ingots used in the thixocasting process must have granular, fine and homogeneous crystal grains. The remaining primary crystals in the liquid alloy play an important role in forming the grains that make up the homogeneous cast structures. Ohno and Motegi have proposed the crystal separation theory to explain granular crystals in castings and ingots; that is, granular crystal nucleate and grow on the mold wall, and then are separated from there by fluid motion.(1) This theory was applied to this continuous casting process to obtain the granular crystals in the solidified structure; that is, many seeds of primary aluminum are formed on the cooling plate and flow into the mold with the molten alloy. The seeds of aluminum grow granular grains during the continuous casting process. In this investigation, a molten Al-Si-Mg casting alloy was poured down the inclined cooling Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
83 plate and then cast into the horizontal continuous casting mold. The best advantage is the very simple process for producing the semisolid alloys.
3
Experimental Procedures
The chemical composition of Al-Si-Mg alloy is shown in Table 1. A liquidus temperature of 614° and a solidus temperature of 545° of this alloy were determined by the thermal analysis method. Figure 1 shows a schematic of the apparatus for horizontal continuous casting machine in this investigation. Figure 2 shows a schematic of the apparatus for vertical continuous casting. Breakout of the molten aluminum alloy often occurred in the vertical type machine. Hence, we used the horizontal continuous casting machine. It consists of the electric furnace, the incline plate with cooling zone, and the watercooling copper mold. Six kilograms of alloy were melted and held at the constant temperature in the furnace. The ceramic rod of 100mm in dia. with a heater as shown in Figure 1 was lowered a constant speed into the molten alloy, then flowed through a taphole near the top of the crucible. This method makes it possible to control the casting speed of the molten alloy.
Figure 1: Schematic illustration of apparatus for the horiztontal continous casting of semisolid alloy
The pouring temperature of the molten alloy onto the cooling plate was 634 to 724°. A 160 mm long inclined plate was used to promote formation of the seeds of primary aluminum crystal. The molten alloy was poured down the plate and then cast into the horizontal continuous casting mold. Each ingot was examined metallographically. In order to examine the sizes and numuers of aluminum crystal at various stages of this process, the rapid cooling of the molten alloy was performed. Each sample obtained was examine metallographically. Each ingot was also examined metallographically and was measured the grain size of aluminum.
84
Figure 2: Schematic illustration of apparatus for the vertical continous casting of semisolid alloy
4
Results and Discussions
The solidified structures obtained by both vertical and horizontal continuous casting were shown in Figure 3. All ingots exhibited a network of aluminum dendrites throughout the ingots as shown in Figure 3 when no cooling plate was used. On the other hand the dendrites became granular crystals by using the cooling plate.
Figure 3: The solified structures of Al-Si-Mg-alloy abtained by continious casting process
85 The crystal separation theory proposes that granular crystals nucleate and grow on the mold wall, and then are separated from there by fluid motion.(1) This theory was applied to this process to explain the presence of granular crystals in the solidified structure; that is, many seeds of primary aluminum are formed on the cooling plate and flow into the mold with the molten alloy. The seeds of aluminum grow granular grains during the continuous casting process. Figure 4 shows the solidified structures obtained by a rapid cooling device at various stages during the horizontal continuous casting. No aluminum crystals appeared before the cooling plate, but numerous crystals appeared in the sample which was taken under the cooling plate. These crystals grew granularly in the tundish as shown in Figure 4. The semisolid alloy containing granular crystals was cast in the mold and solidified structures of the continuous casting ingots were consisted of granular aluminum crystals. Table 1: Chemical compositions of Al-Si-Mg alloy used. Si Mg Fe Cu Zn Mn Ti V 6.79 0.45 0.14 0.04 0.020 0.02 0.15 0.008
Al bal
Figure 4: Primary aluminum crystals obtained by rapid cooling at various positions of the horizontal continous casting process
86
5
Conclusions
To make ingots of the Al-Si-Mg casting alloy used for the thixocasting process, semisolid casting was performed. An inclined cooling plates was used to generate the crystal seeds. The seeds of primary aluminum crystals grew in a molten alloy tundish just beside the mold prior to entering the mold of the casting machine. The results obtained are as follows. • 1. The cooling plate is effective in generating crystal seeds in the molten alloy. • 2. Lower casting temperature in the mold yield more granular crystals of aluminum. • 3.The inclined cooling plate and tundish are useful for refining the cast structure of continuously cast Al-Si-Mg alloy.
6
References
[1] A.Ohno,T.Motegi and H.Soda, Trans.Iron and Steel Institute of Japan, 1971,1,18.
Casting Technology and Processes
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
Yield and Quality Improvements for Semi-Continuously Cast Copper Alloys Carl-Michaël Raihle Outokumpu Process Automation, Västerås, Sweden
1
Abstract
During the last 25 years a lot of work has been done to improve inner quality and yield in continuous and semi-continuous casting. One of the methods proposed to accomplish this is high water cooling at the end of the liquid pool, so called Thermal Soft Reduction (TSR). This paper will discuss how to use TSR in order to improve quality and yield for semi continuous casting of copper alloys.
2
Introduction
Centerline porosity, segregation and yield are central issues in continuous casting of metals. These have all been studied extensively in the literature (1-4) and it is well documented that one root cause is thermal stresses caused by different thermal gradients in the solidifying metal. In copper base alloy production semi-continuous casting is still a frequently used production method. With this method there are frequent starts and stops during production and it is of high interest to decrease the yield losses associated with the finishing of each casting. One major reason for the yield loss is the shrinkage formation, or pipe, that occurs at the end of each casting. Other end defects are inclusions of cover material, refractory and slag. The end section has to be cut off before further processing of the material to avoid defects in the finished product If the end defects can be decreased there is a large possibility for quality and yield improvements. The aim of this paper is to briefly discuss the reasons for pipe formation and centerline inhomogenities and show how they can be influenced in practice by controlling secondary cooling.
3
Theoretical
The solidification starts at the meniscus, i.e. the region where liquid metal comes in contact with the mould. A solidified shell develops rapidly and grows towards the center of the cake. After a certain thickness the shell is strong enough to withdraw from the mould wall and an
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
90 airgap forms between mould wall and cake surface. At this time the heat transfer between cake and mould decreases several magnitudes and the growth of the shell slows down. When the cake exits the mould it enters a secondary cooling zone with water sprays and heat transfer increases dramatically, the surface temperature drops rapidly and shell growth increases. After the sprays there is normally an aircooling zone with low heat transfer, the surface temperature increases and shell growth decreases. Further down there is usually a water bath that cools down the cake to room temperature. The changes in surface temperatures does not only influence shell growth but also stresses and strains in the solidifying shell and the flow of liquid metal inside the cake. For a more detailed theoretical description the interested reader is referred to literature in this field (1-7). Here only a simple qualitative discussion will be made. The solidifying shell can be simplistically looked upon as two beams connected at the ends (figure 1). One side of the beam is the interface solidified shell / liquid pool and the other side interface solidified shell / surroundings. A typical temperature-time curve for the shell is shown in figure 2. As can be seen in the figure there is a large temperature drop along the centerline axis when the center starts to solidify. If there is no corresponding drop at the surface of the cake, thermal contraction will force the shell (or beam) to bulge, figure 1.
Liquid pool Solidified shell
Beam Liquid pool Beam
Deflected beam Liquid pool Deflected beam
Figure 1: Cross-section of a cake. In order to simplify the discussion the solidified shell is assumed to behave like a beam, temperature variations at the surrounding and liquid pool interfaces will cause the beam to deflect as shown in figure.
Liquid will flow downwards to replace the volume change caused by thermal bulging. This downward flow can be very strong and if some foreign particles are introduced in the liquid pool there is a risk that these will be sucked down and trapped in the solidification front, later causing defects in the strip. A number of casting defects are associated with these thermal stresses and induced flow. The most important are: • Centerline porosity • Surface cracks • Centerline cracks
91 • •
Halfway cracks (ghost lines) Centerline segregation The pipe, or shrinkage porosity, is also influenced by the thermal stresses and liquid flow. The size of the pipe can be decreased and in some cases even almost eliminated by adjusting casting speed and / or cooling program.
Figure 2: The figure shows the measured temperature for the surface and the center for a semi-continuously cast cake. Note the difference in cooling rate at the end of the liquid pool
One technique has earlier been used with success to decrease pipe formation and segregation in continuously cast steel billets and slabs, the so called Thermal Soft Reduction (TSR) (4-7). With TSR the aim is to decrease the difference in cooling rates between the center and surface of the cake at the liquid pool end, thus reducing the thermal bulging. This is achieved by increasing secondary cooling at the liquid pool end.
4
Experimental
The production environment at Outokumpu did allow changes in secondary cooling and casting speed during the trials. By varying these two parameters a large number of trials were performed with varying settings to achieve different values of cooling rate ratio: dTcenter (1) dt dTsurface dt Figure 3 shows a pipe from normal production with a normal pipe depth of 0.10 m. As can be seen there is an increased amount of slag and porosities associated with the pipe, these sometimes penetrates much deeper into the casting than the actual pipe itself.
92
Figure 3: Pipe around 0.010 m deep with normal casting procedure
Figure 4 shows another pipe with a different setting of cast speed and secondary cooling, cooling rate ratio has been much closer to unity in this case. Here the pipe depth has been decreased to 0.035 m. It can also be noted that the top of the cake bulges upwards which is a sign that liquid is squeezed upwards and overflowing the top of the cake when casting parameters are correct.
Figure 4: Reduced pipe with the use of TSR, pipe depth is 0.035 m
5
Discussion
From the experimental results it is clear that the location of secondary cooling in relation to the liquid pool end has a very high influence on the pipe formation. When the secondary cooling is located at the correct position the pipe depth is decreased with more than 60%. This result agrees with figure 5 (7), which shows theoretically how the pipe depth and centerline macrosegregation varies with the position of an increased cooling zone in a continuous caster.
93 As seen they can be nearly eliminated if the casting speed and cooling program are adjusted correctly.
Figure 5: The simulated variation in pipe depth and centerline segregation for a steel billet. The figure shows how pipe depth and segregation ratio varies depending on the location of the extra cooling. Two different cooling programs for the secondary cooling are used, the horizontal lines shows the values for pipe depth and segregation ratio for both practices. The abscissa gives the distance from meniscus for start of extra cooling zone in [m]
6 • •
7 [1] [2] [3] [4] [5] [6]
Concluding Remarks Thermal stresses have a large influence on pipe formation, central macrosegregation and cracking. To minimize pipe, central macro segregation and cracking the cooling system must be arranged so that the cooling rate ratio between center and surface is close to 1 at the end of the liquid pool.
References
H. Mori, N. Tanaka, N. Sato, M. Hirai;Trans JSJ Japan, 12,1972, p. 102-111 H. Bauman, W. Löpman; Wireworld Int. 16, 1974, p.149-155 J.K. Brimacombe, K. Sorimadri; Met. Trans. 8B, 1977, p.489-505 G. Engström, H. Fredriksson, B. Rogberg; Scand. J. Met. 12, 1983, p.3-12 K. Miyazawa, K. Schwerdtfeger; Arch. Eisenhuttenwes, 52, 11, 1981, p.415-422 P. Sivesson, C-M Raihle, J Konttinen; The Symposium on Advances in Solidification Processes, E-MRS, 1993, Strasbourg, France, Mat. Sci and Eng., vol. A173, p.299-304 [7] C-M Raihle, H. Fredriksson; Met and Mat Trans B, 25B, 1994, p.123-133
Continuous Casting Technology for Magnesium U. Holzkamp (Sp), H. Haferkamp, M. Niemeyer University of Hanover, Institute of Materials Science, Hanover, Germany
1
Introduction
Magnesium alloys as structural materials with high weight specific properties have again gained in importance as an alternative lightweight construction material in the last years. Because of this development a high increasing consumption of magnesium components is predicted for the future. This development mainly takes place for the magnesium diecasting. The production of components is nearly restricted to 90 % to the diecasting process . The portions of wrought alloys or formed components and semi-finished products will be almost ignored in this article /1, 2/. But particularly components which were produced by forming and a suitable thermal treatment have a great potential for high-tensile and ductile components at the same time (figure 1). By applying the thermal forming like extrusion, rolling and forging a wrought alloy free of pores and sink-holes is guaranteed. Wrought alloys additionally can be distinguished by improved ductile and mechanical strength properties. This predetermines the wrought alloys for applications with high requirements on security relevance like for example in crash behaviour /3/. Mechanical Properties of DC and Wrought Alloys Yield stress (0,2%) MPa
250 200 150
Wrought Alloys
100 50
DC-Alloys
0
0
5
Tensile %
10
15
Figure 1: Comparison of the mechanical properties of DC- and wrought-alloys /4/
The diecasting method permits because of high cycle times and good casting properties of most of the magnesium alloys a very effective production. Production characteristics like locking pressure and limited flow length and mould filling capacity restrict the producable size and thickness of the components. Furthermore with diecasting alloys in fact high mechanical strength properties can be obtained but the reached ductilities depending on the solidification are comparatively low. Magnesium wrought alloys present here a great extended property spectrum. Apart of improved mechanical properties and a more balanced strength and ductility relation, thin and also bulk components can be produced by the rolling, extrusion Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
95 and forging process through thermal forming. Apart of the improved mechanical properties a thermal treatment to optimise the properties is possible and also an extended use of assembling processes like e.g. the arc welding process. Although the basic principles of the magnesium forming are well-known for a long time /4/ components of magnesium wrought alloys are despite of their good properties not used very often. The available potential of wrought alloys can be verified by varied applications in the past, e.g. for the American bomber B-36 approximately 9000 kg magnesium was used, a great part were sheets for the shell /5/. Today investigations and development also show the recognised efficiency of this series of material. Especially in the sector of the alloy development and optimisation of the process parameters for the extrusion and the sheet metal production considerable improvements are made /6/. Despite of these tendencies this series of material is not very important in contrast to the developments in the diecasting process. Apart from deficiency in the alloy sector and necessary adaptation in the metal forming technology this can be especially put down to economical aspects. Higher costs regarding the unmachined parts restrict the increased application in the industry /3/. A greater use of the material potential of wrought alloys requires therefore a material suitable and economical manufacturing techniques at the same time to put down semi-finished products for the metal forming which are cost effective and of a high quality /7/. To produce semi-finished products in a sufficient quantity and quality, for other typical forming materials like aluminium, copper or steel the continuous casting process is used for decades. Apart of the continuous production of great material quantities the continuous casting permits also a high-purity material processing and the setting of structural characteristics which are relevant for the forming. Improvements which were reached for example with applied and magnesium adapted mould technologies seem to be as an application out of the aluminium continuous casting sector which is considerably further developed /8/. Also the security sector with regard to the high danger of fire and explosion of molten magnesium, remembering the accident in the continuous casting plant of Norsk Hydro /9/, Canada, has a great development potential. Above all the choice of the secondary cooling medium of the melt conduction needs to be considerably optimised. At present only 17 % of the produced continuous casting products are used as forming semi-finished products /10/. The rest of the material is used to alloy up the aluminium and for the GGG production. It can be assumed that with the control of the solidification with regard to grain refining and structural homogeneity which are necessary to improve the forming conditions of the magnesium which is difficult to form not a lot of experience was made. In this sector of research extensive investigations are imperative to guarantee a magnesium suitable continuous casting technology in the long-term.
2
Continuous Casting Technology
2.1
Process Control
The continuous casting in general and the processing of the magnesium in particular makes great demands on the materials processing. These demands were the basic conditions of the technique described in the following.
96 In contrast to common ingot casting technologies the continuous casting stands out due to a continuous dross of the billet. This means a lot of control work for conducting the melt. To achieve a high-grade billet quality a constant level in the mould must be guaranteed all the time. In spite of the fact that an overflowing or draining off the mould must be prevented, a stationary state can be produced because of the thermal conditions in the mould and also a homogenous billet. A control must be applied which correlates the proportioned quantity out of the furnace or tundish into the mould with the level of the mould and the discharging velocity. Apart of a correct dosage, this requires especially a level measuring method in the mould itself which is complicated for molten baths (figure 2).
tundish
dosing furnace
level measuring
dosing melt volume
dischargingvelocity
Figure 2: Interrelationship of process parameters of the continuous casting process
2.2
Physical and Chemical Particularities of Magnesium
In contrast to the common metals used in the continuous casting process like aluminium, copper and steel magnesium can be distinguished by some physical and chemical properties which should be not ignored in the development of a magnesium adapted casting method. Here the thermophysical characteristic values like fusion heat and thermal capacity have to be emphasised out of which information about the quantity of heat which the mould dissipates can be received.
4 3
1074
1829
2143
0
2
640 2,23
500
3,07
1000
Mg
Al
Cu
St
1
3
1500
5
(kJ/dm )
5,42
specific heat capacity
3,89
3
specific heat of fusion (kJ/dm )
2000
specific heat of fusion
specific heat capacity at 800K
6
2500
0
Figure 3: Specific heat values of different metals in comparison to Mg
97 Even with regard to aluminium, significant lower thermophysical values can be seen (figure 3.). Because of these values on the one hand less heat must be extracted out of the mould for the solidification but on the other hand a continuous heating of the melt current is necessary to guarantee an optimised casting temperature. Another fundamental problem of the processing of magnesium with the continuous casting method is the great oxygen affinity. Apart of the endangering potential through magnesium fires, especially when the melt overflows uncontrolled, partially appeared oxidation products must be avoided as they have a negative influence on the quality of the casting product. Because of this reason a totally continuous casting process is desirable which avoids any contact with the atmosphere. The existence of a suitable protective gas in the casting installation leads to a high-purity furnace atmosphere. 2.3
A Magnesium Adapted Continuous Casting Concept
bar
billet
dummy
On the basis of these technological and physical conditions a continuous casting concept was developed. The most typical feature of the continuous casting technology is the chosen drawing off direction of the billet. In contrast to conventional methods the billet is not produced in a horizontal or vertical way but it is rising upwards out of the mould. This concept (figure 4) has above all during the processing of magnesium the following advantages in contrast to conventional continuous casting processes: • Robust dosage: Because of the hydrostatic dosing principle no further dosing control besides opening and closing the stopper in the dosing furnace is necessary. The drawing off direction upwards facilitates a continuous filling of the mould from below. The material which is drawn off out of the mould as a (partly) solidified billet is feed continuously because of the level difference between furnace and mould. Therefore a complex level measuring in the mould is not necessary and a thermodynamic stationary condition during the casting process is guaranteed.
mould
furnace
heated dosing tube Figure 4: New magnesium adapted continuous casting concept
•
Melt control poor in contamination: The installation is completely protected from the surrounding atmosphere. The melting and dosage furnace is swept with a protective gas and the heated dosage tube is connected with the furnace and the mould pressure sealed. For this installation model argon is used as a pure inert gas which is mixed with CO2 to pro-
98
•
•
2.4
duce a protective coating.The casting material just gets into contact with the atmosphere in a solidified condition. This sweeping also has in this installation a cooling function. Flexible temperature control: On the basis of the sensitive thermal behaviour (low heat of fusion and heat capacity) the whole flow of the melt from the furnace to the mould is heated. Furthermore the mould has two cooling ducts independent from each other which facilitate a optimal carrying-off of heat over the length of the mould and have a positive influence on the solidification behaviour of the magnesium. Security of the installation: As the area between the molten and the solidified condition at the mould exit is only very short and the melt permanently is filled up an overflow can be expected when the billet tears or brakes. This represents especially magnesium with its high fire and explosion potential a great security risk. The designed installation reduces this risk to a minimum, as in case of a break of the edge shell the melt in the mould does not overflow because of gravitational forces. Furthermore at the mould exit the overflow of the melt is controlled by a sensor. If the melt comes into contact with the sensor the dosage will be automatically interrupted. The protective gas layer above the mould prevents an ignition of this melt and provides the best possible security for the operational staff and the installation. Simulation of the Start Up
Before the installation was initially operated a solidification simulation with ANSYS took place. To restrict the casting parameters like melting and mould temperatures of the starting operation for the real operation a model which bases on the real mould was produced and that represents the solidification time and course. As the mould is rotationally symmetrical only one half was calculated because of symmetrical conditions. Figure 5 shows a mould 3s after the feeding with melt. The solidification contour and the course of the liquid phase in the mould can be clearly seen. The first operational attempts were made on this basis. dummy bar mould (cooling system)
melting
dosing tube (heated)
Figure 5: Simulation of the start up process
99
3
Continuous Casting Model Facility
An installation model was designed and constructed based on the presented new magnesium suitable continuous casting principle (figure 6). This installation concept follows the principle of the hydrostatic dosage with a furnace which is placed a little bit higher and opposite to the mould. The melt dosage is effected by a stopper in the bottom of the converter. The furnace principle is pressure sealed and facilitates with this a melt treatment poor in contamination under a protective gas. For the first preliminary tests the inert gas argon was used.
dummy bar mould dosing tube
complete facility
Figure 6: Magnesium adapted continuous casting model facility
The mould of the installation model has a casting cross section for the billet of 40 mm and therefore makes a production of small billets under the variation of the different parameters which are relevant for the continuous casting possible.In figure 6 the constructed installation with furnace, u-shaped feed tube, mould and the drawing off facility which draws off the starting bar in a velocity interval of 0-30 cm/min upwards by means of an electric motor with a gear. To heat the furnace and the feed tube resistance heatings are used. The mould can be thermally manipulated by a combined heating and cooling system. The main emphasis was put on the investigation of a regulated starting behaviour which is necessary for a continuous casting operation. That is why the results of the simulation were consequently turned in these tests and led to the production of first short billets of pure magnesium.
4
Conclusion
Regarding the deficits and potentials in the field of wrought alloys the development of a magnesium suitable continuous casting technology to increase the spectrum of use is necessary. The specific requirements for such a technology was shown and the implementation of an installation model basing on a new concept was described. Therefore the Institute of Materials
100 Science has an installation at its disposal which first facilitates the determination of all important casting parameters and in the longer term to transfer these on an industrial scale.
5
Acknowledgements
The research has financially been supported by the Deutsche Forschungsgemeinschaft (DFG). Special thanks to the Institutes of Mathematics, University of Hanover, for executing the numeric simulation
6
References
[1] Edgar, R.L.: Magnesium Supply and Demand 1998. IMA, 53th Annual World Magnesium Conference, Rom, 1999, 1 - 6 [2] N.N.: World-wide Demand for Magnesium Diecastings. Business brochure Norsk Hydro, 1998 [3] J. Becker, G. Fischer: Strangpreß- und Schmiedeerzeugnisse aus Magnesium-sicheres und leistungsfähiges Halbzeug für den Leichtbau. 16. Umformtechnisches Kolloquium Hannover, 25-26 Feb. 1999 [4] G. Schichtel: Magnesium-Taschenbuch, VEB Verlag Technik Berlin, 1954, 246-276 [5] N.N.: Magnesium Boosts Performance of Aerospace Systems and other Defense Equipment. Metalscope, Brooks & Perkins, May 1964 [6] Friedrich, H., Schumann, S.: The second age of Magnesium - Research Strategies to Bring the Automotive Industry’s Vision to Reality. Proceeding of the second International Isreali conference on Magnesium Science& Technology, 22-24 Feb. 2000 [7] Pinford, P.M.D.: Wrought Magnesium the next challenge. Light Metal 1997, Proceeding, Ontario, 85-93 [8] Kittilsen, B., Pinfold, P.M.D.: Recent Development in large Format Magnesium Casting. Light Metals 1996, The Minerals, Metals & Materials Society, 1996 [9] Pawlek, R.P.: Magnesium Aktivities at the Turn of the Year 1999/2000. Metall, 54. 3/2000, 96-99 [10] Baker, P.: Issues in Magnesium DC Casting. Light Metal 1997, Proceeding, Ontario, 1997, 355-367
Local Distribution of the Heat Transfer in Water Spray Quenching F. Puschmann, E. Specht, J. Schmidt University of Magdeburg, Institute of Fluid Dynamics and Thermodynamics, Magdeburg, Germany
1
Introduction
In most cases efforts are made to use continuous casting water spray quenching to achieve uniform cooling with a locally independent heat transfer. However, the heat transfer coefficient obtained in this cooling task with spray water depends on various conditions. Among them, the surface temperature of the material subjected to cooling and the characteristics of the spray generated by nozzles exert an influence on the heat transfer coefficients achieved /1/-/3/. The spray characteristics for full-cone nozzles, e.g. impingement density, drop diameter and drop velocity, in a defined plane in front of the nozzle are a function of the radial distance, even in the ideal case of a symmetrical spray jet. In the continuous casting process materials are frequently cooled by means of panels of flat-spray nozzles, often causing areas without water impingement and overlap areas of individual nozzles. The results of the locally non-uniform spray characteristics and interactions in nozzle panels are that water is not applied in a uniform way to the entire material subjected to cooling and, hence, cooling intensities vary locally. Investigations focussed on determining the influence exerted by surface temperature, water impingement density, drop diameter and drop velocity on local heat transfer in water spray quenching in the film boiling range.
2
Experimental Set-up
2.1
Drop Diameter and Velocity
The measuring set-up sketched in Fig. 1 was constructed to investigate drop size and velocity. It served to measure the distribution of drop sizes and drop velocities of water sprays by means of a 2D-Phase-Doppler-Anemometer (PDA). The PDA is an optical measuring system which is suited to performing non-contact and simultaneous measurements of the velocity and the diameter of spherical particles. The sensing volume of the PDA is fairly small, ensuring high resolution in terms of time and area. It is able to identify individual drops passing through the sensing volume. Due to its high laser performance of 4W it is an excellent device for performing measurements in misty and steamy environments. Figure 2 depicts the distribution of drop size and drop velocity for a flat-spray nozzle as used in continuous casting. The mean volumetric diameter is presented as a function of the measuring position. The measuring plane was located at a distance of 200mm in front of the nozzle. The measuring position was the distance measured across the length of the flat jet from its centreline. A pressure of 5bar was applied to the nozzle resulting in a water flow rate of 318kg/h. The mean volumetric diameter was about 110µm in the centre of the spray jet. Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
102 The drop size increased towards the border. Reducing towards the border, the mean drop velocity was about 6.4m/s. It should be noted that the aperture angle of the flat-jet nozzle was 105° and, hence, the overall length of the spray jet in the measuring plane was about 520mm. The figure shows a section of the centre exhibiting a width of 200mm. Air
Water Nozzle
Probe volume
Laser (max. 4 W)
30 ° (off-axis angle)
Transmitter
Receiver Analysis unit
150
7
140
6,5
130
v
120
6
D30
5,5
110 100 90
5
Nozzle: D25381-13-105/20-1 Measuring planes: 200 mm Nozzle pressure: 5bar Flow rate: 318kg/h
80 -100
-50
Mean drop velocity [m/s]
Mean volumetric drop diameter [µm]
Figure 1: PDA System
4,5
0
50
4 100
Measuring position [mm]
Figure 2: Drop Size and Velocity
2.2
Impingement Density
A patternator as depicted in Figure 3 was used to measure the water impingement density. The water drops of a spray were collected by means of collecting tubes, which were arranged in the spray jet and exhibited a diameter of DR=10mm, over a period ∆t. The amount of water mW collected can be used in the equation m& S =
4 ⋅ mw ∆t ⋅ π ⋅ D R2
(1)
to compute the water impingement density. The water impingement density can be also measured by means of the PDA, but due to the high error rate is necessary to employ another measuring system, i.e. a patternator as in this case.
103 Figure 4 shows the distribution of water impingement density obtained with the flat-spray nozzle as a function of the measuring position. As can be seen, the mean impingement density amounting to 4.5kg/m2/s is fairly constant under the operating conditions described above, forming individual skeins at the nozzle. The results of two measurements are shown to document the reproducibility of the measuring results obtained with this measuring procedure. Water spray
Patternator
Collective containment
Figure 3: Patternator System Water impingement density [kg/m²/s]
5,5 5,3 5,1 4,9 4,7 4,5 4,3 4,1
Nozzle: D25381-13-105/20-1 Measuring planes: 200 mm Nozzle pressure: 5bar Flow rate: 318kg/h
3,9 3,7 3,5 -50
-30
-10
10
30
50
Measuring position [mm]
Figure 4: Impingement Density
2.3
Heat Transfer
The measuring procedure presented in Figure 5 was developed to determine the heat transfer within a fairly short period of time and with locally high resolution /4/. It is based on determining the surface temperature by means of infrared thermal imaging. To determine heat transfer, a thin metal sheet was arranged in front of the spray-generating nozzles and supplied with a constant electric current. The water leads away the heat from the hot metal sheet surface. In a stationary measuring process the metal sheet temperature assumed a value as a function of the local heat transfer coefficient. The higher this coefficient, the lower was the local metal sheet temperature. In a non-stationary measuring process the metal sheet was
104 heated to an initial temperature without being cooled by water spray. Subsequently, the spray jet was released cooling down the sheet. Due to the low thickness of the metal sheet of 0.1mm to 0.3mm both measuring procedures yielded an almost identical temperature distribution on both the side sprayed on and the side not sprayed on. The local distribution, and time distribution in the non-stationary case, of the surface temperature of the non-approached side were recorded by means of an infrared camera. On that relevant side the sheet exhibited a specific coating with an emission capability that had been determined before as a function of temperature. By using a telephoto lens with a supplementary lens a local temperature resolution of up to 0.2mm/pixel can be achieved. The local distribution of the heat transfer coefficient can be calculated from the temperature distribution. The difference between the surface temperature distribution measured on the rear side and the required surface temperature distribution can be determined through a numerical solution of the problem of thermal conduction. Thereby the multidimensional conduction of heat within the sheet was taken into account. PDA-Analysis unit
PDAReceiver
PDAData
Metal sheet 0.1 - 0.3 mm
IR Analysis unit
Registration of operation conditions
Nozzle
150 l/h 5bar;22°C
IRpicture
IR Radiation Transmitter Pump
IR Camera 0.2 mm/Pixel
300 A
Water
Direct current source
Laser (4 Watt)
Figure 5: IR System
2.3.1 Stationary Measuring Procedure In the stationary measuring case surface temperature distribution assumed a constant value under cooling conditions which was recorded by means of an infrared camera. This surface temperature distribution can be used to compute the distribution of the heat transfer coefficient. The heat flux resulting from the current flow through the metal sheet and, hence, from the source of heat applied, is calculated by means of the electric power Pel supplied and the area A of the metal sheet as follows q& H (ϑ H ) =
Pel I 2 ⋅ R ρ el = = A b⋅l s
I2 ( ϑ ) ⋅ H b2
,
(2)
where ϑH is the corrected metal sheet temperature, I the electric current passed through the metal sheet, R the electrical resistance of the metal sheet, (ρel/s) the temperature-dependent specific resistance of the metal with reference to the sheet thickness s, and b the width of the sheet. The specific resistance of the metal sheet was established in special measurements as a function of temperature. During the measurement the sheet was not only cooled by the spray jet. Also radiation of energy and a convective heat transfer must be considered under the conditions of high
105 temperatures. This heat transfer is included as the heat loss. Hence, the heat flux q& Sp (ϑH) leaded away by the spray yet is calculated by q& Sp (ϑ H ) = q& H (ϑ H ) − q&V (ϑ H ) , (3) where q&V (ϑH) is the temperature-dependent heat loss. Hence, the obtained heat transfer coefficient αSp can be calculated using the corrected surface temperature ϑH and the spray jet temperature ϑSp using the following equation α Sp (x, y ) =
q& Sp (ϑ H (x, y ))
(ϑ H (x, y ) − ϑSp ) .
(4)
2.3.2 Non-stationary Measuring Procedure Under the conditions of high heat transfer coefficients and high surface temperatures it is difficult to obtain and keep a stationary operating point and in some cases it is even impossible due to the limits of the electric power available. Hence, heat transfer coefficients were determined by means of non-stationary techniques under the conditions of high heat flux. To this end, the metal sheet was heated to an initial temperature supplying a constant current and, subsequently, cooled down using a spray jet. The time dependent distribution of temperature on the metal sheet surface was measured. For calculating the total heat transfer coefficients α with neglection of conduction the differential equation ρ M ⋅ V ⋅ cM ⋅
dϑ H − Pel = α ⋅ A ⋅ ϑ Sp − ϑ H dt
(
)
(5)
can be established through an energy balance at the metal sheet. Here, ρM is the density of the metal sheet, V its volume, and cM its specific thermal capacity. Dividing by the metal sheet area and converting the ratio Pel/A as described above, we obtain ρ M ⋅ s ⋅ cM ⋅
dϑ H I 2 − 2 dt b
ρ ⋅ el s
(
= α ⋅ ϑ Sp − ϑ H
).
(6)
With the well known time dependent surface temperature it is possible to compute the total heat transfer coefficient. This coefficient is corrected by heat losses. The IR-Camera operates in line-scan-modus with a data rate of 2500Hz. In this case a high resolution in time of surface temperature is obtained.
3
Results of Measurements
In preliminary investigations the nozzles were characterised. To this end, various operating points of different nozzles were used to determine the drop size and drop velocity distributions as well as the water impingement density. In the measurements nozzles were examined which were different in only one of the measured variables indicated. Thus, nozzles were examined which exhibited different drop velocities under the conditions of an identical water impingement density and drop diameter distribution. Establishing such operating points of nozzles, it is possible to examine the influence of the drop diameter, drop velocity, water impingement density and surface temperature on spray water quenching.
106 3.1
Influence of Drop Size
Figure 6 depicts the heat transfer coefficient as a function of the surface temperature, the drop size serving as a parameter. Various nozzles were examined which were different in drop diameter while producing identical water impingement densities and identical drop velocities. The heat transfer coefficient was measured by means of a stationary procedure. This procedure could be employed as the nozzles exhibited water impingement densities which, amounting to about 0.25kg/m2/s, were fairly small and suitable for the measuring procedure. As can be seen, in the investigated range the drop diameter exerts no influence on the heat transfer obtained. The same dependence can be established when analysing the influence of the drop diameter for other drop velocities. Heat transfer coeffizient α [W/m²/K]
350 Nozzle 1: D30 = 107µm, w=3,7m/s Nozzle 2: D30 = 46µm, w=3,7m/s Nozzle 3: D30 = 71µm, w=3,7m/s
300 250
m& S = 0,25
kg m2 ⋅ s
200 150 100 50 0 300
350
400
450
500
550
600
Surface temperature ϑ [°C]
Figure 6: Influence of Drop Size
3.2
Influence of Drop Velocity
Figure 7 shows the heat transfer coefficient as a function of surface temperature, the drop velocity serving as a parameter. Two nozzles were examined which were different in terms of velocity of the generated drops, while producing an identical water impingement density of about 0.33kg/m2/s and an identical mean drop diameter of 66µm and 63µm with reference to the volume. As can be seen, the spray with the higher mean drop velocity yielded a higher heat transfer coefficient. The heat transfer coefficients were determined in a stationary measuring procedure. Heat transfer coefficient α [W/m²/K]
350 300
Nozzle 4: D30=63µm, w = 6,7m/s Nozzle 5: D30=66µm, w = 3,8m/s
250 200 150
m& S = 0,33
100
kg m2 ⋅ s
50 0 300
350
400
450
500
Surface temperature ϑ [°C]
Figure 7: Influence of Drop Velocity
550
600
107 3.3
Influence of Water Impingement Density
Figure 8 presents the heat transfer coefficient as a function of the water impingement density. Results of Müller/Jeschar /5/ and Fujimoto /6/ are also presented. The own examinations were performed at a surface temperature of 550 °C, a drop velocity of 8m/s and a drop diameter of 60µm. It can be seen that the water impingement density exerts a major influence on the heat transfer coefficients achieved. When the water impingement density increases, the obtained heat transfer coefficient also increases. The own examinations results higher heat transfer coefficients for constant impingement density. In his investigations Fujimoto detected the same gradient of heat transfer coefficient with increasing impingement density. The curve of Müller/Jeschar has a lower gradient. Heat transfer coeffizient [W/m²/K]
450 Experiment Fijimoto /6/ Müller / Jeschar /5/
400 350 300 250 200 150
ϑ = 550°C w = 8m/s D30 = 60µm
100 50 0 0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1
Water impingement density [kg/m²/s]
Figure 8: Influence of Impingement Density
3.4
Influence of Surface Temperature
Figure 9 contains the heat transfer coefficient as a function of the water impingement density. The surface temperature serves as a parameter. Operating points with constant drop velocities and constant water impingement densities were examined at surface temperatures of 350 °C, 450 °C and 550 °C. It can be seen that the heat transfer coefficient achieved slightly decreases when the surface temperature increases and the drop velocity and water impingement density remain constant. Heat transfer coefficient [W/m²/K]
600 w = 8m/s
500 400 300
350°C 450°C 550°C
200 100 0 0
0,2
0,4
0,6
Water impingement density [kg/m²/s]
Figure 9: Influence of Surface Temperature
0,8
1
108
4
Conclusions
For performing heat transfer measurements with a locally high resolution an infrared measuring device is available measuring the temperature distribution on an electrically heated metal sheet. The metal sheet was cooled from the opposite side by means of water spray. The water spray was analysed in terms of drop size, drop velocity and water impingement density using a PDA. Examinations revealed that in the investigated range the drop size has no influence on the heat transfer coefficient, whereas the surface temperature exerts a low, drop velocity a bigger and water impingement density the biggest influence. The investigated range of low impingement density has to be enlarged to high impingement density used in continuous casing. With the known local spray characteristics then the local heat transfer coefficients are computable. The already investigated flat-jet nozzle used for cooling purposes in the continuous casting process exhibited a fairly constant water impingement density. The drop velocity reduces towards the border.
5
References
[1] Jacobi, Kaestler, Wünnenberg; Heat transfer in cyclic secondary cooling during solidification of steel, Ironmaking and Steelmaking, 11 (1984), S. 132-145 [2] Köhler; Wärmeübertragung von heissen Oberflächen durch Wasserfilmkühlung im Bereich der stabilen Filmverdampfung, Dissertation, TU-Clausthal, 1990 [3] Mizikar; Sprya Cooling Investigation for Continous Casting of Billets and Blooms; Iron and Steel Engineer (1970), S. 53-60 [4] Boye, Schmidt; Einfluss von Oberflächentemperatur und Tropfenparameter auf den Wärmeübergang bei der Sprühkühlung; Chem.-Ing.-Techn. 70, (1998), S 1177-1178 [5] Müller, Jeschar; Untersuchung des Wärmeübergangs an einer simulierten Sekundärkühlzone beim Stranggießverfahren, Archiv Eisenhüttenwesen, 44 (1973), S.589-594 [6] Fujimoto, Hatta, Asakawa, Hasimoto; Predictable modelling of heat transfer coefficient between spraying water and a hot surface above the Leidenfrost temperature; ISIJ International 37 (5), 1997, S. 492-497
Grain Structure, Microstructure and Texture of Copper Ingots Produced during the Continuous Casting Process V. Plochikhine, V. Karkhin, H.W. Bergmann Department of Metallic Materials, University Bayreuth, Germany
1
Introduction
This paper presents the results of an investigation on the solidification structures as well as the texture of copper ingots produced by the continuous casting process. The aim of this study was to develop a relatively simple and effective simulation tool for prediction of the grain structure and the texture of copper ingots and to use this tool to find out the most effective strategy for optimising of the solidification structure quality. Here we present the main results of the investigation, and further detailed information can be found in [1].
2
Experimental and Simulation Methods
A copper ingot produced by a conventional continuous casting process has been experimentally investigated. The correspondent casting parameters are listed in Table 1. Table 1: Parameters of the continuous casting process Parameter Value Diameter, mm 298 Velocity, mm/min 160 Initial casting temperature, °C 1140 Average mould temperature (evaluated value), °C 650 Mould length, mm 350 Standard metallographical techniques were used to investigate the grain structure and the microscopic dendritic structure of single grains. Furthermore, the texture in different zones of the ingot from the outer surface to the central part was measured using X-ray diffraction analysis. The “grain boundary evolution” modelling method has been used in order to reproduce the process of the grain structure formation and the texture evolution in an ingot during solidification. This method has already been described in [2-4]. The temperature field has been calculated using an analytical method of Green functions [1].
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
110
3
Results and Discussion
The results of the simulation directly represent the grain structure. All possible crystallographic orientations of grains (from 0° to 90°) are evenly distributed into nine classes with the colour of the grain reflecting the orientation class of the grain. Figure 1 shows the grain structure calculated under the parameter listed in Table 1 (Figure 1, a). A section of the calculated structure is compared with a photograph of the longitudinal section of the experimentally investigated ingot. The calculated and the experimentally observed grain structures exhibit zones with different alignment of the grains, which occur during solidification due to the process of grain selection.
Figure 1: Grain structure of the copper ingot: a) calculated structure, b) a section of the calculated structure in comparison with the longitudinal section of the experimental ingot
Calculation of the fraction of grains belonging to each orientation class in some section of the ingot gives a histogram, which can be used as a qualitative characteristic of the texture distribution in the correspondent section. Figure 2 illustrates the texture development in an ingot, which is also induced by the grain selection process during the solidification. The experimental measurements of texture (Figure 2, c) in different regions of an ingot exhibit the presence of grains with a preferred orientation, which is enhanced in the direction of the ingot centreline.
111
Figure 2: Microstructure of the copper ingot: relatively coarse microstructure in a zone near the outer surface of the ingot and a fine microstructure in a zone neat the centreline
The grains observed in the zones with the different grain alignment exhibit a different dendritic microstructure. The microstructure in the zone near the centreline is much finer compared to the area near the outer surface of the ingot. This difference occurs due to the sufficiently different growth rates of the grains in these zones during solidification. The growth rate of the dendritic grains R is described using the following equation [5]: R = v cos θ / cos ψ
(1)
v is the casting velocity, θ is the deviation of the normal to the solidification front from the centreline, ψ is the angle between the normal to the solidification front and the crystallographic orientation of the grain. In the zone near the centreline the growth rate is almost equal to the casting velocity, producing a fine microstructure. In the remaining regions of the ingot the microstructure is more coarse, due to the higher value of the angle θ. The central question of the study was to find out the parameters of the continuous casting process, which have a significant influence on the quality of the solidification structure. As the quality criteria the high homogeneity of the grain size and of the texture as well as the fine microstructure have been assumed. It can be simply shown, that the quality of the solidification structure directly depend on the rate of cooling of the liquid metal. A more rapid cooling leads to a smaller liquid bath with a less elongated solidification front, whereby the value of the angle θ decreases, resulting in a higher growth rate and, accordingly, finer microstructure. The change of the form of the solidification front has also a significant positive effect on the grain structure (see below). Numerous simulations have shown, that the most effective means for promoting rapid cooling are to decrease the length of the mould and/or stirring of the liquid metal. Results of simulations, shown in Figure 4, clearly illustrate the changes in the solidification structure, which are expected if the mould length is decreased (Figure 4, b) and, furthermore, if stirring of the liquid metal is applied (Figure 4, c).
112
Figure 3: Texture of the copper ingot: a) calculated grain structure, b) calculated histograms, showing the fraction of a certain orientation class in different sections of the ingot and c) experimentally obtained polefigures
Both measures lead to the formation of long straight grains, which are aligned at an angle approximately 45° to the centreline. About 80% of grains have the crystallographic orientation in the interval 45°±15° relative to the centreline, what indicates a high texture homogeneity. Furthermore, due to the average orientation of solidification front 45° relative to the centreline (Figure 4, a) the growth rate will be about 4 times higher as by the normal process (Figure 4, b). This means, that the two times finer microstructure can be expected [5].
113
Figure 4: Influence of the shorter mould length and of the forced convection on the formation of the solidification structure and its expected quality
4
Summary
The experimentally observed grain structure exhibits zones with different alignment of grains, texture and dendritic microstructure. The results of the simulation exhibit the same features of the solidification structure. Results of the numerical simulation show, that sufficient improvement of the quality of the solidification structure can be achieved by more rapid cooling of the liquid metal. The most effective strategy that can be used to promote the more rapid cooling is to alter the mould construction (decrease of the mould length) combined with the forced stirring of the liquid metal.
114
5
Acknowlegements
This study is carried out in the cooperation with the Norddeutsche Affinerie AG. The authors would like to thank Dr. A. Mathiae for the useful consultations concerned the experimental part. We are also indebted to Mrs. Jacqueline Uhm for her patient help in the preparation of the manuscript. This work was partially supported by the foundation “Stiftverband Metalle”.
6
References
[1] H.W. Bergmann, “Numerische Simulation der Strukturausbildung bei der Erstarrung im Strangguß”, Bericht an Stiftverband Metalle, 1999, pp. 32. [2] V.V. Ploshikhin, H.W. Bergmann: "Simulation of Grain Structures in Laser Beam Welds Undergoing the Planar Solidification Mode", Mathematical Modelling of Weld Phenomena 4 (Ed.: Cerjak J.), The Institute of Materials, 1998, 150-165. [3] V. Plochikhine: "Modellierung der Kornstrukturausbildung beim Laserstrahl-schweißen", Dissertation, Universität Erlangen-Nürnberg, 1998. [4] V.V. Ploshikhin, H.W. Bergmann: "Grain Structures of Laser Beam Welds Undergoing the Planar Solidification Mode", Modelling of Casting, Welding and Advanced Solidification Processes - VIII (Ed.: G. Thomas, C. Beckermann), TMS, 1998, 399-405. [5] M. Rappaz, S.A. David, J.M. Vitek, L.A. Boatner: “Analysis of Solidification Microstructures in Fe-Ni-Cr Single-Crystal Welds”, Metall. Trans. A, 1990, 21A, 17671782. [6] Kurz W. and Fisher D.J. Fundamentals of Solidification, Trans Tech Publications, 1989.
Technologies and Installation for Electrochemical Hardening of Wear Surfaces Radu Boiciuc, Viorel Munteanu, G. Petrache UZINSIDER ENGINEERING S.A. ( ICPPAM) Galati, Romania
1
Introduction
As a result of the investigations performed in specialised laboratories, we started to design plants based on the principle of electrochemical deposition of some hard metal layers (nickel, nickel-alloy). Special electrochemical solutions for nickel and nickel-alloy plating were prepared, having a well-defined composition and parameters kept, in strict limits resulting in the achievement of hard and thick layers (nickel 300 µm approx.and nickel-alloy 1mm approx.). The technology of reconditioning the mould plates by nickel and nickel alloy coating includes operations for the prepration of plate surface (cleaning, pickling), nickel and nickel alloy tanks, special devices for solution stirring, delay of dendrite for occurrence on the plate corners and edges, solution recirculation system, power supply sources, electrical and steam system for solution heating. Following the experiments, the best working parameters were also defined. On the plant designed by our institute we performed the hardening of the mould plate surfaces within the continous casting, obtaining significant savings and an increase of the service life (5 times approx. compared to the olds applied methods). By such coatings, the hardness of plate working surfaces increase from 100 HV, existing after mechanical processing to 500 – 550 HV, after nickel coating, to 800 HV, after nickel alloy coating. For a proper process development, as well as the obtaining of accurate deposition properties, the influence of several extremely important factors on the coating process was taken into account.
2
Influence of Electrochemical Solution Composition
In order to select the optimal nickel coating solution three electrolyte types were tested: • sulphate-based electrolyte obtaining low internal stress depositions but with rather low hardnesses (140- 170 HV); • mixed sulphate-chloride electrolyte with high electric conductivity, high dispersion power, respectively (causing a more uniform distribution of metal on the mould plate surface) and with better hardness of the deposited layer (250- 300 HV); • chloride electrolyte obtaining thicker layers, pinhole-free, with high cathodic performances but with lower hardnesses (200 HV approx.). Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
116 If deposited under certain conditions, nickel increases the metal surface hardness and generally, irrespective of the deposition conditions, nickel hardness is higher than the copper one, which is why the nickel coated plates are much more resistant than the copper ones to the friction wear. The electrochemically deposited nickel is ferro-magnetic. The dull or semi-bright nickel depositions have a columnar structure, they are soft and ductile, and the bright ones have a lamellar structure, with the laminations parallel to the copper surface they deposit on, they are harder, more strained and less ductile. Also, nickel deposits in compact, fine grained, continuous, smooth layer and adherent to the copper plate. The basis of the process for the electrochemical obtaining of nickel layers consists of the following conjugated, cathodic and anodic reactions which develop in acidic medium. For the deposition of a nickel layer on the copper plates a special chloride-sulphate electrolyte was selected where antipitting, brightening (grade I and grade II) and smoothing agents were added. The nickel chloride improves the anodic corrosion (the chlorine ion being an anodic depolarizer), increases the electrolyte conductivity and increases the deposition hardness. The nickel sulphate is the main source of nickel ions and its concentration determines the limits of current densities where good quality depositions are obtained. Generally, the increase of nickel concentration allows the use of higher current densities, implying higher deposition rates. Boric acid is the buffer substance and helps in obtaining light-coloured, smooth and ductile depositions. The antipitting agent (for wetting) improves the coating smoothness and decreases the pitting. This agent acts on the reduction of surface and interface stress of the hydrogen bubbles adhering to the cathode, resulting in their detaching. The release of the hydrogen produced in the secondary cathodic reaction takes place before the bubbles increase to the dimensions that could cause the pitting by coating locking at the cathode contact point. In order to obtain bright depositions the grade I and grade II brightening agents are used. The grade I compounds only develop bright depositions on previously polished surfaces and allow the presence of grade II brightening agents in the electrolyte. The mechanism by which the grade II brightening agents act is based on the molecule adsorbtion on the plate surface by non-saturated links. Adsorbtion can occur on the crystal growing spots (nodes, steps) and at dislocations. The grade II brightening agents are only used together with the grade I ones to obtain bright and smooth depositions that do not feature brittleness. The applied agent is non-saturated (it has the -C≡C- group) and brings carbon in the deposition by the reactions occurring at the cathode. The smoothing agents are additions which, by increasing the cathodic polarization and due to the specific activity of inhibiting crystal growing (on microprominences), cause a preferential distribution of nickel on the plate surface. They are consumed by the cathodic reduction reaction, being differently incorporated in the deposition. The mechanism by which they act is due to the different diffusion rate on the cathodic surface. After the performed experiments, the mixed sulphate-chloride electrolyte (with a higher chloride content than the sulphate one) was selected for the mould plate surface hardening, due to the obtained hardnesses, compared to the other electrolytes.
117 A Ni-P alloy was selected as a final layer for the mould plate surface hardening. The main advantage of this coating is the high increase rate of the deposition and the solution stability in time. Also, these coatings have a considerably lower porosity than the nickel depositions. The depositions containing 2% phosphorus are smooth, dull, with small crystals and look like the nickel ones, those containing 5% phosphorus are semi-bright and the ones with 10% phosphorus are bright. The Ni-P alloy hardness is higher than the nickel one and increases with the increase of the phosphorus content. The Ni-P alloys with 2% P are plastic. The density of these alloys decreases with the increase of the phosphorus content. The Ni-P alloy containing more than 8% P is not magnetic. The electrochemically deposited Ni-P alloys have a 30% lower friction coefficient than chrome. Thus, these alloys are completely suitable for the surface hardening, aiming at increasing the part wear resistance. The Ni-P bright alloy deposition has an advantage over the bright nickel coating in that the electrolyte is more easily controlled, taking into consideration the fact that the electrolyte does not contain organic substances; however, Ni-P alloys are not as light-coloured as the bright nickel. The conditions for the electrochemical deposition of this alloy differ considerably from the conditions the nickel coating requires. The 15% P alloy is deposited from acidic solutions. For the mould plate coating an electrolyte was selected, out of which a 12- 15% P content was obtained in the deposition. The bath operates at high temperature (75- 95°C) because at room temperature the current output is low. At a 75°C temperature the strongest pitting defect is obtained, the best depositions being obtained at a temperature of 95°C. The current output decreases with the temperature increase. The Ni-P alloy has higher internal stresses and at higher thicknesses, it cracks. In order to prevent this phenomenon a small amount of saccharine is introduced in the coating bath. In order to remove the pitting phenomenon low wetting addition quantities can be used. At low current densities (4A/dm2), the mould plate is not completely coated sometimes. The phosphorus content in the deposition decreases with the increase of current density. In order to obtain higher thickness depositions, the electrolyte needs stirring from time to time, besides the mould plate stirring.
3
Influence of Current Density
The higher the current density, the higher the ion migration rate in the cathodic area, favouring the formation of fine structure deposits. This can be explained by the increase of the cathode active surface due to the number increase of the crystals on it. The grain sizes increase with the current density, reach a maximum, after which they decrease, resulting in an increasingly finer structure of the deposit. In the case of nickel coating the increase of the current density does not result in the exaggerated dendrite increase, but it can lead to brittle depositions, end-burnt depositions, rough depositions, pitting. The optimal current density was within 2- 4 A/dm2, the defects above mentioned occurring after almost 24 hour operation at higher current densities.
118 For the Ni-P alloy deposition one cannot use current densities below 5 A/dm2, otherwise rough and brittle depositions would be obtained. Therefore, the current density used was 5-10 A/dm2.
4
Temperature Influence
The temperature increase, keeping the other conditions constant, usually reduces cathodic polarization, contributing to the formation of macrocrystalline deposits. This temperature influence can be explained, first of all, by the increase of the ion diffusion rate (the concentration polarization decreases) and, secondly, by the fact that chemical polarization decreases with the temperature increase. Thus, the permissible current density and, therefore, the process rate can be increased. The current density increase promotes the deposit crystal reduction, thus removing the temperature effect on the structure. In the case of nickel coating a low temperature, associated with insufficient stirring, can result in rough depositions by pitting. The optimal temperature range for nickel coating was selected within 35- 45°C. In the case of Ni-P alloy deposition the optimal temperature used for the obtaining of hard depositions was 75- 80°C (the current output being very low at lower temperatures).
5
Influence of the Mechanical Stirring of the Mould Plates (Cathode)
The main role of stirring consists in the promotion of the ionic concentration homogenizing in the cathodic film and in the rest of the solution. Plate stirring along the process baths was used to increase process output. It keeps the solution concentration constant, preventing the concentration polarization. Due to this fact, compact, smooth, fine grained deposits can be obtained at higher current densities and with a high current output; the lower the current density, the higher the cathode stirring should be. The mould plates were longitudinally stirred in relation to the coating pot, with a frequency of 15- 20 complete strokes/min, obtaining fine, smooth and compact depositions.
6
Influence of Electrolyte Filtering
The mechanical contaminants cause high difficulties when nickel coating, being deposited with the metal, resulting in the occurrance of porosities and deposition roughness. These contaminants come from: • anodes - non-soluble components (oxides, carbides); • metal contaminants forming non-soluble compounds in the electrolyte; • non-soluble compounds formed in the electrolyte; • dust from the environment. In the nickel coating electrolyte the maximum grain size is 26 µm, the concentration of grains smaller than 10µm is 15%, and that of grains smaller than 1µm is 2%. Under such
119 conditions an electrolyte recirculation system was developed, consisting of one pump and two filters.
7
Influence of Current Direction Periodical Change
It was found that the periodical change of the current direction in a 6:1 ratio when nickel coating the mould plates results in the obtaining of high thickness, smooth, prominence-free and internal stress-free depositions, the nickel depositions being more compact, smoother, less porous and having a lighter colour. Also, the capacity of dispersion increases, especially in the low current density zone. The favourable influence of this factor on the structure is explained by two aspects: • the possibility of increasing the current density, resulting in microcrystalline deposits; • the growth arresting of crystals formed in the cathodic stage. Besides these factors taken into account during the electrodeposition process, solutions had to be found for the uniform distribution of current on the plate surface, the smoothening of nickel or Ni-P alloy layer thickness on the entire deposition surface, respectively. Thus, two main solutions were developed: • the optimal anode-cathode distance (20- 25 cm) was determined, a higher distance resulting in the stress increase due to the increase of ohmic strength; • additional (protective) copper bar cathodes were used, surrounding the plates as a frame, the dendrites moving on it.
8 • • •
Conclusions The structure of the electrolytically deposited metals is determined by a series of factors such as: the base and deposited metal type, electrolyte type, current density, temperature. The correct control of the electrolytical process allows to obtain higher hardness metal layers (nickel, Ni-P alloy) than the basic metal out of which the plates are made, improving their operation behaviour. Within the metal coating laboratory of UZINSIDER ENGINEERING S.A. a technology was developed and laboratory and pilot plants were constructed.
Modelling – Heat and Fluid Flow; Nucleation
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
Numerical Mass and Heat Flow Predictions in Aluminum DC Casting: A Comparison of Simulations with Melt Pool Measurements Andreas Buchholz 1, Benoît Commet2, Gerd-Ulrich Grün3, Dag Mortensen4 1
Corus Research, Development & Technology, P.O. Box 10000, NL-1970 CA Ijmuiden, The Netherlands Pechiney Centre de Recherche Voreppe, BP 27, F-38340 Voreppe, France 3 VAW aluminium AG, Research and Development, P.O. Box 2468, D-53014 Bonn, Germany 4 Institute for Energy Technology,P.O. Box 40, N-2027 Kjeller, Norway 2
1
Abstract
Melt flow is one of the key factors for process and quality control during DC casting of aluminum ingots. Nowadays, numerical modeling is an important tool to support the design of optimized melt distribution systems and to describe the solidification process. Nevertheless, the predictions of such calculations are rarely tested, since temperature and especially flow measurements in the liquid pool are difficult to do. This paper presents a comparison of threedimensional steady state calculations of mass and heat transport with temperature and sump profile measurements during the DC casting of rolling ingots of an AA3004 and AA5182 alloy. The calculations are made with different CFD codes, which have been cross-checked in a previous benchmark. The temperatures are measured in specific planes of the melt pool using an array of thermocouples. In addition, the shape of the solidification front is scanned with a mechanical device. Simulated thermal fields and measured temperatures are compared for two completely different flow situations: In the first case the usual casting technique using a spout and distributor bag is applied, while in the second case the distributor bag is removed. Since certain features of the temperature field pattern are closely related to specific flow phenomena the comparison provides a means to assess the validity of the flow calculations. In both cases the results agree well with the related measurements.
2
Introduction
Recent research clearly shows the importance of an appropriate design of the metal distribution system with regard to temperatures in the liquid part of the solidifying ingot and related microstructures or casting defects [1-3]. Because a direct determination of the complex flow pattern within the liquid aluminum under real casting conditions is extremely difficult and rather unreliable, mathematical modeling has become the standard tool for the optimization of metal distribution systems. A variety of different models, either based on general purpose CFD codes or specially developed programs is used for the solution of this kind of problem and several application examples concentrating on different aspects of the DC casting process have been published [2-4].
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
124 Currently used models within the European aluminum industry were subject of an intensive evaluation of the influence of different flow models on the complex interaction between flow pattern and thermal development. This work was carried out within the context of a joint European research project on the development of modeling tools on aluminum casting technology (EMPACT), partly funded by the European Commission. The comparison of the results of all three-dimensional calculations revealed a strong influence of the chosen flow model on the high temperature regime above the melting point [5]. Based on these findings the work in that research project was continued with the validation of these models against temperature measurements during experimental DC castings. This paper partly summarizes the DC casting trial results and the comparison with the related simulations. In the following section a brief description of the casting trials and the used measurement technique is given, while in the subsequent part the influence of different alloys and distribution situations as well on the real thermal field in the liquid aluminum as on the modeling results is critically discussed.
3
DC Casting Experiments and Measurements
During the previously mentioned EMPACT project, several experimental casting trials were carried out at different locations throughout Europe. In order to get comparable data for all different alloys all rolling ingots were cast with the same casting tools (mould + bottom block) for ingot sizes of 1.85m x 0.51m. The used mould is a conventional waterjacket system with continuous oil lubrication. The DC casting trials related to this paper took place at the Research Center of Pechiney (CRV) in Voreppe (France). Details related to the casting parameters are given in the following sections with the comparison of measurements and modeling results. TC
Z X
Y nozzle
mould
Combo-bag
melt pool
sensor
rolling ingot
Figure 1: Principle of Pechiney CRV’s device to measure temperatures at various locations in the liquid sump and the sump depth by mechanical contact
The experimental device used for the temperature mapping in the sump is a 3-axis crane as shown in Figure 1. The sensor can be accurately translated along the three main axes of a
125 rolling ingot DC casting mould. The translation direction X is parallel to the width, Y to the thickness and Z to the casting direction. The location {X, Y, Z} of the sensor is recorded with time. Each location is maintained at least 10 s in order to get a stable measurement. In some regions, the temperature varies with time due to turbulence; there a mean temperature is calculated by the measuring device. Two types of sensors are used for the measurements: • A series of thermocouples fixed on a metallic rod. The acquisition of several temperature data at the same time allows the rapid mapping of the whole sump space. Different rod shapes enable the access of remote locations, for example the region below the Combobag. • A single straight rod. It is used to touch the bottom of the sump, which corresponds to an isothermal surface at the coherency temperature of the aluminum alloy. Every measurement is then stored as location with a related temperature: {x, y, z, T}. In order to achieve a continuous description of the temperature field pattern those data are interpolated by use of the mathematical package Matlab. For the current model validation data in two planes are calculated: • In a horizontal plane {X,Y}, Z=0.1m below the top of the mold. • In a vertical plane {X,Z}, Y=0.0m, which corresponds to the symmetry plane parallel to the rolling face of the ingot.
4
AA3004 Alloy DC Casting
The simulation of the stationary period of the AA3004 alloy casting is done with ALSIM [4], which is a FEM code especially designed for the calculation of DC-casting processes. The coupled heat and mass transport equations are solved on the basis of a continuum-mixture model of the solid-liquid metal, and a Darcy force accounts for the interfacial friction due to the different velocities of the solid and the liquid. Turbulence is modeled with a low Reynolds number (LRN) k-ε model. Due to assumed symmetry only a quarter of the cold geometry (0.51 m x 1.86 m) of ingot and mould is used as calculation domain. Corresponding to the casting trials the geometry of a standard Combo-bag with a dimension of 0.3m x 0.1m x 0.1m is included as melt distributor. The bottom of that bag is 0.06m below metal level and 0.02m below the nozzle tip. The openings against the small faces of the slab are 0.1m wide and 0.03m high. 0.105m from the center of the symmetrical bag geometry on each side an additional 0.05m x 0.05m hole is cut into the bottom of the Combo-bag. Following the average casting temperature of the trial the inlet temperature is chosen with 680°C and a casting speed of 0.001m/s (corresponds to 60mm/min) is applied. The metal level is 0.06m below the top of the mould. Thermal data for the AA3004 alloy are calculated by Alstruc [6], and boundary conditions are similar to those applied in [4]. 4.1
AA3004 DC Casting with Combo Bag
The values of the interpolated measured and the calculated temperatures in a vertical plane parallel to the rolling side (about 0.02 to 0.03m off the center of the slab) are shown in Figure 2. Vertical zero level corresponds to the top of the mould.
126 0
0
662.5
665
670
660 −0.1
−0.1
−0.2
−0.2 655 −0.3
660
650
−0.4
−0.4
652.5
−0.5
−0.5 −0.6
0
655
−0.3
652.5
−0.6
measured sump 0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
650
measured sump 0
0.9
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Figure 2: Measured (left) and calculated (right) temperatures in a vertical plane. Equidistant isotherms every 2.5 °C. AA 3004 alloy DC casting trial with Combo-bag. The thick line indicates the sump determined by the mechanical scanning device. The device is sensitive for a fraction solid of about 50%
Depending on the measurement array the related contour plot covers only a part of the calculated results. Main differences between both occur in an area below the distributor where the measurements show no thermal disturbance, while this is predicted by the model due to flow out of the bottom hole of the Combo-bag. A possible reason may be that the real Combobag is made out of woven fibres and still deformable, while it is assumed to be rigid in the model. Therefore, in reality this may cause a slightly disproportion in the shape of the bag, since the forces due to the fast flow out of the nozzle tends to bend down the bottom of the bag right below. Consequently, the following lateral flow is directed slightly upwards when it passes the bottom hole of the bag and outflow through the bottom hole is not possible. The large and nearly isothermal sump below the bag is, however, evident in both, measurements and calculations. Buoyancy driven convection is the explanation for this large volume of isothermal liquid and (possibly) dilute mush. The liquidus temperature calculated by Alstruc for this alloy (652.4°C) is probably too high, since the isothermal value in the calculation is a few degrees higher than the corresponding measured temperature. The results of the same cast (c.f. Figure 3) in a horizontal plane 0.040m below the metal level support the explanation concerning the outflow through the bottom holes. 665
0.2
662.5
667.5
0.2
665
657.5 0.1
662.5
657.5
660
0.1
665 662.5
660 0 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Figure 3: Measured (left) and calculated (right) temperatures of the AA3004 alloy casting in a horizontal plane, 0.1m below the top of the mould. Equidistant isotherms every 2.5°C
The measured temperatures reveal that the hot jet out of the distributor seems to turn upwards (out of the plane) towards the melt surface, where it is somehow reflected. This is explained by the existence of the hot spot on the outer right area in the measurements, while the calculations show a steady decrease from the distributor to the cooling boundaries, which corresponds to a straight outflow of the melt.
127 0
0
−0.1
−0.1
−0.2
−0.2
660
−0.3 −0.4
655
−0.3 657.5 655
−0.5
−0.4 652.5
650
652.5
647.5
−0.5
−0.6
−0.6
−0.7
−0.7
measured sump
−0.8
−0.8
−0.9
−0.9
−1 0
660
0.2
0.4
0.6
0.8
−1 0
measured sump
0.2
0.4
0.6
0.8
Figure 4: Measured (left) and calculated (right) temperatures in a vertical plane. Equidistant isotherms every 2.5°C. AA3004 alloy DC casting trial without Combo-bag. The thick line indicates the sump determined by the mechanical scanning device. The device is sensitive for a fraction solid of about 50%
4.2
AA3004 DC Casting without Combo-bag
In this case the measurements (left part of Figure 4) do not show how deep the jet out of the nozzle penetrates into the mushy zone and the solid due to spacial limitations of the TC array. But it is of particular interest that the whole sump is a dilute mush or the aluminum has a temperature close to the liquidus (except from the hot jet in the center). Concerning a later comparison with the calculation results it should be kept in mind that the temperature representation near the hot center "surface" (it is not the surface, but 0.04m below) in the measurements depends on the interpolation between only one TC result directly below the nozzle (680°C in average) and a colder value (650°C) 0.22m besides the center. In the calculations, which deliver the temperatures on the right of Figure 4, the Darcy term in the momentum equations is not activated before the temperature drops below 649.4°C (corresponding to a selected solid fraction of 30%). Between the liquidus temperature and this value the viscosity is increased linearly by a factor 20 compared to the molecular viscosity of 1.35 10-3 Pa.s. This is done for consideration of, ad hoc, the effect of solid grains transported with the flow. The selection of this "Darcy temperature" is decisive for the penetration depth of the jet into the mush. Here, this seems to be overestimated and further adjustments necessary.
5
AA5182 DC Casting Results
As described in the section on the experiments the AA5182 alloy was cast with the same mold and distributor configuration as that used for the AA3004 alloy trials. But regarding this case the numerical modeling is carried out on the basis of the general purpose CFD control volume code CFX 4.3. This model incorporates a source based latent heat release and a Darcy approach to couple momentum and thermal equations [1,5]. A standard k-ε model accounts for turbulence, while buoyancy driven convection is approximated by a Boussinesq approach. Special damping terms guarantee, that turbulence quantities and buoyancy effects vanish in
128 the solid. Material properties of the AA5182 alloy are as well supplied by Alstruc calculations [6], which predict a liquidus temperature of 638.3°C and a solidus of 448°C for the chosen composition. The secondary cooling is described by temperature dependent heat transfer coefficients determined by a special experimental setup and inverse modeling [7]. Similar to the AA3004 alloy two cases are considered: Firstly, an ingot is cast using a distributor bag, and in the second trial this bag has been removed. 5.1
AA5182 DC Casting with Combo-bag
The measured and calculated temperatures in the vertical plane of the first experimental set-up are shown in Figure 5. The general pattern is very similar to what has been found earlier for this kind of melt supply [5]. The distributor bag generates a mainly horizontal outflow. Due to buoyancy the steadily supported hot melt forms a layer near the surface of the liquid pool. The aluminum then mainly descends at the small face of the rolling ingot and along the cold solidification front. Corresponding to the previous AA3004 alloy trial (c.f. Figure 2) a significant discrepancy between the computation and the measured temperatures is the hot jet coming out of the bottom hole in the simulation. 0
0 660
650
645
−0.1 −0.2
−0.2
640
645
640
−0.4
−0.4
−0.5
−0.5
Tl=638.3
635
−0.6
−0.6
measured sump
measured sump 0
0.1
0.2
0.3
650
−0.3
635
−0.3
665
675
−0.1
0.4
0.5
0.6
0.7
0.8
0
0.9
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Figure 5: Measured (left) and calculated (right) temperatures in a vertical plane. Equidistant isotherms every 2.5°C. AA5182 alloy DC casting trial with Combo-bag. The thick line indicates the sump determined by a mechanical scanning device. The device is sensitive for a fraction solid of about 50% 0.3
0.3
650
655
665
660
0.2
0.2
650
0.1 0 0
670
0.1
0.2
0.3
0.4
655
0.1
645
0.5
0.6
0.7
0.8
0.9
0 0
650
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Figure 6: Measured (left) and calculated (right) temperatures of the AA5182 alloy casting in a horizontal plane, 0.1m below the top of the mould. Equidistant isotherms every 2.5°C
But here the beam emerges beside the front face of the distributor as indicated in the horizontal temperature plot on the right of Figure 6, while the temperature measurements predict no significant flow in this area (left of Figure 6). Concerning this deviation the explanation regarding the deformability of the Combo-bag given in the previous section is still valid. On the other hand, it is interesting to note that the calculations for the AA3004 alloy predict a different direction of the flow out of the bottom
129 hole, although the liquid alloy properties should have no significant impact in this region. This different behavior may be caused by the applied turbulence models. The ALSIM program uses a low Reynolds number k-ε model in the AA3004 computations, whereas the CFX4.3 model includes a standard high Reynolds number model. Again, the interesting correspondence between measurements and computation is the broad, almost isothermal region in the lower part of the sump, i.e. the area between the bottom of the sump and the 635°C isotherm in the measured data and between the 635°C and 640°C in the calculations (c.f. Figure 5). This is partially due to buoyancy driven convection, but mainly because of the formation of fraction solid. The fraction solid – temperature dependency of the AA5182 alloy shows a very steep slope near the liquidus temperature. This means that small changes in temperature require a comparably large extraction of heat. Thus, the formation of solid close to the liquidus temperature works as thermal buffer leading to this large area with low thermal gradients. Although, the actual liquidus temperature of the alloy was not measured during the casting trial, following this consideration it can be concluded that the real value is a few degrees lower than in the computations. Actually, the composition used for the Alstruc calculations differs slightly from that of the cast ingot and can account for this deviation. 5.2
AA5182 DC Casting without Combo-bag
On the basis of the above considerations the results of the measurements and computations for the casting without distributor bag can be understood. Both, measurement and computation in Figure 7 indicate as in the related AA3004 alloy case that a vast region of the sump besides the inlet jet is filled with almost isothermal liquid. Again, it can be concluded that the initial formation of fraction solid works as a thermal buffer. 0
0 660
−0.1
−0.1
−0.2
−0.2
−0.3 −0.4
645
−0.4
635
−0.5
−0.5
−0.6
−0.6
−0.7
−0.8
−0.9
−0.9 0.2
0.4
0.6
645 635
−0.7
measured sump
−0.8
−1 0
640
−0.3
633
640
Tl=638.3
0.8
−1 0
measured sump 650
0.2
0.4
0.6
0.8
Figure 7. Measured (left) and calculated (right) temperatures in a vertical plane. Equidistant isotherms every 2.5°C. AA5182 alloy DC casting trial without Combo-bag. The thick line indicates the sump determined by the mechanical scanning device. The device is sensitive for a fraction solid of about 50%
Compared to the usual casting with a distributor bag, it seems that without distributor heat is stronger dissipated. A reason for this could be the higher turbulence generation in the configuration without bag. The immersed inlet jet somehow reverses the natural thermal
130 gradient. Buoyancy additionally generates instability in the flow leading to increased turbulence, which yields more mixing and higher effective diffusive transport.
6
Conclusions
Based on this extensive work the following conclusions can be drawn: • The accurate modeling of the fluid flow in the vicinity of the distributor still has to be improved and requires additional investigation. • The model results have to be interpreted with special care where a strong interaction occurs between fluid flow and solidification front. • The encouraging overall agreement of fluid flow and heat transport between the model predictions and the measurements proves that the numerical models are a valuable tool to improve the casting process.
7
Acknowledgements
This research was carried out as part of the Brite-Euram project BE-1112 EMPACT (Contract N° BRPR-CT95-0112). It included the partners: Alusuisse-Lonza Services Ltd., Switzerland, Calcom SA, Switzerland, Delft University of Technology, Netherlands, École Polytechnique Fédérale de Lausanne, Switzerland, Elkem Aluminum ANS, Norway, Hoogovens R&D, Netherlands, Hydro Aluminum AS, Norway, Institute National Polytechnique de Lorraine, France, Péchiney CRV, France, VAW aluminum AG, Germany, and SINTEF, Norway, as a major subcontractor. Funding by the European Community and by the Office Fédéral de l’Education et de la Science (Bern) for the Swiss partners is gratefully acknowledged.
8
References
[1] S.C. Flood, L. Katgerman, A.H. Langille, S. Rogers, C.M. Read, Light Metals 1989, 943947 [2] G.-U. Grün, W.Schneider, Light Metals 1997, 1059-1064 [3] H. G. Fjær, D. Mortensen, A. Håkonsen and Einar A. Sørheim, Light Metal 1999, 743748 [4] D. Mortensen, Metall. Trans. B, 1999, 30B, 119-133 [5] G.-U. Grün, A. Buchholz, D. Mortensen, Light Metals 2000, 573-578 [6] L. Dons, E. K. Jensen, Y. Langsrud, E. Trømborg, S. Brusethaug, Metall. Trans. A, 1999, 30A, 2135-2146 [7] Opstelten, J. Rabenberg, Light Metals 1999, 729-735
Investigations of the Primary Cooling in Sheet Ingot Casting Hallvard G. Fjær1, Andreas Buchholz 2, Benoît Commet3, Jean-Marie Drezet4, and Dag Mortensen1 1
Institute for energy technology (IFE), Kjeller, Norway Corus Research, Development & Technology, Ijmuiden, The Nederlands 3 Pechiney Centre de Recherche de Voreppe, Voreppe, France 4 Laboratoire de Métallurgie Physique, EPFL, Lausanne, Switzerland 2
1
Abstract
A comprehensive series of full-scale sheet ingot DC casting experiments has been carried out within the Brite Euram project EMPACT (European Modelling Project on Aluminium Casting Technology). Temperatures measured by thermocouples situated inside the mould wall provide extensive information about the heat flux during casting from the solidifying metal to the water cooled mould, i.e. the primary cooling. During the start-up period, the primary cooling is seen to be irregular and high compared to the stationary period of casting. A strong correlation between the metal level and the mould temperatures is revealed. A regular fluctuation in the mould temperatures with a 6-12 seconds period is generally observed which is explained by a periodic overflow of liquid metal at the meniscus. Also an example of fluctuations having a period of approx. 30 seconds is registered. The heat flux as function of distance from the metal surface has been estimated by inverse modelling techniques. The heat transfer to the mould is seen to be strongly affected by the development of an air gap induced by the pull-in phenomenon. Several of the observations from the experimental castings have been reproduced by 2D coupled thermomechanical simulations where the heat flux boundary conditions at the mould surface is based on the computed air gap. An important mechanism seems to be the effect of metallostatic pressure on the weak semisolid shell that undergoes a partial re-melting in the air gap zone. The heat transfer in the coupled simulations conforms well with the heat flux distribution found in the inverse modelling analyses.
2
Introduction
In direct chill (DC) casting of aluminium sheet ingots, a solid surface is formed by the thermal contact with a water cooled mould, i.e. the primary cooling. Although only a minor amount of heat is extracted through the mould compared to the direct water cooling zone, the primary cooling is of vital importance for the surface grain structure and the surface segregation [1], as well as for defects like cold shuts, bleed-outs and surface cracks [2]. A comprehensive series of full-scale sheet ingot DC casting experiments has been carried out within the Brite Euram project EMPACT. In this project several European partners have Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
132 joined their efforts to test and improve their numerical models of the aluminium DC casting process. Some of these experiments included temperature measurements by thermocouples situated inside the mould wall, providing extensive information about the heat flux from the solidifying metal to the water cooled mould. In this paper a selection of these mould temperature measurements are presented, and various phenomena connected to the primary cooling are discussed. In addition, by comparing measurements with results from numerical models, the potential of modelling tools to reproduce the complex phenomena related to the primary cooling is illustrated. Numerical models may provide important information complementing the experimental data, and they provide insight into which parameters and properties that might influence the different mechanisms.
3
Experimental
The EMPACT project included experimental DC casting of sheet ingots with cross section 1860mm x 510mm. In these castings the same conventional open mould made of a 6xxx alloy was applied for the casting of different alloys at different cast houses. These experiments included numerous kinds of measurements [3,4]. However, here are only considered measurements by thermocouples (TCs) situated in 1.5mm ∅ holes drilled vertically 2mm below the cooling surface of mould. These TCs where arranged in groups of 20 with a vertical and horizontal spacing of 2mm and 6mm respectively. The lowermost was positioned 30mm from the lower end of the mould, see Figure 1. Four such groups of TCs were applied at different positions along the periphery of the mould, see Figure 2. The development of the mould temperatures has some general characteristics, as can be seen in Figure 1. During the initial phase of casting, the temperatures are relatively high and they display erratic variations, which indicates both a strong and a variable thermal contact between the ingot and the mould. These characteristics are respectively assumed to be related to the minor mould air gap during start-up (i.e. lack of pull-in resulting in a butt swell [2]), and the “bumping” phenomenon caused by boiling of cooling water in the gap between the ingot and the bottom block. When stationary casting conditions are approached, the mould cooling becomes more stable. In the right part of Figure 2 the average temperature for each TC group as well as the metal level (with reference to the lower end of the mould) is plotted as function of time from a casting of an AA5182 ingot. The low temperatures at the short side of the mould (TC group 4) between 200 and 400 sec may be related to the butt curl deformation [5] that during this period involve a bending of the short ingot surface inwards from the mould. Later this TC group has the highest temperatures which may be caused by the flow of hot melt from the distributor positioned at the centre of the ingot towards the short ingot side. The low temperatures in TC group 1 after 800sec may be related to the pull-in effect being strongest at this position [6] or to the flow pattern out of the Combo-bag resulting in the lowest ingot temperatures at the middle of the rolling side [7]. In this casting, a manual regulation entailed large metal level variations. After the start-up period of casting, an interestingly strong correlation between the metal level and the mould temperature was revealed. Two different types of regular oscillations in mould temperatures are depicted in Figure 3. One type with a period of approx. 30sec is assumed to be caused by a periodic freezing/remelting of the surface shell. This type was most evident for AA1050 cast at low speed. The
133 other type with a shorter period (6-12sec) is probably related to the observed “waves” of hot melt overflowing the meniscus and solidifying at the mould surface. 180 160 "TC 0101" "TC 0103" "TC 0105" "TC 0107" "TC 0109" "TC 0111" "TC 0113" "TC 0115" "TC 0117" "TC 0119"
140
2mm
Temperature (°C)
TC x20
10mm
38mm
100 80 60 40
TC x01
30mm
120
"TC 0102" "TC 0104" "TC 0106" "TC 0108" "TC 0110" "TC 0112" "TC 0114" "TC 0116" "TC 0118" "TC 0120"
20 0 0
200
400
600 800 Time (sec)
1000
1200
1400
TC group 3
TC group 2
TC group 1
Temperature(°C)
TC group 4
110
90
100
80
90
70 60
80
50
70
40
60 TC Group 1 TC Group 2 TC Group 3 TC Group 4 Metal level
50 40 30 0
200
400
600 Time (sec)
800
30
Metal level (mm)
Figure 1: Schematic drawing of mould cross section with thermocouple positions (left) and measured mould temperatures where an AA3004 ingot was cast at 60mm/sec (right)
20 10 0
1000
Figure 2: Position of thermocouple groups depicted on a top view of the mould (left). Metal level and (smoothed) average temperature for each TC group plotted as function of time for an AA5182 ingot cast at 60mm/sec (right) TC 0103 TC 0106 TC 0109 TC 0113 TC 0117 TC 0120
Temperature (°C)
60 55 50 45 40
TC 0103 TC 0106 TC 0109 TC 0113 TC 0117 TC 0120
70 Temperature (°C)
65
60
50
35 30 1250
1300
1350 Time (sec)
1400
40 1000
1010
1020 Time (sec)
1030
1040
Figure 3: Regular oscillation in mould temperatures. For AA1050 cast at 50mm/min (left) and for AA5182 cast at 60 mm/min (right)
134 Temperatures (time averaged) for different TC groups as function of vertical position are shown in Figure 4. The measurements from casting of AA5182 show a strong effect of casting speed (and metal level) whereas the mould temperatures for AA3004 reveal both a strong impact of the inflow condition (with/without Combo-bag [4]) and some variation between different castings with the same speed. The differences between the TC groups, especially the fact that the TC group 2 possesses both the highest and lowest temperature, indicate that the cooling conditions could be quite uneven along the periphery of the mould. 100
Temperature (°C)
90
70mm/min TCgroup 2
80 70
50mm/min TCgroup 2
60
60mm/min TC-group 1 (1) 60mm/min TC-group 1 (2)
50
Temperature (°C)
AA5182
60
Cast32 TC-group 1
55
Cast32 TC-group 2
50
Cast32 TC-group 3
45
Cast32 TC-group 4
40
Cast35 TC-group 2
35
Cast35 TC-group 2 without Combo-bag
30 25 20
40 30
40
50 Z-position (mm)
60
30
70
40
50 Z-position (mm)
60
70
Figure 4: Mould temperatures (time averaged for periods with small variations) as function of distance from lowermost TC. For AA5182 (left) and AA3004 (right)
4
Simulations and Discussions
A number of 2D coupled transient thermomechanical simulations of AA5182 castings has been carried out with the thermal model ALSIM [8] and the stress model ALSPEN [9,10] with a solution domain corresponding to a cross section at the centre of the wide ingot side. The heat transfer coefficient (htc) between the ingot and the mould was dependent on the computed displacement at the ingot surface. A gap was assumed where the displacements exceeded 0.2mm. However, the maximum (contact) htc was also set temperature dependent with a maximum of 3000W/m2K for ingot surface temperatures exceeding the liquidus temperature. In addition, the transition between the contact and gap conditions (conduction and radiation) was smoothened. This is considered as reasonable due to surface roughness (segregation). 110 Z=30mm Z=42mm Z=58mm Z=68mm
100
Temperature (oC)
90 80 70 60 50 40 30 20 0
200
400
600
800
1000
1200
1400
Time (sec)
Figure 5: Computed mould temperatures at different z-positions (height above the lower end of the mould) as function of time (left) and temperatures shown on a distorted mesh illustrating the air gap between the ingot and the mould (right)
135 2D simulations are presently requisite in order to allow for a sufficient spatial resolution of the solidifying shell inside the mould. The fluid flow was not computed directly, but it was necessary to account for the 3D flow conditions by applying a reduced melt inflow temperature (on top of the 2D-domain) in order to avoid too high temperatures in both the ingot and the mould compared to the measurements. The computed mould temperatures as function of time shown in the left part of Figure 5 seem to reproduce some of the experimentally observed features quite well. The temperatures are seen to be high during the initial phase of casting when only a minor air gap is predicted. It should also be noted that the metal level in this simulation reached 75mm after 300sec and that the metal level was reduced to 55mm during the period from 700 to 900sec. The short period oscillations associated with the liquid melt overrun is not reproduced as this physics is not included in the model. However, oscillations with a period of about 30sec are obtained (corresponding to the ones that are depicted in the left part of Figure 3). 100 50mm/min Met.lev.65mm
Temperature (°C)
90 80
60mm/min Met.lev.65mm
70 60
70mm/min Met.lev.65mm
50
60mm/min Met.lev.45mm
40
60mm/min Met.lev.75mm
30 30
40
50 Z (mm)
60
70
Figure 6: Computed temperatures 2mm inside mould for different combinations of casting speed and metal level
In the simulation results shown in Figure 5 they are most pronounced between 500 and 800sec. The period of these oscillations has been seen to depend on the metal level, and their amplitude has been shown to be reduced by a stronger temperature dependence of the htc. The computed amplitude has also been found to have some maximum for an intermediate metal level and it was reduced when a higher casting speed was applied in the simulations. These computed oscillations are seen to be created from varying mechanical deformations in the solidifying shell driven by the (small) metallostatic pressure. A weak (and warm) shell deforms easily, resulting in a longer contact zone that in turn makes the whole shell colder and stiffer. A rigid solid shell is then drawn away from the mould surface by the pull-in effect (that is a global inwards bending of the ingot side [11]) and this, in turn, entails re-melting and a weaker shell. The temperature field plotted on a distorted mesh in the right part of Figure 5 illustrate both the formation of the air gap and the deformation of the surface shell. The largest isotherm value indicated in the figure (620°C) corresponds approximately to the coherency temperature defining the extent of the solution domain in the stress model ALSPEN. Investigations have also revealed that the mechanical properties assigned to the mushy zone has a clear influence on the computed ingot shell behaviour. In the mushy zone the flow stress σmush was set equal to the flow stress for the solid σsol multiplied by a factor depending exponentially on the liquid fraction fl, σ mush = exp(−kfl )σ sol . By reducing k, the oscillations grew severely. This corresponds well with the fact that such thermal oscillations were mainly observed for commercially pure aluminium where the extent of the mushy zone is smallest.
136 In Figure 6 are shown mould temperatures at the TC-positions averaged over the last 100sec of simulations with different combinations of casting speed and metal level. The mould temperatures are clearly seen to increase with a higher casting speed and a higher metal level. An increased heat transfer may be explained by the formation of a weaker surface shell due to a longer distance from the initial freezing point to the zone strongly affected by the secondary (water-) cooling below. These simulation results correspond well with the measured values seen in the left part of Figure 4. Inverse modelling, using the same types of techniques as in [3], has been carried out in order to estimate the heat flux from the ingot to the mould. Mould temperatures for two different metal levels from AA5182 cast at 60mm/min (see left part of Figure 4) was applied as input, as depicted in Figure 7. mould
ingot
distribution of heat flux
water chamber
measured temperature profile
adiabatic
x x x x x x x
38 mm cooling water: htc and Tw
Z
measurement at 20 locations
adiabatic
Figure 7: Experimental situation and computation domain. The thermal field inside the mould is shown on the right side (inverse modelling (2) in Figure 8) 1400 1200
Q (kW/m2)
1000
Simulation
800
Inverse modelling (1)
600
Inverse modelling (2)
400 200 0 30
40
50 Z (mm)
60
70
Figure 8: Heat flux at end of simulation compared with results from inverse modelling on heat flux through mould (right)
In Figure 8 the heat flux distribution at 1400sec in a simulation is compared with results from the inverse modelling. The results are in rather good agreement, showing that the contact zone is rather narrow and emphasising the importance of the air gap. However, some of the observed discrepancy might be ascribed to some difference in the applied htc to the water channel inside the mould (6000W/m2 in the inverse modelling compared to 8000W/m2 in the ALSIM simulation), and the fact that adiabatic boundary conditions were applied at the top and at the bottom of the solution domain for the inverse analysis extending over a vertical range corresponding to the TC positions.
137
5
Conclusions
Several of the observations from the experimental castings have been reproduced by 2D coupled thermomechanical simulations where the heat flux boundary conditions at the mould surface is based on the computed air gap. An important mechanism seems to be the effect of the metallostatic pressure on the weak semisolid shell that undergoes a partial re-melting in the air gap zone.
6
Acknowledgement
This work is an addendum to the Brite-Euram Project No. BE-1112 with Contract No. BRPRCT95-0112, which included the following partners: Alusuisse-Lonza Services AG, Switzerland: Calcom SA, Switzerland; Ecole Polytechnique Fédérale de Lausanne, Switzerland; Elkem Aluminium ANS, Norway: Hoogovens Research & Development, The Netherlands; Hydro-Aluminium ASA, Norway; Institut National Polytechnique de Lorraine, France; Pechiney Recherche, France; Technische Universiteit Delft, The Netherlands and Vereinigte Aluminium Werke AG, Germany, and SINTEF, Norway, as a major subcontractor. IFE participated as a sub-contractor to Hydro and Elkem. The authors thank the European Commission and the project partners for financial support and permission to publish this work. The Norwegian contribution to this paper has been financed by the research program PROSMAT supported by the Norwegian Research Council and the Norwegian aluminium industry.
7
References
[1] H. Thevik, A. Mo, T. Rusten, Metall.Trans. 1999, 30B, 135-142. [2] W. Droste, W. Schneider, in Light Metals 1991 (Ed.: E. L. Rooy), TMS, 1991, pp. 945951. [3] J.-M. Drezet, G.-U. Gruen, M. Gremaud, Metall.Trans. 2000, 31A, 1627-1634. [4] A.Buchholz, B. Commet, G.-U. Grün, D. Mortensen, (This book). [5] H. G. Fjær, E. K. Jensen, in Light Metals 1995 (Ed.: J. W. Evans), TMS, 1995, pp. 951959. [6] J.-M. Drezet, M. Rappaz, Metall.Trans. 1996, 27A, 3214-3225. [7] G-U. Gruen, A. Buchholz, D. Mortensen, in Light Metals 2000 (Ed.: R.D. Peterson), TMS, 2000, pp. 573-578. [8] D. Mortensen, Metall.Trans. 1999, 30B, 119-133. [9] H. G. Fjær, A. Mo, Metall.Trans. 1990, 21B, pp. 1049-1061. [10] H. G. Fjær, D. Mortensen, A. Håkonsen, E. A. Sørheim, in Light Metals 1999 (Ed.: C. E. Eckert), TMS, 1999, pp. 743-748. [11] H. G. Fjær, A. Håkonsen, in Light Metals 1997 (Ed.: R. Huglen), TMS, 1997, pp. 683690.
Boiling Curve Approach for Thermal Boundary Conditions in DC Casting J. Zuidema jr. Netherlands Institute for Metals Research, Laboratory for Materials, Advanced Materials and Solidification Technology, Delft, The Netherlands
I.J. Opstelten Corus RD&T, Ijmuiden, The Netherlands
L. Katgerman Netherlands Institute for Metals Research, Laboratory for Materials, Advanced Materials and Solidification Technology, Delft, The Netherlands
1
Introduction
A relation for the heat flux or heat transfer coefficient in the secondary cooling zone in DC Casting is necessary to be able to describe the cooling conditions close to the surface of a billet or slab. A time averaged heat transfer coefficient is normally good enough for describing the thermal behavior in the center region. Because surface thermal history is very important for the quality of the cast product, a relation for the heat flux as a function of process parameters needs to be found. Unfortunately literature does not give the answer in this case. No good theoretical framework exists for establishing thermal boundary conditions in convection and water boiling mixed cooling regime. The best description on this subject can be found in Weckman and Niessen(1). Mr. Jensen from Elkem has derived key values to produce two boiling curves, one for the impingement region and one for the down streaming water region, which in combination produce (almost) the same heat transfer values as a function of distance from the impingement point as those measured (2). In the current research Mr. Jensen’s approach was extended to get better resemblance with experiments.
2
Theory
Between the end of the mould and the impinging water jet advance cooling takes place just above the jet and air-cooling further away. From the point where the water stream from the secondary-cooling water jet first impinges on the surface of the solidified shell to the point one meter downstream, where convection cooling is dominating the heat flux, the cooling behavior can be divided in two regions. In the region were the jet impinges on the surface the heat flux is a function of the velocity of the water jet and the temperatures of the surface and the water. If the surface temperature is above some critical temperature, the boiling mode goes from nucleate to film boiling resulting in a significant lower heat flux. This behavior is not encountered in the current research.
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
139 Further down away from the impinging zone the cooling behavior is gradually changing from boiling dominated to convection dominated. In this region the Weckman and Niessen relation is able to describe the heat flux, q/A in W/m² as function of temperature. The relation is given by q 3 (1) = (− 1.67 *10 5 + 704 * T )Qwater ∆T + 20.8(∆Tx ) 1 3
A
In this equation T is the average of bulk fluid temperature and wall temperature, ∆T is the temperature difference between bulk fluid and Twall, ∆Tx is the temperature difference between Twall and Tsat and Qwater is the water flux in m²/s. Because the process of water-cooling on a DC Casting slab is a process that cannot be modeled easily, some kind of experiment was needed to get heat flux data for different boundary conditions. The heat flux data was obtained at Corus RD & T in IJmuiden.(3) The apparatus used is given in Figure 1. The rig consists of a thick Aluminum plate that can be heated to a uniform temperature and a water-spraying device. This device can be moved in vertical direction along the plate to simulate the DC Casting slab moving along a stationary water-cooling jet. The temperature is measured at some locations close to the surface. From the temperatures measured it is possible to extract the temperature versus distance along the plate (y-location) data. This data can be used to inversely calculate the heat flux as function of the distance from the impinging water jet (impingement point). The software used for this is Calc®MOS. 130 mm
y x
250 mm -z
movement waterfilm 1000 mm
Figure 1: Sketch and Photograph of experimental set-up
The y-axis can be divided in 4 regions: air-cooling zone, advance-cooling zone, impingement zone, and nucleate boiling+convective cooling zone. For every zone a heat flux versus temperature curve can be produced for different boundary conditions. If the heat flux versus temperature curves are fitted against theory, the relations sought for are found: heat flux versus temperature. These relations are valid for different casting speeds, different water flow rates, and different water temperatures. The formulation is not only valid for one alloy. The fit parameters will be different for other alloys however. For regions of air- and advance-cooling behavior no general relations were used in this research. For the impingement region a data-fitting approach based on linear functions for heat-flux as function of temperature was taken. For the nucleate-boiling + convection region the approach is based on a modification of equation 1.
140
The impingement zone is defined here as the point where the surface temperature gradient in vertical direction is at maximum value to 7.5 mm downstream that position. Therefore the impingement region has a size of only 7.5 mm. In this zone the heat flux as function of surface temperature is plotted. As a value of the heat flux in the impingement region the distance based integrated average value of the heat flux is taken. For the value of the surface temperature the average surface temperature in the impingement region is taken. This assures that energy conservation in this region is met.
3
Relations Found from Inverse Calculated Heat Fluxes
The mean impingement heat flux versus temperatures as obtained from the experiments and the relations deduced from it are shown in Figure 2. 6.0E+06
5.0E+06
q [W/m²]
4.0E+06
3.0E+06
2.0E+06
1.0E+06
0.0E+00 100
110
120
130
140
150
160
170
180
190
200
Tsurf [ºC]
Figure 2: Heat fluxes obtained from inversely calculated results and relations fitted for this data points
The results for the impingement region are given by equations 2a, 2b and 2c. q T < 120º C = 2.73 *10 4 T − 1.27 *106 , A
(2a)
q = 9.43*104 T − 9.24*106 , (2b) 120 º C ≤ T < 150 º C A q = 1.23*10 4 T + 3.06*106 , (2c) T ≥ 150º C A The q-T relations for the impingement region plotted in Figure 2 gave the closest resemblance to the experimental temperature fields when used as a boundary condition in a direct temperature field calculation. The q-T relation for the nucleate boiling + convection region further downstream from the impingement point was found after adapting equation 1 to the inversely calculated q-T data. The function that was found to give the best match to experimental temperature fields is given by q 3 (3) = ( −1.67 *105 + c * T ) Qwater ∆T + 100 ( ∆Tx ) 1 3
A
where
141 2
c = −1.11*10−3 * Qw + 1.15* Qw + 628 (4) The fit parameter c is only valid for interpolation between water flow rates Qw of 120 l/m/min to 500 l/m/min. The bulk water temperature was taken constant for equation 3. Its value was chosen to be 45 ºC. The ‘saturation’ temperature was chosen to be 90 ºC. Saturation is shown in quotes because it is not the boiling temperature of the water. It has a close relation to the boiling temperature however. To be able to use equation 3 it must be bounded for high temperatures and must go to convective-only behavior for low temperatures. Equation 3 is therefore modified to 1 ( −1.67 *105 + c * T ) Qwater 3 ∆T , q 6 (5) = max 0, min 3.0 *10 , max 1 3 5 3 A 1.67 *10 c * T Q T 100 T − + ∆ + ∆ ( x ) ) water (
4
Calculating the Temperature Field
With the relations obtained in the previous chapter it should now be possible to calculate temperature profiles in the aluminum block for all water flow rates between 120 l/m/min and 500 l/m/min, for cast speeds from 1.66 to 6.64 mm/s and uniform start temperatures between 300 and 500 ºC. From the 20 cases with different cooling parameters 2 examples are shown. In figure 3 a block cooled from 400 ºC with a water traverse speed of 6.64 mm/s and a Qw of 240 l/m/min is shown. 5.00E+02
4.50E+02
4.00E+02
3.50E+02
T calc0 T calc1 T calc2 T calc3 T calc4 T1 T2 T3 T4
T [ºC]
3.00E+02
2.50E+02
2.00E+02
1.50E+02
1.00E+02
5.00E+01
0.00E+00 9.50E-01
1.00E+00
1.05E+00
y [m ]
1.10E+00
1.15E+00
1.20E+00
Figure 3: Measured and calculated temperature profiles along slab for case a140897
In figure 4 a block cooled from 500 ºC with a water traverse speed of 3.32 mm/s and a Qw of 500 l/m/min is shown.
142 500
450
400
350
T1 T2 T3 T4 Tcalc0 Tcalc1 Tcalc2 Tcalc3 Tcalc4
T [ºC]
300
250
200
150
100
50
0 0.95
1
1.05
y [m ]
1.1
1.15
1.2
Figure 4: Measured and calculated temperature profiles along slab for case a051297
5
Discussion and Conclusions
The heat transfer database produced at Corus RD&T has been analyzed with help of inverse modeled heat flux versus temperature relations. Empirical relations for two of the four different flow regions, defined in chapter 3, in secondary heat transfer in DC casting were obtained. The Impingement and the down-streaming region are described by a functions of temperature and casting parameters. For the air-cooling region and the advance cooling region the results from the inverse modeling were used. In the inverse modeling a linear increase in heat flux from air-cooling region to the impingement region was taken. The advance-cooling region was 10 mm in the inverse modeling. In the approach taken here the advance cooling was taken constant and the advance cooling region size was only 5 mm. The difference in total heat subtracted in this region compared to the inverse modeling results is less then 10% except for case a090997. The calculated temperature fields with the current approach were found to be within close range from measured temperature fields indicating that a general set of relations for different water flow rates is capable of describing accurately the temperature of the surface of a DC casting slab. The results get better with higher water flow rates.
6
References
[1] D. C. Weckman and P. Niessen, Metall. Trans. B 1982, 13, 593-602. [2] Einar K. Jensen, Report on evaluation of laboratory experimental heat transfer data by comparing with ALSIM modelling of the experiments, subtask 3.3, Febr. 1999, 11 p. [3] Ivo J. Opstelten and Jan M. Rabenberg, Light Metals 1999, TMS [CD ROM].
Theoretical and Experimental Study of Vertical Continuous Casting of Copper Markku Uoti Helsinki University of Technology, Laboratory of Metallurgy,P.O. Box 6200, FIN - 02015 HUT, Finland
Mikko Immonen, Kalle Härkki Outokumpu Poricopper OY, Kuparitie, P.O.Box 60, FIN - 28101 Pori, Finland
1
Abstract
This theoretical and experimental study was carried out to investigate heat transfer and solidification in vertical continuous casting of copper. The study combined plant scale temperature measurements at the Outokumpu Poricopper foundry and mathematical modeling. The commercial FEM based software FIDAP was used to calculate heat transfer and solidification phenomena. Liquid pool depth, mould temperatures and temperatures of the cast slab were determined. After tuning the model, the mathematical model was used to investigate the effects of parameters such as heat transfer coefficient and casting velocity on the temperature distribution of the cast slab, especially on the shape and location of the solidification front.
2
Introduction
In the continuous casting process the heat transfer in the mould is one of the main factors limiting the maximum productivity. By increasing the casting speed more heat has to be transported through the mould. But it is not simply a maximization of heat extraction, it is also of great importance to remove the heat from the slab in a controlled way. Uneven heat transfer can cause quality problems such as cracks. Experimental tests were made to study the heat flow and solidification in the copper slab casting case. Results from the measurements were used to calibrate the mathematical model. The mathematical model was then used to investigate the impact of different heat transfer coefficients and casting velocities to the temperature distribution of the cast slab.
3
Heat Transfer in the Mould
The mould consists of a graphite die and a copper mould. A schematic presentation of the temperature profile in the mould and the cast slab during the solidification phenomenon is presented in Fig. 1. Heat flow is generally controlled by the formation of an air gap between the cast slab and the graphite die. Further significant heat barrier can exist between the graphite die and the copper mould. [1] Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
144 Water cooling
Copper mould
A i r g a p
Liquid Solid shell
Graphite die T e m p e r a t u r e
Figure 1: Schematic presentation of the temperature profile in the mould and cast slab
3.1
Heat Transfer Across the Air Gaps
If the gap thickness variation with time would be known exactly, the heat transfer across the gap could be estimated. However, this is very difficult in practice, so the heat transfer associated between two surfaces with an air gap is usually treated empirically with a heat transfer coefficient. The total heat flux can be calculated with Eq. 1, q = hgap (T1 − T2 ) (1) where hgap is the total heat transfer coefficient (W/m2K) including the three heat transfer mechanisms: convection, conduction and radiation. T1 and T2 are the surface temperatures on the opposite sides forming the air gap [2]. At high temperatures above the liquidus the contact between the casting shell and the graphite die remains good. As soon as the air gap forms, the contact between the surfaces is weakened, which leads to decreased heat extraction. For this reason it was necessary to use a temperature dependent heat transfer coefficient at this interface. The graphite die and the copper mould have different heat expansion coefficients, and therefore an air gap can form at their interface, which also leads to reduced heat extraction. Because the degree of the contact between the graphite die and the copper mould determines the heat transfer, a temperature dependent heat transfer coefficient is not reasonable to use. In our study a constant value was supposed.
4
Experimental Part
In the casting operation of copper slabs the melt is transferred from the holding furnace via a submerged feeding tube into the mould. The casting machine consists of a primary- and secondary cooling area. The primary cooling area consists of a water-cooled mould where the solidification of liquid copper begins. Respectively, the secondary cooling area is constructed of a water spray area and a water pool. A schematic presentation of the experimental setup and measurements is shown in Fig. 2.
145
Copper mould Graphite die
Mould temperature measurements
Liquid
Slab temperature measurements
Stick test
Water spray
Water pool
Solid
Figure 2: Schematic presentation of the experimental setup and measurements
In order to investigate heat flow and solidification phenomena several measurements were made. To adjust the mathematical model the liquid pool depth, mould wall temperatures, cooling water temperatures and water flow rates were measured. The cast slab temperature profile was also determined in order to test the model. 4.1
Description of the Experimental Measurements
The liquid pool depth was measured by a stick test. Several steel rods were put into the liquid pool of the cast slab and removed when touching the solidification front (Fig. 2). To measure the temperatures from the mould wall, several thermocouples were inserted between the graphite die and the copper mould at different heights. The objective was to determine the temperature profile in the mould along the casting (Fig. 2). Total heat flux in the mould was estimated by measuring the flow rate of the cooling water and the increase in the cooling water temperature at every inlet and outlet. This information was applied in the determination of the boundary conditions used in the mathematical model. The target of the thermocouple measurements in the cast slab was to determine temperature profiles in the slab during the casting process. Three thermocouples were placed in a base that was inserted in the liquid pool close to meniscus level. Afterwards it was allowed to follow the solidifying slab downwards in the casting machine (Fig. 2). The temperature profile was used to test the numerical model.
5
Numerical Modeling
5.1
Software
The commercial software FIDAP (Version 8.0) was used to perform the numerical modeling. The solution method used in FIDAP is based on the finite element method (FEM). FIDAP can be considered as an integrated set of program modules designed to perform all aspects including model generation, problem setup, solution and post-processing. FIDAP provides a
146 number of different methods which are appropriate for phase change problems depending on the type of material involved. In this case the enthalpy model was used. [3] Special gap entities, which consist of boundary gap heat transfer elements were also utilized in FIDAP. The purpose of these entities was to impose heat transfer coefficients as a function of temperature between the slab – graphite die interface and the graphite die – copper mould interface. To obtain reliable results from the numerical modeling, accurate thermophysical material data are needed. Typical data are density, thermal conductivity and specific heat. The heat transfer equation in the simulations was solved in 3D. The governing differential equation (Eq. 2) in this steady-state calculation was ∂T ∂ ρ ⋅ H ⋅ VX ⋅ = ∂x ∂x
∂T ∂ ∂T ∂ k + k + ∂x ∂y ∂y ∂z
∂T k ∂z
(2)
where ρ is the density, H is the enthalpy including the latent heat, k is the thermal conductivity and Vx is the velocity component of the copper in the casting direction. [4] 5.2
Assumptions and Boundary Conditions
The cast slab was assumed to have a fixed grid. This means that the element does not shrink with the temperature drop. Therefore the density can not change either. In such cases, the density should be that of the initial liquid and constant. However, during solidification, the fluid in the interdendritic space is free to move and it more or less compensates for the solidification contractions. To take this feeding into account, the density should be the density at the solidus temperature [5]. The flow field of the copper was assumed to be constant. The liquid flow was set to be equal to the casting speed. The heat transfer coefficient between the cast slab and the graphite die was iteratively found out to be a function of the slab surface temperature as shown in Fig. 3. Heat transfer coefficient (W/m 2K)
2000 1800 1600 1400 1200 1000 800 0
200
400
600
800
1000
1200
Temperature (°C)
Figure 3: Heat transfer coefficient as a function of the slab surface temperature
As mentioned earlier, it is not useful to assume a temperature dependent heat transfer boundary condition at the graphite die – copper mould interface, because heat transfer depends greatly on the surface quality and the contact of the opposite interfaces. A constant value was used in this study. The value was iteratively simulated to be 1000 W/m2K.
147
6
Results
The simulation showed good agreement with the measured temperature profile as seen in Fig. 4. This is an indication that FIDAP software can be used in casting related simulations. T e m p e r a tu r e p r o file o f th e c a s tin g s la b 1200
Temperature (°C)
1000
C a lc u la t e d te m p e r a tu r e
800
600
M e a s u re d te m p e r a tu r e
400
200
0
D is t a n c e f r o m t h e t o p o f th e m o u ld
Figure 4: Comparison between calculated and measured temperature profiles
The effect of the casting speed on the shape and position of the solidification front was investigated at three different casting speeds. In Fig. 5 a cross-section through the broad and narrow sides of the cast slab is shown, including the graphite die and the surrounding copper mould. Due to the symmetric nature of the mould, only half of it is presented.
V cast
1.5*V
cast
2.0*V
cast
Figure 5: Shape and position of the solidification front at different casting speeds
Fig. 5 shows the shapes and positions of solidification front at three different casting speeds ( Vcast , 1.5 * Vcast and 2 * Vcast ). The liquid pool depth increases when the casting velocity increases. The wedge-shaped solidification front is formed due to the weak cooling efficiency of the narrow side in comparison to the broad side of the cast slab.
148
7
Concluding Remarks
Heat transfer and solidification were studied numerically and experimentally. The heat transfer coefficients in the calculations were adjusted to match the measured temperature profiles with the calculated ones. Therefore the results are only valid for this casting process geometry and casting material. The effect of different casting parameters was examined only on the length and shape of the solidification front. To be exact, modeling of this kind of processes would have implied simultaneous treatment of heat transfer, solidification, fluid flow and stress formation [5]. Due to the highly non-linear mechanical behavior of copper just under the solidification temperature, the numerical simulations are very difficult and time demanding.
8
References
[1] K. Härkki, Heat Transfer and Solidification in Copper and Brass Upcasting, Acta Polytechnica Scandinavica, Chemical Technology Series No. 253, Espoo, 1997, 69 pp. [2] D.R. Poirier, E.J. Poirier, Heat Transfer Fundamentals for Metal Casting. Second edition with SI Units. The Minerals, Metals & Materials Society, 1994, p. 1-41. [3] Fluid Dynamics International, FIDAP 7.0 FIPREP Users Manual, 1993 [4] D.C. Prasso, J.W. Evans and I.J. Wilson, Heat Transport and Solidification in the Electromagnetic Casting of Aluminium Alloys: Part II. Development of a Mathematical Model and Comparison with Experimental Results. Metall. Trans. B. 1995 Vol. 26B, p. 1281-1287. [5] S. Louhenkilpi, Simulation and control of heat transfer in continuous casting of steel, Acta Polytechnica Scandinavica, Chemical Technology Series No. 230, Helsinki, 1995, 37 pp.
Modeling of Grain Refinement in Aluminum Alloys A. L. Greer and A. Tronche University of Cambridge, Cambridge, UK
1
Abstract
An assessment is made of progress in predicting grain size in inoculated Al alloy melts. A new ‘free-growth’ model gives good fits to observed behavior, but fails in the presence of a steep temperature gradient or of solute ‘poisoning’ of the nucleation itself. Nevertheless, the model is useful in considering possible improvements in refiner design.
2
Introduction
In aluminum casting it would clearly be useful to have quantitative predictions of grain size, taking account of the nature of the alloy, the nature and amount of added refiner, and other processing variables. Based on better understanding of the underlying mechanisms, it may be possible to develop improved grain-refining practice, giving finer grain size, more uniform grain size, and equiaxed structures for lower addition levels. There appears to be much room for improvement, as existing refining practice is rather inefficient — at best 1% of the added particles nucleate grains, even in the absence of any impairment of the nucleation process itself through ‘poisoning’ by particular solutes. In the present work the predictive capabilities of the ‘free-growth’ model [1] are assessed.
3
Free-Growth Model
Maxwell and Hellawell [2] considered that the nucleation of new grains in an undercooled melt would most likely be limited by recalescence of the melt. They noted that typical thermal diffusion lengths are much greater than solute diffusion lengths, and greater than grain diameters; it is then reasonable to treat the solidifying melt as spatially isothermal, with crystal growth limited by solute diffusion. Assuming classical spherical-cap heterogeneous nucleation on inoculant particles, they showed that recalescence could provide a semiquantitative understanding of low refiner efficiency. However, nucleation can occur at very low undercoolings (0.01 to 0.5 K [3]) at which the spherical-cap nucleation model breaks down [4]. Furthermore a nucleated spherical cap could grow into a grain only if it breaks through the critical (minimum radius) hemispherical condition (Fig. 1a, inset). The undercooling for grain initiation by such ‘free growth’ is inversely proportional to the diameter of the relevant inoculant particle face; for typical inoculant particle sizes (Fig. 1b), this undercooling (Fig. 1a) is greater than or comparable with that required for nucleation. Thus free growth may be the rate-limiting step for grain Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
150 initiation. The key input parameter for modeling is then the particle-size distribution, which is directly measurable (unlike the contact angle involved in heterogeneous nucleation). Figure 1b shows the measured distribution in a commercial Al-5Ti-1B (wt.%) refiner. When the melt is cooled, grains are initiated first on the largest particles and then on progressively smaller particles until recalescence. The details of the free-growth model, the input parameters and the application to Al-Ti-B refiners have been given elsewhere [1]. Here, the work is extended to Al-Ti-C refiners and to considerations of refiner design.
(b)
(a) TiB2 d
Figure 1: The free-growth model. (a) The undercooling necessary to initiate free growth for the face of a discshaped TiB2 particle is inversely proportional to the particle diameter. (b) The measured diameter distribution of TiB2 particles in a commercial Al-5Ti-1B refiner. The particles are hexagonal platelets, approximated as discs. Under normal conditions, only the largest particles (shaded area) initiate grains
4
Results
Standard TP-1 tests [5] were conducted on commercial-purity aluminum (CP-Al) inoculated with commercial Al-Ti-B or Al-Ti-C refiners [6]. Following standard techniques, the grain size (i.e. mean lineal intercept, l) was measured on sections cut from the test samples. As an example, Fig. 2 shows the grain size as a function of addition level of Al-Ti-B refiner, and for comparison the corresponding prediction from the free-growth model. (To permit this comparison it is necessary to relate the number of grains per unit volume NV to l; as justified in [1], the relationship is taken to be NV = 0.5/l3.) Both model and experiment show that the grain size falls rapidly with addition level at first but then tends to saturation. The agreement between model and experiment is remarkably good, especially as the calculations are based on input parameters (solid-liquid interfacial energy, latent heat of fusion, specific heat of the liquid, inoculant particle-size distribution, liquidus slope m and partition coefficient k for the effective solute level C0) all of which are known independently and not adjustable. Restriction of crystal growth by solute is essential for grain refinement, and can be characterized by the parameter Q = m (k – 1) C0 [7]. The CP-Al was modeled as Al - 0.0089 wt.% Ti, this Ti content providing the same overall growth-restriction factor Q as the sum of the Qs from the various solutes (of which Ti is dominant) at their measured contents [1]. The addition of refiner increases the Ti content, e.g., by 0.0028 wt.% for each 1 ppt (parts per thousand, or kg tonne–1) of Al-5Ti-1B refiner. The inset in Fig. 2 shows, again with good agreement between measurement and modeling, that the efficiency of inoculation (i.e., no. of grains per inoculant particle) is very low (at best 1%) and is worse at higher addition levels. There is a smooth variation of efficiency with addition level, unlike the two-regime behavior predicted by the earlier Maxwell-Hellawell model based on heterogeneous nucleation kinetics [2].
151 700
Grain Size (µm)
600 500 400 300 200 100 0
prediction 0
2
4
6
8
Addition Level of Refiner (ppt)
10
Figure 2: Comparison of measured (TP-1 test) and calculated (free-growth model) grain sizes in CP-Al as a function of the addition level of refiner. The inset shows the same data plotted as numbers of grains and particles
The general form of the grain-size curve in Fig. 2 is found not only as addition level L is increased (at constant Q and cooling rate dT/dt), but also for increases in Q (at constant L and dT/dt) and in dT/dt (at constant L and Q) [1]. These experimental findings are satisfactorily reproduced by the free-growth modeling [1]. Figure 3 provides a summary of the performance of the modeling, using a wide range of experimental data (newly acquired and from the literature [8]), over ranges of L = 0.01 to 10 ppt, Q = 1 to 18 K, dT/dt = 0.5 to 5.5 K s–1, for Al-5Ti-1B and for Al-3Ti-0.15C (for which the particles are octahedra rather than discs). The agreement between experiment and modeling is good for grain sizes less than 400 µm, but for larger sizes, the real grain structures in the TP-1 tests are columnar not equiaxed, and the model does not apply.
Computed Grain Size ( µm)
1000 800 600 400 200 0
0
200
400
600
800
1000
Measured Grain Size µ ( m)
Figure 3: Comparison of grain sizes computed using the free-growth model and those found in experiment (mostly TP-1 tests), over ranges of refiner addition level, cooling rate and solute types and levels: • Al-5Ti-1B; × Al-3Ti-0.15C
The success of the modeling permits it to be used in considering how to design better refiners. The size distribution of the refiner particles is central in the modeling; it might in principle be changed by altering the manufacturing route, and it is of interest to know what size distribution would be optimal. Preliminary calculations have been carried out on Gaussian distributions of particle diameter, with varying average diameter and dispersion in
152 diameter. Narrow size distributions (for constant total volume fraction of inoculant particles) tend to give finer grain size, but this is not a major effect. The main effect is of the average particle diameter, as shown in Fig. 4, which shows that there is a clear optimum size. At larger size (for given total volume fraction) there are fewer inoculant particles and this restricts grain refinement. For smaller size, grain initiation occurs at greater undercooling and the more rapid growth facilitates the restriction of grain refinement by recalescence. The optimum particle diameter, ~5 µm, is larger than in a commercial refiner and perhaps larger than could be tolerated in many products. 600 σ = 10% λ
Grain Size ( µm)
500
Commercial Al-5Ti-1B Refiner
400 σ = 25% λ
300 200 100
0
2
4
6
8
10
12
Average Particle Diameter, λ (µm)
Figure 4: Computed grain size for a Gaussian particle-size distribution of average particle diameter λ, and with widths 10% or 25% λ. The grain size obtained with the size distribution in a commercial refiner is also indicated. (Addition level 2 ppt; cooling rate 3.5 K s–1)
In casting of aluminum alloys, uniformity of grain size may be more important than fine grain size. Grain size may vary across a casting for many reasons; the modeling suggests that one of these reasons may be variation in cooling rate. For idealized Gaussian size distributions, Fig. 5 shows the sensitivity of grain size to cooling rate for different average diameters of inoculant particle. Clearly, larger particles give greater insensitivity to cooling rate, and an average diameter of ~5 µm gives a similar performance to a commercial refiner. 600 λ = 1 µm
Grain Size (µm)
500 400
Commercial Al-5Ti-1B refiner λ = 10 µm
λ = 5 µm
300 200 100 0 0
2
4
6
8
10
12
-1
Cooling Rate (K s )
Figure 5: Computed grain size for Gaussian particle-size distributions as a function of cooling rate, compared to the variation obtained for the distribution in a commercial refiner. (Distribution width = 25% λ; addition level 2 ppt; cooling rate 3.5 K s–1)
153
5
Discussion and Conclusions
The predictions of free-growth modeling match experiment well for both the Al-5Ti-1B and Al-3Ti-0.15C refiners used in the TP-1 test. This appears to justify the key assumptions: that grain initiation is controlled by the free-growth condition and not by nucleation per se; that the melt can be treated as isothermal; that grain refinement is limited by recalescence; and that solutes play a key role in slowing crystal growth, permitting greater undercooling to be reached. However the model deals only with equiaxed solidification and cannot predict the transition to columnar growth which is expected at higher temperature gradients; this transition is analyzed in [9]. Also, grain initiation in for example DC-casting may arise not only by direct inoculation, but also through processes such as dendrite fragmentation promoted by convection in the melt. Nevertheless, application of the free-growth model to understand the results of TP-1 tests is of more general significance in identifying additional phenomena; the key area of interest here is poisoning, in which particular solute elements (Zr or Cr for Al-Ti-B, Si for Al-Ti-C) lead to an increase in grain size. The free-growth model cannot itself account for this, and poisoning appears to be largely through degradation of the nucleation mechanism itself.
6
Acknowledgements
ALG’s research on grain refinement is supported by EPSRC (UK). financial support from LSM Co. Ltd and Pechiney (CIFRE studentship).
7
AT acknowledges
References
[1] L. Greer, A. M. Bunn, A. Tronche, P. V. Evans, D. J. Bristow, Acta Mater. 2000, 48, 2823 - 2835. [2] Maxwell, A. Hellawell, Acta Metall. 1975, 23, 229 - 237. [3] M. Johnsson, L. Bäckerud, G. K. Sigworth, Metall. Trans. A, 1993, 24, 481 - 491. [4] W. T. Kim, B. Cantor, Acta Metall. Mater. 1994, 42, 3045 - 3053. [5] Aluminum Association Standard Test Procedure for Aluminum Grain Refiners, The Aluminum Association, Washington DC 20006, 1990. [6] kindly supplied by London & Scandinavian Metallurgical Co. Ltd. [7] Tronche, A. L. Greer, these proceedings. [8] J. A. Spittle, S. Sadli, Mater. Sci. Technol. 1995, 11, 533 - 537. [9] M. Vandyoussefi, A. L. Greer, these proceedings.
Modeling of the Grain Refinement in Directionally Solidified Al-4.15 wt.% Mg Alloys using Cellular Automaton – Finite Element Approach M. Vandyoussefi and A. L. Greer University of Cambridge, Cambridge, UK
1
Abstract
Grain refinement of Al-4.15 wt.% Mg alloys is studied using a 2-D CAFE (Cellular Automaton - Finite Element) model, in combination with Bridgman experiments. At low solidification rates and addition levels of grain refiners, short columnar grains, elongated and twinned grains were observed. Fully equiaxed dendrites were observed only at addition levels greater than conventionally used, and at higher solidification rates. Showing similar variations of grain size with processing parameters, both modeling and experimental results demonstrate that grain refinement is limited in the presence of a temperature gradient. A columnar/equiaxed transition (CET) map is calculated, indicating that the CET occurs progressively by varying solidification rate and temperature gradient. The microstructures observed in the samples inoculated with 2 and 5 ppt of Al-Ti-C refiner confirm this prediction.
2
Introduction
Prediction of processing parameters appropriate for the grain refinement of aluminum (i.e., density of nucleation sites, temperature gradient (G) and solidification rate (V) or cooling rate (R)) is of great technical importance. In the early model developed by Maxwell and Hellawell [1], it is assumed that grains initiate by nucleation on the heterogeneous particles of added inoculants. Then they grow as spheres in an isothermal melt during the initial stages before morphological breakdown. However, the spherical-cap model used for heterogeneous nucleation is not valid at very low undercooling. The nucleation undercooling may be as low as 0.01 K in the case of Al-Ti-B refiner [2]. For such potent nucleation the contact angle would be so low as to be unrealistic at an atomic level. Greer et al. have recently modified the Maxwell and Hellawell model, suggesting that nucleation is not the rate controlling process for grain formation; rather a “free growth” criterion controls the initiation of new grains [3]. Although this model has had some success, it still assumes an isothermal melt and may therefore be invalid in the presence of a significant temperature gradient. Also a strong gradient affects the shape of grains otherwise taken to be equiaxed [4]. The prediction of the CET is also important in the grain refinement as it determines the amount of refiner necessary to obtain fully equiaxed grains, a quantity not predictable using the isothermal model of Greer et al. [3]. A probabilistic approach is an alternative to treat grain refinement and solidification, and can overcome this limitation. Several stochastic Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
155 models have been developed for the prediction of dendritic grain structure with differing degrees of consistency with experiments [5-10]. The aim of the present work was to study the grain refinement of Al alloys under directional growth in a temperature gradient, characterizing the CET via both experiment and modeling approaches. Using commercial software (CalcoMOSTM) [11], a stochastic 2-D Cellular Automaton - Finite Element (CAFE) model has been applied to the grain refinement of Al-Mg alloys [6-8]. Although previous workers had investigated the CET in an Al- 7 wt.% Si cylindrical sample cast in a bottom chilled mold using this model [7], it was not tested for the case of grain refinement including the effects of the addition level of refiner (affecting the nucleation density), and the separate effects of G and V.
3
Experimental Procedure
Bridgman solidification was used to study the effect of process parameters on the grain structure of Al- 4.15 wt.% Mg alloys. Details of the apparatus were given elsewhere [12]. Inoculated alloys were prepared by adding different amounts of carbide grain refiner (Al- 4.91 wt.% Ti - 0.14 wt.% C) varying from 2 to 20 ppt (kg/tonne) to the base melt. Each charge was contained in an alumina tube 20 cm long with 2 mm diameter. The temperature gradient in the melt was determined to be 10 ± 2 ºC/mm [12]. The experimental conditions and chemical analyses of the alloys are given in Table 1. The specimens were etched in an HF solution (5% HF, 95% H2O). The grain sizes were measured using the lineal intercept method.
a)
b)
c)
Figure 1: The microstructures of the Bridgman samples at the quenched interface (a) non-grain refined sample 1 solidified at 0.1 mm/s, (b) sample 5 inoculated with 2 ppt and solidified at 0.2 mm/s and (c) sample 7 inoculated with 2 ppt and solidified at 1 mm/s
156 Table 1: Experimental conditions for the Bridgman tests and summary of observations. The concentrations of major impurities (Si, Fe and Mn) are less than 0.01 wt.%. Samples Pulling Addition rate level [mm/s] [ppt] 1, 2, 3 0.1, 0.5, 1 0 4 0.1 2
4
Composition Microstructure [wt.%] Mg Ti 4.15 < 0.001 columnar 4.12 0.007 short columnar
5 6
0.2 0.5
" "
" "
7 8 9 10 11 12 13 14 15 16 17
1 0.1 0.2 0.5 1 0.1 0.5 1 0.1 0.5 1
" 5 " " " 10 " " 20 " "
" 4.15 " " " 4.16 " " 4.13 " "
" "
" equiaxed + twinned tweltwielonagteddendrites " equiaxed 0.021 elongated + twinned " elongated + equiaxed " equiaxed " " 0.036 " " " " " 0.083 " " " " "
Experimental Results
Table 1 summarizes the microstructures observed in the samples. All of the non-grain- refined samples showed columnar dendrites growing anti-parallel to the heat flow; Figure 1a shows a typical example in sample 1 solidified at 0.1 mm/s. At 2 ppt addition level, no equiaxed grains were observed in samples 4 and 5 solidified at 0.1 to 0.2 mm/s. Instead the microstructure consisted of short columnar grains which are different from those in non-grain-refined samples (Fig. 1b). At 0.5 mm/s, equiaxed dendritic grains with non-uniform size and shape and also elongated twinned grains were observed. A fully equiaxed grain structure was observed only at 1 mm/s (Fig. 1c). As shown, the interdendritic liquid solidified as very fine dendrites upon quenching. At 5 ppt addition level, relatively elongated grains with varying aspect ratio were seen at 0.1 mm/s. The microstructure was a mixture of short elongated and equiaxed grains at 0.2 mm/s. At higher solidification rates, the microstructures were fully equiaxed (samples 10 and 11). At 10 and 20 ppt addition levels (sample 12-14 and 15-17), the grain structures were thoroughly equiaxed over the entire solidification range. Al- 4.15 wt.% Mg shows eutectic solidification at 450º C, a temperature outside the region shown; in steady state Bridgman growth, full solidification occurs far behind the front shown in Fig. 1.
157
5
Discussion
Generally, addition of 1 to 5 ppt of grain refiners produces fine equiaxed grains in the standard TP1 grain refinement test. In contrast, very little refinement was observed in the Bridgman samples inoculated with 2 and 5 ppt grain refiners and solidified at low and intermediate solidification rates (Table 1). This is likely due to the effect of higher temperature gradient in Bridgman samples. When G is increased, the extent of constitutionally undercooled zone ahead of the advancing solidification interface decreases. Consequently, a smaller number of nucleants becomes active and can grow for a relatively long distance down into undercooled melt before impinging on elongated grains which merge into short columns. This mode of growth leads to the formation of short columnar and elongated grains. Higher amounts of grain refiner are required to promote the formation of fine equiaxed grains in the Bridgman experiments although the cooling rates may be similar to ones encountered in the TP-1 test. In the present work, the required level was >10 ppt at 0.1 mm/s (sample 12) and >5 ppt at 0.5 mm/s (sample 10). 1 mm
3 K/mm
6 K/mm 1011 m-3
10 K/mm 1012 m-3
1013 m-3
Figure 2: CAFE modeling of effect of G and Nmax on the grain structure in a Bridgman sample solidifying at V= 0.1 mm/s. Only bulk nucleation was considered for which a Gaussian distribution was used with variable Nmax and ∆Tm = 3 and ∆Tσ = 0.5 K. For the calculation, phase diagram properties were obtained from a standard reference [13]
In the CAFE model, a two-dimensional regular network of square cells (cellular-automaton grid) is defined in which nucleation happens randomly over a spectrum of undercoolings, both at the surface of mould and bulk of melt. Nucleation behavior is treated using a Gaussian distribution which relates nucleation density (dn/d∆T) to undercooling. It is defined by median undercooling (∆Tm), standard deviation (∆Tσ) undercooling and maximum number of nucleants (Nmax). Growth is dendritic with the preferred direction (<100>). The growth of dendrite tips controls the evolution of grains with kinetics calculated using an analytical dendritic growth model [14]. The temperature distribution is calculated using a 2-D enthalpybased finite element (FE) algorithm which is coupled with the latent heat released in the CA cells and with changes in specific heat. Figure 2 shows examples of the modeled grain structures. At constant nucleation density Nmax and V, the grains are equiaxed at low G but higher G produces elongated grains. The aspect ratio of these grains increases by increasing G. Similarly at constant G and V, decreasing the number of nucleants leads to the formation of elongated grains.
158 20 ppt 10 ppt 5 ppt 2 ppt
(a)
Grain size [µm]
200
1012 [m -3]
100
(b)
1013 [m -3]
00
0.2
0.4 0.6 V [mm/s]
0.8
495
179
equiaxed region 522
165
199
increasing aspect ratio
265 183
-4
-5 1
150 154 160
-4.5
50
391
-3 -3.5
1011 [m -3]
150
-2.5
Log V [m/s]
250
-5.5 2
CET line Hunt model
2.5
3
elongated/columnar region
3.5 4 4.5 Log G [ºC/m]
5
5.5
Figure 3. (a) Comparison between the measured (data points, for selected refiner addition levels) and calculated grain sizes (solid lines, for selected Nmax). (b) A CET map showing the results of CAFE modeling with Nmax = 1011 m-3; solid points indicate equiaxed grains with sizes as shown, open points indicate columnar grains. The CET line between these (dashed) is similar to that predicted by Hunt model (solid line) with nucleation density 5x1010 m-3 and ∆Tnucleation= 0.2 K [15]. Measured grain structures are also indicated ×: equiaxed and +: elongated/columnar for 2 and 5 ppt addition levels.
Figure 3a compares measured and calculated grain sizes versus solidification rate for different addition levels of grain refiner and Nmax. The beginning of each curve corresponds to the suppression of elongated grains. For a given Nmax or addition level, increasing solidification rate (or cooling rate) reduces grain size consistently in both measured and calculated results. At low addition level (2 and 5 ppt), there is reasonable agreement between the measured and calculated results. Also, the growth of short columnar and elongated grains in the samples inoculated with 2 and 5 ppt grown at 0.1 and 0.2 mm/s (samples 4, 5, 8 and 9) is consistent with the growth of elongated grains at Nmax=1011 m-3. The measured and calculated grain sizes display similar trends at higher addition levels of grain refiner (10 and 20 ppt) and Nmax, although better fits could be made by adjusting Nmax. The experimental (samples 4-10) and modeled results show that the CET occurs gradually in the Al-Mg alloys, rather than abruptly. Figure 3b shows a CET map in which this transition and the size of equiaxed grains are given as a function of G and V for a given set of nucleation parameters. The CET line divides the map into regions of equiaxed and elongated grains. For a given G, increasing V yields finer grains in the equiaxed region as the numbers indicate in Fig. 3b. However, the grain size close to the transition line does not change substantially by varying G and V. Also, at and beyond the CET, the grains are not columnar (infinite aspect ratio) but they tend to be elongated. As the transition is progressive, this line has been drawn somewhat arbitrarily. Obviously, the transition is also controlled by the nucleation density, a higher number of nucleants expanding the equiaxed region. The CET was earlier modeled by Hunt without a stochastic treatment of microstructure [15]. This model predicts a similar trend, but is not able to predict whether the transition is gradual or abrupt. Experiments are reasonably consistent with this map for the samples refined with 2 and 5 ppt of refiner.
6
Conclusions
It was observed in Bridgman tests that the grain refinement is less effective with higher temperature gradient and it is necessary to use greater amounts of refiners than in the conventional TP-1 tests. The formation of short columnar and elongated grains is due to the
159 lack of sufficient nucleants and a high temperature gradient. The microstructural observations indicate that the columnar/equiaxed transition (CET) occurs continuously in Al- 4.15 wt.% Mg alloys, in agreement with the modeled microstructures. A calculated CET map based on CAFE modeling shows that the equiaxed grains do not become finer in processing conditions close to the transition line and afterward their shapes become elongated. This map shows good consistency with experimental observations at 2 and 5 addition levels of refiner.
7
Acknowledgements
This work was supported by the Engineering and Physical Sciences Research Council (UK). Alcan International Inc.-Banbury laboratories is thanked for provision of experimental facilities, and technical assistance of Mr. J. Worth is acknowledged. The authors acknowledge helpful discussions with project partners: Dr P. Schumacher and Dr K. A. Q. O’Reilly (University of Oxford); Dr R. G. Hamerton and Dr M. W. Meredith (Alcan International Inc.); Dr P. S. Cooper and Dr A. Hardman (London and Scandinavian Metallurgical Co.); and A. Tronche (University of Cambridge).
8
References
[1] Maxwell, A. Hellawell, Acta Metall., 1975, 23, 229 - 237. [2] M. Bunn, P. V. Evans, D. J. Bristow, Light Metals 1998, (Ed.: B. Welch), TMS, Warrendale, 1998, 963 - 969. [3] L. Greer, A.M. Bunn, P.V. Evans, D.J. Bristow, Acta Mater., 2000, 48, 2823 - 2835. [4] M. Rappaz, Ch. Charbon, R. Sasikumar, Acta Metall. Mater., 1994, 42, 2365 - 2374. [5] J. A. Spittle, S. G. R. Brown, Acta Metall., 1989, 37, 1803 - 1810. [6] M. Rappaz, Ch.-A. Gandin, Acta Metall. Mater., 1993, 41, 345 - 360. [7] Ch.-A. Gandin, M. Rappaz, Acta Metall. Mater., 1994, 42, 2233 - 2246. [8] Ch.-A. Gandin, Ch. Charbon, M. Rappaz, ISIJ Int., 1995, 35, 651 - 657. [9] L. Nastac, D. M. Stefanescu, Model. Mater. Sci. Eng., 1997, 5, 391 - 420. [10] D. J. Jarvis, S. G. R. Brown, J. A. Spittle, Light Metals 2000, (Ed.: R. D. Peterson), TMS, Warrendale, 2000, 603 - 608. [11] Calcom SA, Lausanne, Switzerland. [12] M. W. Meredith, Ph.D. Thesis, University of Cambridge, 1999. [13] T. B. Massalski, Binary Alloy Phase Diagrams, ASM International 1990, Vol. 1. [14] R. Trivedi, W. Kurz, Int. Mater. Rev., 1994, 39, 49 - 73. [15] J. D. Hunt, Mater. Sci. Eng., 1985, 65, 75 - 83.
Modelling – Stress and Structure
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
ContiSim™: Process and Material Modelling of Continuous Casting in Macro and Micro Scale J.R. Boehmer Process Modelling and Informatics, Betzdorf/Sieg, Germany
1
Background and Focus
During the last two decades, mathematical modelling has grown to a powerful tool for both process and product research in basic materials industries. The overall goal is to get a quantitative description of the process influencing the development of final structure and properties of the products. Today two trends emerge: integrated modelling of the whole process route, and the combination of material micro and macro models. ContiSim™ is a systematically modular software built on more than 16 years of experience in the field of modelling of continuous casting processes [1]. It is focused on the coupled computation of solidification, cooling, phase change, contraction and stress evolution during the continuous casting of metals, the material bandwidth is from nonferrous metals like copper alloys to steel and cast iron. It allows to describe and to check existent or planned configurations virtually, to optimize operating conditions by parameter investigation, to determine crucial process states to develop measures for safe production, and to predict (and to control) interim and final product qualities. The special quality of computer simulation is to visualize even not measurable inner states and product qualities as well as to make understandable temporal sequences. The vision behind the design of ContiSim™ is to provide an open software with defined interfaces, and with modular program components sharing common resources, being extendable and modifiable economically. The basic equipment is populated with a library of models for different plant configurations and casting conditions, strand dimensions (billet, tube, bloom, slab, thin slab), model requirements (geometrically 1D, 2D or 3D, for steadystate or transient case), and for different material specifications, Figure 1. To combine the two modelling trends mentioned, the programs are designed and implemented modularly according to object-oriented principles. Emphasis is on easy transfer of data between applications, and on easy integration of alternative models. In the same way, the open framework structure allows an integration into holistic (through-process) models as well as networking with other program systems.
2
The ContiSim™ Simulation Kernel
In physical respect, continuous casting is a spatially heterogeneous and time-controlled high temperature process with phase transformations, nonlinear material characteristics, transient load development and transient and heterogeneous boundary conditions, partially a contact problem. The inner state in the solidifying strand and in the mould is of interest, being, for the Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
164 most part, inaccessible to measurements. In mathematical respect the process is a nonlinear thermal-mechanical-metallurgical coupled transient field problem with superimposed transport movement. Its quantitative representation necessitates iterative solution and a computation grid adaptable to solidification progress. The state variables are to be passed on in time, regardless of the kind of grid used.
Figure 1: Structure of the ContiSim™ concept
The model equations of ContiSim™ process temperature-dependent and deformation /-ratedependent material properties for different phase kinds of the material. The specific cooling history during the production process determines development and transition of any phase kind, and therefore the local structure and quality data which again come into the model equations. Primary cooling and solidification are to be explained as a result of the thermal contact in the mould which depends on the cooling-dependent thermal contraction of the strand shell and the inner geometry of the mould. This requires a model that both includes phase change and heat flux/transfer phenomena, “thermal model“ (1), and the thermal contraction and thermo-mechanical load history of the solidified, “mechanical model“ (2). If hot-cracking tendency is to be assessed, dependent on the evolution of the casting stresses during the process, the state-dependent load capacity must be known locally, and compared to the thermoplastic strain at any time. A formally linearized characterisation gives a coupled system of equations: (1) C ( T ) ⋅ T& + A ( T ) ⋅ T ( x, t ) = Q& ( T , T U , δ ) ,
(
)
K σ t , δ t , σ t −∆t , δ t −∆t , δ& t , T , T& ⋅ δ t = F ( T , T& , σ t −∆t , δ t , δ t −∆t , δ U ) ,
(2)
where x denotes the space coordinates, t the time, T the temperature, Q the surface heat flux, F the load vector, δ the displacement vector, σ the direction components of the material tensions. Via the temperature T and its local temporal gradient T and via the material deformation δ there are formal links between thermal (1) and mechanical (2) model. Via
165
temporal fineness (casting time / temporal resolution)
ambient temperature TU and via the deformations of the machine components δU there are links between strand model and surroundings models. The system of equations is in fact in high degree nonlinear by nonlinear functions of the thermophysical properties, phase transformation, nonlinear functions of the heat transfer coefficients, e.g., for mould and for direct water cooling, and by the nonlinear thermoplastic deformation behavior of the solidified. In the transient models, the contiprocess is influenced by temporal modifications of the casting conditions (e.g., melt supply temperatures) or by changes of process parameters by intervention or malfunction.
arrangement of items varies with focus of consideration
mould oscillation grain structure, texture formation
strand deformation, evolution of casting stresses solidification rates
spherical / tapered mould geometry, friction start-up and transient operation
cooling zone design sector heat flux, energy balance, overall solidification times
local spray intensities mould wall cooling design
steady-state conditions
geometrical fineness (strand length / element resolution)
Figure 2: Typical requirements and sensitivity of selected problems on model fineness
3
Levels of Model Requirements
Mathematical models of the past considered more or less isolated phenomena of continuous casting. Among these were on macroscopic level the estimation of solidification times or the design of secondary cooling under purely thermal aspects, or the physics of mould oscillation. More sophisticated models concern combinations of physical aspects with about the same geometry and/or time scale, for example, mould cooling and strand contraction or the evaluation of residual casting stresses. If today, in addition to high-productivity solidification and cooling, the adjustment of defined material properties on both macroscopic level (tensions, deformations) and microscopic level (phase distribution, crystalline structure) is objective, this requires an integrated modelling of nothing less than the entire casting plant and its operation, of the cooling conditions, of solidification, of phase and texture evolution, and the resulting material properties, of contraction and stresses. This covers several orders of magnitude in both geometrical and temporal dimension. Figure 2 shows the diversity of requirements on geometrical and temporal fineness of different aspects to be modelled. The arrangement of items there is due to modelling goals. Therefore, it can not be considered firm but varies with the modelling task. Also schematically in Figure 2, arrows characterize the direction to be expected for model refinements.
166 Today, an efficient full model is expected to break the isolated view of continuous casting. Further demands are to consider larger strand lengths without accuracy loss in the geometrical model, to extend model focus in or beyond the casting plant, to simulate longer casting times, to implement higher geometrical resolutions by element refinements and/or to guarantee finer temporal resolutions, up to the resolution of mould oscillation. ContiSim™ makes an attempt to take this into account by flexible scale-ability, optimized model representation and efficient calculation algorithms. It covers correlated problems of different time scale (such as mould oscillation, strand contact and heat transfer), of different length scale (such as design dimensions and local length scales), and the correlation of the micromechanical changes in time-temperature dependent material properties with the macroscopic approach of heat flux and structural constitution. Figure 3 shows the positioning of ContiSim™ both in cooperation with compatible submodels of different orders of magnitude and in interplay with superordinate / holistic production models.
models of further production steps
process route coupling
10 +2
...
continuum level
10 0 . . .
10 -2 . . .
phase level
open interface
1
transformation and development models
2
caster level
state quantities and boundary conditions
2 material data update and effects on macroscopic state variables
TM framework CONTISIM : interface plant, process and product models
process chain
1 macroscopic
microstructure
10 -3 . . .
10 -4 . . .
crystal level
dislocation level
10 -6 . . .
10 -8 . . .
atomic level 10 -10
problem size in m
Figure 3: Positioning of ContiSim™ in the pattern of orders of magnitude of mathematical-physical examination and in conjunction with models of further processing in a framework of holistic production line models
4
Model Linking and Interfacing
The scale problem shall be explained by two examples: (1) Provided that information is not watered-down by aggregation downstream the process route, a geometrical model fineness of 1 cm of resolution requires for a typical continuous-casting plant for copper about 100,000 computation elements, for a steel plant maybe 3 million, the time increment being in the range of some seconds. If implications of mould oscillation are of interest, the geometrical raster is to be refined to the millimeter range, and the temporal resolution to some hundredths of a second. Outside the mould, strand information should only be clustered without significant
167 quality loss, and oriented at physical aspects, e.g., using increments of the size of oscillation mark spacing or fractions of it. (2) Only if considering material factors, comprehensive statements about product quality at the beginning of the next processing stage can be made. While macro-modelling of thermodynamics and non-linear mechanics are subject of choice for process-relevant aspects, grain texture descriptions are terrain of microstructure models of metal physics. Between both model worlds, the gap of the geometrical order of magnitude exists. In ContiSim™ it is attempted to overcome the problem of "scale-bridging" by a phenomenological approach. The thermodynamic-mechanical macro-model provides the backbone of the approach, but considering non-homogeneous property profiles in model calculation. To approximate the governing partial differential equation system, discrete model equations are solved simultaneously for small elements constituting the considered strand volume. For each element and time the temperature and the tension/deformation state are determined as well as corresponding gradients in the neighborhood. Temperature-dependent material properties (including, if necessary, additional pre-deformation and deformation-rate dependent strength data) are considered iteratively in the calculation equations. The calculation grid structure employed anyhow is used as basis for the phenomenological approach: For each element a phase kind index decides on the material properties set to be valid. The specific cooling history during the production process determines the structure and property data of an element. The modular programming and interfacing concept of the ContiSim™ program family allows an easy integration of both physical and empirical-phenomenological transition models (transformation equations, TTT diagrams and characteristic maps or mathematicalphysical microstructure models). The actual volume fractions are computed and taken into account for each subsequent thermal-mechanical iteration via phase kind index and property profile. geometry
time
Figure 4: Problem diversification by different screen definitions and physically founded feedback
The procedure of combining model worlds is outlined in Figure 4. When required, aspects are refined and the results are transferred back via physically founded evolution models for
168 the next iteration step. By this, compared to the certainty of the available data, a sufficiently precise projection appears to be guaranteed.
5
Aspects of Practical Use
A platform with the claim to be a comprehensive model of continuous casting like ContiSim™, is exposed to be assessed complicated and therefore to be less manageable. Actually, the user himself determines the desired degree of complexity and the extent of the required data, i.e., information on plant geometry, process data, on the material properties, and the micromodels used. It is also a fact that the entirety of data possible for model calculation is not always available, or that in other cases the full model potential may not be used for a preliminary estimation. This is taken into account by different levels of detail. In such a way, a relatively compact minimum database is required to perform a first model calculation (e.g., with constant material properties and/or relatively coarse modelling of the plant). If extended model depth is desired, further data may be necessary. The data structure is described in the user's guide and in digital help. If the interactive data input interface is used, a model data assistant directs to input only data relevant for the desired calculation. Already during the input phase all data is subject to a knowledge-based plausibility and compatibility check. With it, typical application mistakes can be avoided. Where applicable, default values are offered for input. Beyond the plausibility check, a material data assistant offers the possibility for inter- and/or extrapolation of incomplete data. The model user should be relieved from the organization of parameters of mathematical model calculation. This is the task of a numerics assistant. It recommends problem-oriented suitable discretisation parameters and suitable values of other computation parameters for confirmation or modification. To assist input control, data can be prepared graphically and visualized when needed. In this way, for Windows or NT platforms, DLL submodel referencing, GUI-based, menudriven user levels and assistant programs accompany the modelling path of data acquisition, plant, process and material modelling to model computation up to purpose-oriented result analysis, competent visualization and documentation. It is completed by links with experimental data, HTML-based instruction manual as well as E-mail support.
6
Reference
[1] J.R. Boehmer, Methodik computergestützter Prozeßmodellierung - dargestellt am Beispiel der Modellbildung für kontinuierliche Erstarrungsprozesse, Oldenbourg, München, 1997.
Crystal Growth Morphology during Continuous Casting Christoph Caesar Munich
1
History of Solidification Models
Continuous casting processes are known at least since 1846, as Bessemer patented his double roller continuous caster. Dated 1908, the first single roller caster was patented. Very much effort has been put into the modeling of these continuous solidification processes. It is, however, astonishing that not any geometrical model of the solid growing into the melt clearly states a growth vector for the different positions of the melt pool. A growth vector can be defined for any crystallization velocity of a planar growth front. The vector is perpendicular to this front thus indicating magnitude and direction of the crystal growth. In steady state condition (continuous process) this vector will certainly be correlated with the feed or casting velocity via an angle and a simple trigononmetric relation.
a)
b) Figure 1: a) Historic view of a laser melt pool solidification front (taken from [21]). b) Velocity V for a horizontally thrown ball; Vw = horizontal velocity, Vs = vertical velocity, [22]
Schematic diagrams like the vector geometry assessment of Fig. 1 a have been published occasionally for various continuous solidification processes as melt spinning or laser surface melting. But there is only one way of resolving a vector as the sum of a known and constant velocity component (as that of a casting wheel) in x-direction and another orthogonal y - direction (e.g. Fig. 1 b). In the example, Vw stays constant, as Vs (or Vy) grows with time. The resulting velocity vector V always is greater than the initial velocity and the movement of the ball obeys this relation as well as a crystal growth front - see eq. 1. Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
170 V = Vs / cos (α) and V = Vw / cos (90 - α) (1) Why has the basic vector geometry of Fig. 1b - taken from an engineering physics book not been used for modeling the melt pool - e.g. in a twin roller or a belt caster - in the past years? The answer is that the approach leads to growth velocities much higher than the casting wheel surface speeds for - assumed or real - flat angles between the wheel surface and the growth front. This seems improbable at first view. It shall be shown in the following that there are no contradictions between crystallization velocities calculated by the approach above and the observed microstructures even for a rapid single roller casting process.
2
Approach and Experimental Methods
Simple model alloys from the completely miscible Copper - Nickel system were selected for this work with additions of Boron or Zirconium to stabilize the as-formed microstructures [4]. All tests were performed with material cut from the same plates of alloy melted and cast under Argon in an induction RF - furnace. The details are summarized in [2, 3]. Droplets of 5 to 6 mm dia. of CuNi30 were dynamically undercooled in a container-less electromagnetic levitation unit [6, 8] and solidified rapidly after nucleation. Temperature and crystallization velocity could be measured accurately, as the crystallization front passed across the surface of the specimen. The same alloy was cast on a single roller caster from 5 m/s to 40 m/s roller surface velocity (see [4]). Single roller casting will be regarded as one half of the symmetric double belt or double roller process in a first approximation. The advantage of the single roller process is that the cast tapes reveal a free surface like the bulk undercooled droplets, which gives valuable information on the solidification history, In glass forming alloys, the higher casting velocities are sufficient to undercool the melt so far that it stays amorphous even at room temperature. In most Aluminum-, Copper- or Nickel - alloys, a segregation-free solidification and a significant increase in solid solubility of alloying elements can generally be observed.
3
Results and Discussion
By dynamic undercooling, crystallization velocities up to 40 m/s could be measured in CuNi30 [6, 8] for a degree of undercooling ∆ T- of 265 K; in the first reference, nearly 80 m/s have been measured for pure Nickel. A square growth law describes the crystallization velocity of the fcc crystal at degrees of undercooling between 0 and 190 K. At higher undercooling, a linear relation is found. The growth law has been established earlier in [2,3] and is attributed to the primary crystallization of a metastable phase (most likely bcc, tetragonal or (Ortho-)rhombic acc. to [19]). Figures 2 a and b show two free surfaces - the free surface of a tape cast at 10 m/s and one solidified after an undercooling of 197 K with 20 m/s measured growth velocity (a).
171 The structures are identical and it is stipulated that undercooling and crystallization history had been similar. Fig. 2 c and d give a comparison of the free surfaces of a droplet undercooled by 265 K (crystallized with a measured growth velocity of 36 m/s) and a tape cast at 30 m/s. The surface of the tape reveals an identical microstructure on a much finer scale, indicating an even higher undercooling prior to crystallization than after having reached an undercooling degree of 265 K.
Figure 2: a. Surface of a CuNi30 tape cast at 10 m/s. b. Free surface of a CuNi30 droplet undercooled by 197 K (measured growth velocity of 20 m/s). c. Free surface of a CuNi30 droplet undercooled by 265 K (measured growth velocity 36 m/s). d. Surface of a CuNi30 tape cast at 30 m/s with very fine pearl - like structure (each dash = 10 µm)
Something similar had been found by TEM within the bulk of the tape as spherical segregation traces being independently from a recrystallized grain structure [3, 4]. Identical microstructures had been presented by Müller and Löhberg [23] as early as 1969 in splat cooled CuNi30. Microstructures [8,3] and surfaces of rapidly cast tapes therefore indicate a high degree of undercooling prior to solidification, which can directly be correlated with measured growth velocities of 15 to over 36 m/s. It was established earlier in [3] that the fraction solidified in the first step of crystallization (visible in the pearls on the surfaces) and in the bulk metal corresponds to a degree of undercooling of approximately 320 K for the tape in Fig. 2 d. A crystallization velocity can then be derived from the linear growth law given in equation 2. For ∆T = 320 K, CuNi30 primarily crystallizes with a linear growth velocity of V = 0.129 m/(s K)*∆T ([3], derived from [6]) = 41 m/s (2) Here, a velocity higher than the wheel surface velocity has been determined for the tape cast at 30 m/s. Often discussed are the curved columnar crystals found in melt spun alloys and in pure metals (see Fig. 3 b), which are likely to mislead the observer. As the crystals are actually inclined away from the melt pool, they cannot result from a growth into the melt. No heat transfer coefficient will be sufficient to solidify the melt beneath the melt pool [25]. By many authors, this inclination is attributed to a solute wash effect. However, the same inclination is observed in pure metals like copper or aluminum [9, 10], where the constitutional undercooling should be no major factor.
172 Fig. 3 shows the principles of the solidification of undercooled and of melt-spun CuNi30 as a summary of the observations described above.
b)
a)
c)
d)
Figure 3: a) Microstructure of sample undercooled 265 K (measured growth rate 36 m/s). b) Longitudinal section of melt spun strip (30 m/s) with fine shading parallel to the drawn in arrow . c) pattern of the solidification front lines (see Colligan and Bayles [13]). d) model of the solidification front geometry during tape casting with stationary growth direction Ug (V)
The fast solidification front penetrating the undercooled droplet leaves a metastable primary structure, which recrystallizes to the observed fine-grained microstructure. Analogous to this solidification the front in a rapid continuous caster stationarily moves into the undercooled melt at local conditions nearly identical to those in the bulk undercooled samples. The columnar crystals then are the result of the recrystallization of this metastable structure. These relations are given in Fig. 3 d. The genuine crystallization direction is seen as an angle of the fine shading of the crystals in fig. 1 b (see also Figs. 4, 5) of 40°. The vector resolution and application of equation 1 gives for the growth velocity V or Ug V = Ug = (30 m/s / cos (40°) = 39 m/s This is in very good agreement with the velocity of 41 m/s calculated from a completely different approach from the degree of undercooling (equation 2). Conventional casters - e.g. a Hazelett belt caster with up to 10 m/min casting velocity (0,166 m/s) may have a melt-solid interface under an angle of approximately 2 degrees to the belt surface [24], which equals to an angle α of the growth vector of 88 degrees to the surface. With the equation above, this corresponds to a growth velocity of 4.75 m/s. This velocity would be achieved at a degree of local undercooling of only 50 to 60 K in the CuNi30 - alloy - which is to be expected in the vicinity of an intensively water film cooled steel belt. The alloys Cu-B0,34 and Cu-Zr 0,17 (w/o) were also cast at velocities of 5 to 30 m/s. The effect of "freezing" an originally formed microstructure by these alloying elements can be seen in Figs. 4 and 5. The fine primary crystallization traces form an angle α with the casting direction of 45 °, which results in a growth velocity of 42 m/s in Fig. 4. Copper-Boron is known as a eutectic composite material; the primary dendrites of the longitudinal section of Cu-B 0.34, cast at 5 m/s [4] have grown at ca. 5.8 m/s primary dendrite growth velocity under an angle of 30 ° to the casting direction (Fig. 5).
173
Figure 4: Rapidly cast double tape, Cu-Zr 0.17; 30 m/s wheel velocity, longitudinal section. Fine lines pass visible "fine grained" crystal structure (direction of solidification)
Figure 5: Rapidly cast tape, Cu-B 0.34; 5 m/s wheel velocity, longitudinal section. Primary dendrites cross the visible columnar crystal structure under an angle of 30 degrees, same magnification as above
Microstructures with fine traces of the original solidification like in Fig. 4 are also seen in other works, e.g. by Tenwick [9, his Fig. 4 center]. They are also regarded as proof that the model of Fig. 3 d is correct.
4
Conclusion
A unified model of crystallization front geometry in the melt pool of a single roller caster is presented, which correlates microstructures with the degree of undercooling, the cooling wheel velocity and the actual crystal growth direction. A very good analytical and mathematical match is found for growth velocities calculated from surface features related to undercooling and from observed growth direction angles and wheel speed. The results can be useful in modeling the solidification in continuous casting in general, as they supply a known variable in form of independently determined growth laws for a local melt undercooling.
174
5
References
[1] R.B. Pond, US patents 2,825,108; 2879,566; 2,900,708; 2.976.590 Metallic filaments and method of making same [2] C. Caesar, in: A. Ludwig (Ed.) "Erstarrung metallischer Schmelzen in Forschung und Gießereipraxis", Wiley-VCH, Weinheim 1999, 43 [3] C. Caesar, Advanced Engineering Materials 1999, 1, p. 75 ff [4] C. Caesar, "Gefüge, Anlaßverhalten und mechanische Eigenschaften schnell erstarrter Kupferlegierungen", VDI -Verlag Fortschr.-Ber. Series 5 no. 172, Düsseldorf 1989 [5] C. Caesar, Patent DE 38 06 451 C2 "Verfahren zur kontinuierlichen Herstellung von langgestreckten Formkörpern und Halbzeugen aus der Schmelze" 1988 [6] E. Schleip, R. Willnecker, D. M. Herlach, G. P. Goerler, RQ6, Mat Sci. Eng. 1988, 98 ,39 [7] B. Cantor, in P. R. Sahm, H. Jones, C M. Adam, Science and Technology of the Undercooled Melt, Dordrecht, Boston, Lancaster 1986, 3 [8] C. Caesar, U. Koester, R. Willnecker, D. M. Herlach, RQ6, Montreal, August 1987, "Materials Science and Engineering"; Elsevier 1988, 339 [9] M. J. Tenwick, H. A. Davies, Mat. Sci. Eng. Lett. 1984, 63 L1; (see RQ5, p 67) [10] M. van Rooyen, N.M. v.d.Pers, L.Katgerman, T.H.de Keijser, E.J. Mittemeijer, RQ5 823 [11] R. Ichigawa, H. Taniguchi, O. Asai, Z. Metallkde.1980, 71 260 [12] T.Z. Kattamis, S. Skolianos, RQ5 1985, 51 [13] G.A.Colligan, B.J. Bayles, Acta Met., 1962, 10 ,895 [14] W. Kurz, B. Giovanola, R. Trivedi, Acta Met. 1986, 34 ,823 [15] T. Suzuki, S. Toyoda, T. Umeda, Y. Kimura, J. Crystal Growth 1977, 38 ,123 [16] K. Kobayashi, L. M. Hogan, Metals forum 1 1978, 165 [17] S. Eucken "Schmelzspinnen von Legierungen mit Formgedächtnis" PhD-Thesis.; Bochum 1986 [18] T. W. Clyne, Met. Trans. B 1984, 15B, 369 [19] H.H. Liebermann, Bye, Jr. in: S. K. Das, B.H. Kear, C.M. Adam (Eds.): Rapidly Solidified Crystalline Alloys Proc. Morristown, TMS-Publ., Warrendale, Pa. 1985, 61 [20] F. Fayard, F. Duflos, A. Lasalmonie, RQ5 1985, 811 [21] W. Kurz, Advanced Engineering Materials 2000, 2, p. 295 ff [22] Dobrinski, Krakau, Vogel, "Physik für Ingenieure", Teubner, Stuttgart 1974,p. 14 [23] K. Löhberg, H. Müller, Z. Metallkde. 60 (1969) p. 231 [24] D. Altenpohl, Z. Metallkde. 60 (1969) H 9, p. 678 [25] P. H. Shingu, RQ4, 1 (1982) 57
3D-Modeling of Ingot Geometry Development of DC-Cast Aluminum Ingots during the Start-Up Phase Werner Droste1, Jean-Marie Drezet2, Gerd-Ulrich Grün1, Wolfgang Schneider1 1
VAW aluminium AG, Research and Development P.O. Box 2468, D-53014 Bonn, Germany CALCOM SA, Parc Scientifique, PSE-EPFL, CH-1015 Lausanne, Switzerland
2
1
Abstract
Constant demands on scrap reduction and increased process performance in aluminum DCcasting require surfaces which need less scalping and start-up procedures which diminish butt sawing and decrease cracking tendencies. During the last years, there has been a big progress in thermomechanical simulation of aluminum DC-casting with focus on ingot shape and butt curl. The aluminum industry is on the way of transferring this more scientific method to a practical tool for calculating mold shapes, minimizing butt scrap and optimizing start-up procedures. In the first section of this paper, a brief description of the model and its implementation are given. The 3-D model is based on the general purpose software Abaqus and uses a thermomechanical approach to describe the process taking into account the viscoplastic behaviour. It delivers the contraction of the rolling faces, the butt curl during the transient casting phase and the final ingot geometry. Also, strain rate and stress field at the end of casting are calculated. For evaluation and adaptation of the model to specific boundary conditions in a cast house, a production mold and bottom block system is chosen in order to compare calculations and corresponding casting trials on measured temperatures, curls and ingot geometries, which are finally presented.
2
Introduction
During the last years, there have been joint activities of the European aluminum producers on developing tools for modeling aluminum DC-casting. A major part of this work was funded by the European Commission and is known as the EMPACT-program. Now these tools are going to be adapted by the aluminum companies for internal use. At VAW aluminium AG, this is carried out by integrating a 3D-thermomechanical model into their toolbox, which is based on the development work done first at Alusuisse [1] and later on within the EMPACTprogram [2]. In this paper, a standard hot-top casting of 1650x600mm sized rolling ingots of an AA1050 alloy is simulated. The calculated results are compared with measured data coming from corresponding full size experiments. Special focus of the study is put on the transient start-up phase including butt curl and ingot geometry development.
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
176
3
Model Description
The 3D-model is realized with the general purpose software Abaqus. A detailed description can be found in [1-5], but a brief presentation of the concept will follow. Based on a finite element formulation to compute the thermomechanical development within the solidifying ingot, the model can calculate temperatures, stresses and stress induced deformation. The stress-strain dependency is handled by a viscoplastic material model within Abaqus using a rate dependent plasticity analysis. The model allows to predict the butt curl as well as the rolling faces pull in. Fluid flow influence on temperatures is taken into account by an increased heat conductivity in the molten state. The computational domain is limited to one quarter of an ingot because of assumed symmetry. The design of the mould and the enmeshment of the ingot and the bottom block are given in Figure 1. In the present example, 96 nodes per layer are used and with a thickness of 30 mm per layer and 30 layers, a total cast length of 0.9 meters is simulated. The rather coarse enmeshment is a compromise to get a stable mathematical solution and an acceptable calculation time. Due to the bending phenomenon during butt curl, the elements in the related regions experience a large vertical displacement. Using smaller elements might cause severely distorted elements, which would stop the calculation due to numerical errors.
Y
330 mm 306 mm
842 mm
X
Figure 1: Mold dimensions and enmeshment of ingot and bottom block
According to the real conditions, a bowl shaped bottom block is used. Its outer rim has a height of 120 mm and follows the contour of the mould. The bottom block thickness is 180 mm , while the rim thickness decreases from 60 mm to 20 mm from bottom to top. The material for the trials and for the simulation is an AA1050 alloy. The thermophysical and thermomechanical properties are taken mainly from the results of the EMPACT-program [2,5]. The bottom block material is an alloy of the 6000 series and it is assumed that it has constant material properties and that the bottom block will not deform at all (rigid body). The hot top, because it is considered as an insulation with adiabatic characteristics, will not influence the rolling faces pull in. Therefore, the whole region including the liquid melt is excluded in the present model. Although, the transient start-up phase of casting is studied here, only a simplified casting recipe is applied. The liquid level in the mould is considered as constant and the filling phase starts with that melt level. Further on, a constant casting speed of 60 mm/min is chosen.
177
Z hot top
75 mm
mold
50 mm
liquid level
Zo
30 mm 150 mm
cast metal
water cooling
Z = 0.
bottom block
Figure 2: Initial configuration of the calculation
The initial situation is depicted in Figure 2. Five layers of metal representing a height of 150 mm are present in the bottom block at the start of the calculation. Note that the filling phase is not modeled with the present model: owing to the large convection that takes place in the liquid pool, the temperature remains rather uniform except on the outer surface of the ingot where heat transfer to the bottom block and to the mould takes place. The computation starts at the very moment when the casting table begins its descent, assuming a constant casting speed of 60 mm/min. The initial temperature field is uniform at 670°C except for the surface of the ingot which is assumed to be at 600°C. Because Abaqus uses a Lagrangian description, metal and bottom block positions are fixed within the solution domain and the boundary conditions move upwards at casting speed on the external faces of the ingot. The top surface of the domain is adiabatic and the temperature of each new layer is set to the temperature of the incoming metal, 670 °C, which is a reasonable choice with a temperature of about 700 °C in the launder. According to experience, the mold length of 50 mm is subdivided into a primary cooling zone of 30 mm and a related air gap of 20 mm. All thermal boundary conditions on the external surfaces are defined as a function of the liquid metal level z and they are split in three main areas, in which different cooling mechanisms are considered: • the primary cooling (contact with the mould) with Φ = h (T - Tmould) for z = (z0 - 0.030) • the air gap with Φ = 0 for (z0 - 0.050) ≤ z ≤ (z0 - 0.030) • secondary cooling with Φ = reduct(time)*hwater(T)*(T-Twater) for z = ( z0 - 0.05) where hwater is a surface temperature-dependent heat transfer coefficient and the reduct(time)-function is a coefficient to simulate the reduced amount of cooling water during start-up, which rises from 32.4 m³/h to a steady state value of 45.9 m³/h. A more detailed description about the determination of the heat flow and the heat transfer coefficients is given in [6]. Starting with an initial temperature of 25 °C the bottom block is cooled down on its vertical sides with a constant heat transfer coefficient and a water temperature of 25 °C. On its bottom side, an air cooling is applied using a much lower heat transfer coefficient.
178 It is known from experience that the water intrusion into the gap between ingot and bottom block can has an influence on the butt curl. Therefore, a strategy, which simply increases the gap conductance, when water starts to intrude, is partially wrong since the intruding water represents rather a heat sink for both, the bottom block and the ingot. Models, which do not account for that, deliver butt curl results, which are too small compared to measured data. One intention of this work is to improve the agreement with measured butt curl. In order to do so, a more realistic model of the water intrusion is tested: 1.)
remove the contact elements in the interface, where water intrudes and
2.) apply instead a special boundary condition on both sides of the interface, i.e. on the ingot and on the bottom block, assuming a constant heat transfer coefficient hintrusion and a constant water temperature.
Figure 3: Top surface elements of the bottom block showing the area, where water intrusion is modeled (elements marked dark)
The water intrusion takes place at 60 s after the start of withdrawing, i.e. about 10 s after the water of the secondary cooling system starts to impinge the ingot surface (time 50 s). The dark element surfaces in Figure 3 show the area wetted by the water and where the special conditions are applied. This zone extends more or less down to the drainage holes, remaining the central area dry (white area).
4
Casting Experiments
The experiments were carried out using a VAW aluminium AG hot top mold system for the mentioned ingot size of 1650x600mm. Temperature in the launder and the casting temperature had the values described above. The melt distribution inside the solidifying ingot was done with a standard Combo-bag. Initially, the upper rim of the bottom block was placed about 10mm below the hot top. Water flow amount was rising during the start-up as given above, the water temperature was 23 °C. After filling the mold up to the hot top within about 4 min, the casting table was lowered first with a speed slightly slower than 62 mm/min. Entering the direct cooling zone the speed was increased slightly above the final casting speed until the final casting speed was adjusted. During the trial casts butt curl and butt temperature development was recorded. Additionally, a detailed measurement of the butt geometry followed after the casting. The curl measurements were carried out in the middle of both small sides of the ingot using displacement transducers connected to the ingot via a 0.8mm steel rods (Figure 4).
179 The development of the curl displayed in the following figures is the mean of the measured left and right value. The grid of thermocouples was mounted inside the bottom block to measure the temperatures at a depth of about 10 mm within the ingot butt all along the x- and the y-axis and within one quarter of the ingot. In this paper only the temperatures measured along the x-axis at distances x = 0, 150, 350, 450 and 600mm are presented, because they will be used in the comparison with the calculations. The ingot geometry scan was done at cast lengths of 50, 150, 250, 350, 650, and 1550 mm.
Figure 4: Butt curl measurement equipment
5
Results
Figure 5 shows the comparison of the time-dependent vertical displacement of the mesh point situated at the middle of the short face of the simulated ingot (named calculated curl) with the results of a representative butt curl measurement from experiments. Additionally, the time derivative of both curves are given, which represent the related curling speed. Note that with the assumption of a rigid bottom block, the vertical displacement of the mesh corresponds to the measured butt-curl. Moreover, since the filling phase is not modeled, the time scale used for the computation has been shifted by the filling time so that the results can be compared with the measurements. max. 158 mm/min
70
800
Butt Curl in [mm] - Curling Speed in [mm/min]
60 50 40
3 2
1.) Start Filling 2.) Start Withdrawing (Start of Calculation) 3.) Start Water Intrusion
700 600 500
1
30
400
20
300
10
200
0
100
-10 0
100
200
300
400
500
600
Time in [s]
Figure 5: Measured and calculated butt curl and curling speed
700
0 800
Cast Length in [mm]
Measured Curl Calc. Curl Meas. Curling Speed Calc. Curling Speed Cast Length
180
Ingot Thickness in [mm]
Looking on the resulting butt curl development, two important events can be identified: 1. In the calculation, no curl is observed until the cooling water starts to impinge the ingot surface (time 361 sec in Figure 5). In fact, during that period, the already solidified shell is still hot and weak and therefore the ingot does not deform on the bottom block. 2. The butt curl starts at the time the secondary cooling water hits the ingot surface (time 361 sec). As soon as the water intrusion occurs, the cooling of both, the ingot and the rim of the bottom block, becomes very high. This leads to a separation of the metal from the bottom block and the butt curling starts. The measured data also seem to have some curl right at the beginning of filling though the steel rod for the curl measurement has not yet contact to the aluminum. The explanation for this behavior is the expansion of the bottom block, which starts at the beginning of filling. Due to that expansion the small gap between bottom block and mold wall is closed and the steel rod is pinched in. This causes the small offset error of about 5mm in the curl measurement (also visible at the start of withdrawing, (2) in Figure 5), which has to be subtracted from the measured final value of about 50 mm regarding the comparison with the model results. Comparing the measured and calculated data, one can notice two significant points: 1. the calculated curl (~45mm) compares well with the measurement (~(50-5)mm) 2. the curling speed derived from the measurements and from the computation are both highest right at the beginning of curling. They also compare very well except a too pronounced computed maximum owing to the coarse mesh, which is the reason of the step like substructure in the calculated curling speed Figure 6 shows the calculated and measured rolling face contours of the butt after it has cooled down to room temperature. The design of the mold is also shown. The butt swell is most visible at cast lengths between 50 and 350 mm. Then, the steady state contraction is measured at lengths 650 and 1550 mm. The contraction development in the butt zone is somewhat underestimated by the model by an amount of 10 mm for the ingot width, that is 5 mm on one ingot side. The use of a rather coarse mesh and/or not fully adjusted material properties or boundary conditions can explain this behavior. Also the casting speed in the calculation was constant while in the experiment a slightly higher speed was choosen. Cast length
660
Measured
Calculated
650
Y-mold L=30mm Calc L=480mm Calc L=780mm Calc 50mm Meas 350mm Meas 650 mm Meas 1550mm Meas
640 630 620 610 600 590 -1000
-500
0
500
1000
Ingot Width in [mm]
Figure 6: Measured and calculated rolling face contour at different heights of the ingot
181 Figure 7 shows a three-dimensional view of the simulated temperatures 50 seconds after the start of withdrawal. Water intrusion has not yet occurred. At this time the model shows a continuous partially solidified shell (temperatures lower than 650 °C, the coherency temperature of the alloy) along the rim of the bottom block. This shell is still too weak to deform. Note that the heat transfer to the rim is rather efficient. Beyond this time, the cooling of the ingot and the rim of the bottom block induced by the water intrusion begins to work.
Figure 7: Temperature distribution on the deformed mesh at time =50s
Figure 8 shows the temperature distribution 180 seconds later (time = 50 + 6*30 s). At this time, the cooling induced by the secondary cooling is particularly efficient on the ingot. Now, the continuous shell that has built up all around the liquid pool, gets thicker where the cooling is efficient, that is near the central part of the bottom block (good thermal contact) and near the ingot surface. This shell seems to be necessary to create the stresses and strains, which cause the curling. Note that the cooling induced by the water intrusion has been very efficient for the rim but not for the ingot which remains hot.
Figure 8: Temperature distribution on the deformed mesh at time = 230 s
Figure 9 shows the temperature distribution 620 seconds after the start of withdrawal (time = 50 + 19*30s, cast length = 570 mm). Due to the development of the butt curl, the heat
182 transfer to the bottom block is now limited to a confined area near the center, where the contact remains. Now, the increased cooling due to the water intrusion is obvious: the rigid shell shows remarkably cooler temperatures in all areas stricken by cooling water.
Figure 9: Temperature distribution on the deformed mesh at time = 620 s
700
60
600
50
500
40
400
30
300
20
200
10
100
0
0
Butt Curl in [mm] - Curlin Speed in [mm/min]
Temperatures in [°C
These calculated temperature distributions can be compared with measured temperatures at about 10 mm within the butt, shown in Figure 10.
X=0,Y=0 X=150,Y=0 X=350,Y=0 X=450,Y=0 X=600,Y=0 Länge ms Curl Curling Speed
-10 0
200 400 600 800 1000 1200 1400 Time in [s]
Figure 10: Measured curl and temperatures along the x-axis of the ingot 10 mm inside the butt
The figure also includes the data of the curl measurement, which was done in parallel. The casting conditions with this experiment were exactly the same as with the curl measurement shown in Figure 6. Within the first 240s during filling, a shell has formed, and temperatures reach a minimum of about 600 °C at x = 350mm from the center. As soon as the secondary cooling hits the ingot, the butt curl starts and the temperatures rise within the butt. Only in the center at x = 0mm and 150mm the butt is still cooled by the bottom block. Even in the outer
183 region (x=600mm), the temperatures remain at a level close to the solidus temperature for about 100s after the start of curling. This indicates, that the water intrusion needs a sufficient large gap to start a cooling effect on the bottom side of the ingot. At x = 350mm the effect of the fluid flow out of the Combo-bag is visible. This area stays close to the liquidus - solidus interval till about 700s or 500mm casting length. The general evolution of the thermal field is reproduced in an acceptable manner by the model as can be seen in Figures 7 to 9.
6
Conclusion
With the addition of the water intrusion effect, it is intended to better model the heat flow between ingot and bottom block. The evolution of the temperature field is now better reflected and reproduces the high cooling of ingot and rims induced by the intruding water. Although the filling procedure has not been modeled and a rather coarse enmeshment was used, the computed butt curl compares very favorably with the measurement. As outlined, a continuous weak solidified shell around the ingot is responsible for the butt curl development. Further work will focus on how to reduce the butt curl by notably adjusting the cooling conditions [6,7] and casting recipes and by using different bottom block geometries.
7
References
[1] J.-M. Drezet, M. Rappaz, B. Carrupt, M. Plata, Metall. Trans. B, 26, 1995, 821-829 [2] Brite-Euram Program BE-1112 EMPACT, European Modeling on Aluminum Casting Technologies, 1996-2000, Task 1.1 (thermo-mechanical tests) and Task 1.2 (sensitivity tests). [3] J.-M. Drezet, Ph.D. Thesis 1509, Swiss Federal Institute of Technology (1996) [4] J.-M. Drezet, M. Rappaz, Metall. Trans. A, 27, 1996, 3214-3225 [5] Burghardt, B. Commet, J.-M. Drezet and HG Fjaer, “A numerical study of the influence of geometry and casting conditions on ingot deformation in DC casting of Al. Alloys” in MCWASP IX, TMS, Aachen, August 2000. [6] J.-M. Drezet, M. Rappaz, G.-U. Grün, M.Gremaud, Metall. Trans. A, 31, 2000, 16271634 [7] W. Droste and W. Schneider, "Einfluß der Angießbedingungen auf die [8] Fußgeometrie bei Aluminiumwalzbarren" in E. Lossack, Stranggießen, DGMInformationsgesellschaft Verlag, 1991, 143-158
The Influence of Casting Practice on Stresses and Strains in 6xxx Billets – A Statistical and Modelling Study
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
185
186
187
188
189
190
Modelling of Macrosegregation in Continuous Casting of Aluminium T. Jalanti1, M. Swierkosz2, M. Gremaud2 and M. Rappaz1 1
Ecole Polytechnique Fédérale de Lausanne, Laboratoire de Métallurgie Physique, CH-1015 Lausanne, Switzerland 2 Calcom SA, PSE, Ecublens, CH-1015 Lausanne, Switzerland
1
Abstract
Within the framework of the Brite-Euram research program EMPACT (A European Modelling Programme on Aluminium Casting Technology), macrosegregation in continuously cast Al-Mg slabs has been modelled. For that purpose, the average conservation equations of mass, momentum, heat and solute were derived, including the shrinkage contribution and the transport of the solid. Solvers for these equations were implemented in the Finite Element software calcoMOS. At each time step, the variations of enthalpy and average solute contents were computed by successive solution of the above equations. A microsegregation model then allowed to determine the increments of the volume fraction of solid, temperature, liquid and solid concentrations, and average density. This model encompassed back-diffusion in the primary phase and non-linear phase diagrams. It was found that the negative and positive segregations measured at the centre and surface of DC cast ingots, respectively, could be explained to a fairly large extent by the solidification shrinkage contribution.
2
Introduction
Unlike microsegregation which can be modified by subsequent heat treatments, macrosegregation, i.e., alloy composition heterogeneity at the scale of a whole casting, remains during the various treatments applied to DC cast ingots (rolling, homogenisation, etc.). Therefore, it is essential to control it and simulation is a means of achieving that. Macrosegregation is the result of the coupling of two phenomena : microsegregation at the scale of the dendrites and relative movement between the solid and liquid phases at the scale of the casting [1]. The relative movement between the solid and liquid phases can be induced by : natural or forced convection, solidification shrinkage, sedimentation of equiaxed grains or deformation of the mushy zone [1]. In the present calculations, only the first two mechanisms were taken into account. The model has therefore to solve first the heat flow equation in order to predict the location of the mushy zone and the enthalpy field (see Fig. 1). Second, the movement of the liquid phase is calculated with the help of the momentum and mass conservation equations. Finally, the solution of the average solute concentration equation(s) gives the evolution of the average solute content at any location. These last equations must be coupled with a microsegregation model, which allows to calculate the solidification path at every location. Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
192 The calcoMOS Finite Element (FEM) software [2], which simulates solidification phenomena for two-dimensional (2D) cartesian and axisymmetric geometry, was used to implement macrosegregation computations. This paper briefly presents the model and its application to DC cast aluminium ingots. Initialization of the problem t :=t+∆t Solution of macroscopic conservation equations
Fluid flow and mass conservation
Heat flow equation
vlt+∆t, pt+∆t
t+∆t
Iterations
Solute conservation equation <wi>t+∆t
Calculation of microsegregation
t+∆t Tt+∆t , f lt+∆t , wt+∆t i,l , wi,s , <ρ>
End
Figure 1: Flow chart of the macrosegregation computation
3
Model Description
In the presence of shrinkage and solid transport at a uniform velocity, vs, the average conservation equations for heat, mass and solute species are given by [3]: ∂<ρh> Heat : + div <ρhv> - div (<κ> grad T) = 0 (1) ∂t ∂<ρ> Mass : + div <ρv> = 0 (2) ∂t ∂<ρwi> Solute (i) : + div <ρvwi> = 0 (3) ∂t where <ξ> denotes the average over the liquid and solid phases of the field ξ, i.e., <ξ> = ξsgs + ξl gl, where ξs and ξl are the values in the solid and liquid phases, gs and gl being the volume fractions of solid of liquid, respectively. ρ is the specific mass, h is the enthalpy per unit mass, v is the velocity, κ is the thermal conductivity, T is the temperature and wi is the mass fraction of solute species (i). For this last entity, the local averaging of the solute concentration, <ρwi>, is slightly different if microsegregation does not occur at equilibrium (i.e., not according to lever rule) :
193 gs
<ρwi> = ρlwi,lgl + ⌠ ⌡ ρswi,s(g,t)dg , i = 1, N
(4)
0
where N is the number of solute species. The momentum conservation equation is usually written only for the liquid phase. In the case of global transport of the solid, such as that encountered in DC casting, this can be done for the relative velocity of the fluid u = (v - vs) = (vsgs + vlgl) - vs = gl(vl – vs). One has : µgl ∂u ρl ρl + g (Grad u) ⋅ u - µ ∆u + K u + gl grad p = glρbg (5) ∂t l µ is the dynamic viscosity of the fluid, K the permeability of the mushy zone and ρb the Boussinesq approximation of the density of the fluid : ρb = ρl,o [1 - βT(T - To) - ∑ βi(wi,l - wi,l,o)] (6) i
ρl,o is a nominal specific mass of the liquid taken at some reference temperature, To, and concentrations in the liquid, wi,l,o. βT and βi are the thermal and solutal expansion coefficients, respectively. The schematic flowchart of the program is given in Fig. 1. Without giving the details of the formulation which can be found in [3], these equations are solved at the scale of the whole casting as follows : 1. The heat flow conservation equation is solved according to an enthalpy scheme [2], i.e., using the enthalpy as the variable and linearising the temperature-enthalpy relationship. The velocity field of the previous time step is used. 2. The mass and momentum conservation equations are solved simultaneously using a GLS (Galerkin Least Squares) formulation for the pressure-velocity fields [4]. The temperature, volume fractions, solute concentrations and variations of the average specific mass (shrinkage) are taken from the previous time step. 3. Once the relative velocity of the fluid is known, the solute conservation equation(s) are solved in order to deduce the new average solute concentrations. Knowing the average enthalpy, t+∆t, and solute concentrations, <wi>t+∆t, at all the nodes of the mesh, the new temperature, volume fractions of phases, solute concentrations in the liquid and solid phases, and average specific mass are calculated according to a local microsegregation model. In the present case, the model developed by H. Combeau and A. Mo for binary alloys has been used [5]. It allows to consider non-linear phase diagrams, eutectic reaction and back-diffusion. It should be pointed out that the ternary model developed by X. Doré within the framework of this project [6] for Al-Mg-Si alloys has also been implemented in calcoMOS. The microsegregation model is equivalent to solving at each nodal point the N equations (4) with the following ones [5,6]: Equilibrium at the interface :
*
wi,s = ki wi,l , i = 1, N
(7)
Liquidus relationship : T = TL(wi,l) (8) Enthalpy relationship : <ρh> = <ρcp> T + <ρ>Lgl (9) The ki's are the partition coefficients, TL is the equation of the liquidus, <ρcp> is the average volumetric specific heat and L is the latent heat of fusion. When precipitation of secondary phases occurs, the problem becomes more complex and is not detailed here [5,6]. Providing a back-diffusion model is given (i.e., evolution of wi,s(g,t) appearing in Eq. (4)), Eqs. (4,7-9)
194 provide (2N + 2) equations for the (2N + 2) unknowns : wi,s, wi,l, T and gl. This back-diffusion model can be given either by a polynomial function approximation of wi,s(g,t) [5] or by the solution of Fick’s second law in the solid phase using a 1D FDM technique [6]. In the case of DC casting of Al alloys in which one is interested mainly in the stationary solution, the time stepping is used as a means of iteration among the equations. In this case, however, the Eulerian description of the macroscopic conservation equations (i.e., in a reference frame attached to the mould) must be coupled with a Lagrangian description of microsegregation (i.e., in a reference frame attached to the dendrites). Details of this coupling can be found in [3].
4
Results
The results which are presented in this section are for an Al-Mg alloy, using the microsegregation model of [5]. Fig. 2 shows the calculated stationary 2D velocity field near the liquidus isoline for a small DC cast ingot 5-cm thick. For symmetry reason, only half the ingot has been calculated over a length of 15 cm. The metal was supposed to be injected uniformly from the top at a velocity of 1 mms-1 with a nominal concentration of 4.5% Mg. Three calculations were performed under identical conditions but considering various sources of fluid movement : buoyancy only (case (a)), shrinkage only (case (b)) and buoyancy plus shrinkage (case (c))
1 cm (a)
.01 m/s
(b)
0.0002 m/s
(c)
0.01 m/s
Figure 2: Calculated stationary field of the relative fluid flow velocity for a small DC cast Al-4.5%Mg ingot (casting speed : 1 mms-1) with isolines of fraction of solid. a) buoyancy only ; b) shrinkage only ; c) buoyancy and shrinkage. Microsegregation model of Ref. [5] with a Scheil approximation
As can be seen, the relative velocity of the fluid in the fully liquid region (i.e., above the first isoline of fraction of solid also represented in this figure) is of the order of cms-1, regardless of whether shrinkage is included or not in the calculations (compare Figs. 2a and
195 2c). It is induced by thermal buoyancy, the solutal expansion coefficient associated with magnesium being small. Despite this fairly large relative velocity of the liquid, the induced macrosegregation is almost negligible when only buoyancy is considered (see Fig. 3). This is due to the fact that the flow pattern in this case is essentially parallel to the liquidus isotherm, i.e., perpendicular to the solute gradient, and vanishes quickly in the mushy zone. As reviewed in [1] and pointed out many years ago by Flemings, it is the component of the velocity along the thermal gradient which induces macrosegregation. Although the shrinkageinduced velocity is nearly two orders of magnitude smaller than that associated with thermal buoyancy (see Fig. 2b), it has a much more pronounced influence on the final concentration profile at the exit of the ingot (Fig. 3). In the mushy zone, the relative velocity of the fluid is on the order of 0.2 mms-1 only, but it is nearly perpendicular to the isofractions of solid. (Please note that the same flow exists in the mushy zone of Fig. 2c but is not visible with the scale used to visualise the overall flow pattern). Since the streamlines of the interdendritic fluid flow deviate from the ingot centerline (Fig. 2b), this induces a negative segregation at the center of the ingot and a positive one at the surface (Fig. 3). This shrinkage-induced segregation, commonly labeled “inverse segregation” in static castings [1], increases with the depth of the liquid pool, i.e., with the casting speed [3].
(a) (c)
Figure 3: Mg concentration profiles at the exit of the small DC cast ingot for the cases involving buoyancy only (a) and buoyancy plus shrinkage (c) of Fig. 2
The same phenomenon can be observed in real-size ingot computations (Figs. 4 and 5). Unlike small DC castings for which a good resolution can be obtained with a structured mesh following the coordinate axes [3], large scale simulations have to be performed with an unstructured mesh in order to obtain a sufficient accuracy within a reasonable CPU time. A first thermal calculation allowed to determine the approximate position of the mushy zone in this 0.25 x 1 m2 domain and to refine the mesh in this region (Fig. 4a). Feeding of metal through a distribution bag was simulated by inserting a horizontal plate inside the domain. Since the Eulerian-Lagrangian algorithm implemented in calcoMOS for back-diffusion calculations requires, at present, structured meshes (i.e., mesh points aligned along verticals), the macrosegregation result shown in Fig. 5 was obtained with the lever-rule approximation.
196
10 cm (1)
(2a)
(2b)
(3a)
(3b)
Figure 4: Real size DC casting simulation for an Al-4.06%Mg alloy solidified at 1 mms-1 : Mesh size (1), velocity streamlines (2) and isolines of fraction of solid (3) for the cases with shrinkage only (a) and with buoyancy plus shrinkage (b). Microsegregation module of [5] used with the lever rule. 16417 nodes, about 24h CPU on SGI2000
In Fig. 4, the streamlines and isolines of fraction of solid are shown for the two cases of shrinkage only (case (a)) and shrinkage plus buoyancy (case (b)). In the first case (Fig. 4(2a)), the streamlines turn around the horizontal plate and then are fairly straight : they directly outline feeding of the ingot from the upper gate to the mushy zone. On the contrary, thermal buoyancy in the liquid pool is turbulent if a laminar viscosity value is used : it gives rise to a complex fluid flow pattern, which never reaches a stationary state. Since no turbulent model was implemented in calcoMOS, an artificially increased viscosity by a factor 100 was used to obtain the result shown in Fig 4(2b). Results obtained with such an increased-viscosity approximation have been compared recently with those calculated with a turbulent model [7]. One can notice the influence of the primary and secondary coolings on the isolines of fraction of solid (small cusp near the top of Fig. 4(3)). The corresponding values of the heat transfer coefficients were deduced from experimental measurements and inverse method [8]. As can be seen, thermal buoyancy slighltly modifies the depth of the molten pool. However, the two corresponding segregation profiles at the exit of the ingot calculated for this real-size casting (Fig. 5) do not differ much, thus indicating again that shrinkage-induced macrosegregation is dominant over that associated with natural convection. A negative centerline segregration is again predicted by the simulation. The amount of segregation predicted by this model compares fairly well with the concentrations measured within the framework of the EMPACT project [9].
197
(a) (b)
Figure 5: Calculated Mg concentration pofiles at the exit of the real-size DC cast Al-4.06%Mg ingot (Fig. 4) for the cases with shrinkage only (a) and with buoyancy plus shrinkage (b)
5
Conclusion
Computation of macrosegregation in real-size DC cast ingots is a real challenge, even in two dimensions. The size of the region where both the fluid flow and solute gradients are non-zero is very small compared with the overall size of the ingot. Unstructured meshes offer clearly an advantage in this respect, but they are also more complicated to implement (e.g., for a mixed Lagrangian-Eulerian description). It has been shown in the present contribution that shrinkage-induced macrosegregation can already account for a fairly large portion of the concentration inhomogeneity measured in Al-Mg ingots. Thermal buoyancy has a minor influence for this alloy. This does of course not preclude anything about the influence of other phenomena such as grain movement, deformation of the mushy zone or solutal convection induced by other alloying elements.
6
Acknowledgements
This research was undertaken as part of the European Brite-Euram program EMPACT (A European Modelling Programme on Aluminium Casting Technology). It was sponsored by the European Community under contract CEC 0112 and by the Office Fédéral de l’Education et de la Science, Bern, under contract 95.0037-2. The authors would like to thank Dr H. Combeau, Ecole des Mines de Nancy, for providing the binary microsegregation model and for his help during its implementation in calcoMOS.
198
7
References
[1] Ch. Beckermann, Modeling of Macrosegregation : Past, Present and Future. To appear in: Flemings’ Symposium (TMS, Warrendale, PA, 2001). [2] N. Ahmad, H. Combeau, J.-L. Desbiolles, T. Jalanti, G. Lesoult, J. Rappaz, M. Rappaz and C. Stomp, Met. Mater. Trans., 29A, 1998, 617-30. [3] T. Jalanti, Etude et modélisation de la macroségrégation dans la coulée semi-continue des alliages d'aluminium, EPFL PhD Thesis No 2135, Lausanne, 2000. [4] L. P. Franca and S. L. Frey, Comput. Methods Appl. Mech. Engng., 99, 1992, 209-33. [5] H.Combeau and A. Mo, Met. Mater. Trans., 28 A, 1997, 2705-14. [6] X. Doré, H. Combeau and M. Rappaz, Acta mater., 2000, to appear. [7] G.-U. Grün, A. Buchholz and D. Mortensen, in Light Metals 2000 (TMS Publ., Warrendale, PA, USA, 2000) p. 573-78. [8] J.-M. Drezet, M. Rappaz, G.-U. Grün and M. Gremaud, Met. Mater. Trans., 31A, 2000, 1627-34. [9] Joly, G.-U. Grün, D. Daloz, H. Combeau and G. Lesoult, in Materials Science Forum (Trans Tech Publ., Switzerland, 2000) p. 111.
The Effect of the Differencing Scheme on the Numerical Diffusion in the Simulation of Macrosegregation B.C.H. Venneker Netherlands Institute for Metals Research, Delft, The Netherlands
L. Katgerman Delft University of Technology, Laboratory for Materials, Advanced Materials and Solidification Technology, Delft, The Netherlands
1
Introduction
A reliable calculation of macrosegregation during the casting of alloys depends on the accurate modeling of the associated physical mechanisms. Besides that the particular microsegregation model (Scheil, lever-rule) is of importance, the relative movement of the liquid and solid phase inside the mushy zone controls the amount of macrosegregation. In solving the solute concentration equation, the accuracy of the velocity field is thus of great concern. From the literature on Computational Fluid Dynamics, we also know that in high Peclet number flows, the incorrect treatment of the convection terms causes numerical diffusion which can completely overshadow the actual physical diffusion. Throughout the history of CFD, a great number of differencing schemes for the convection term has been proposed in order to reduce the numerical diffusion. In the current research several of these schemes are examined on their ability to correctly predict macrosegregation in the DC casting of an Al-4.5wt% Cu alloy.
2
Numerical Errors in Modeling
If it is assumed that in a particular CFD simulation the right equations are solved, there is still the concern of solving the equations in the right way, see (1) for an important discussion on this topic. The (partial) differential equations of the complete model are transformed into algebraic discretisation equations. Besides that we have to be sure that we have reached convergence, a potentially source of errors lies in the transformation of differential equations to discretisation equations. Particularly the convection term needs special attention. Incorrect treatment of the convection term can result in two types of errors: numerical diffusion - the spreading out of profiles – and numerical dispersion – the appearance of wiggles (oscillations) on the profile. Roughly speaking, numerical diffusion is common to odd-order schemes when convection dominates physical diffusion (high cell-Peclet numbers). Numerical dispersion is common for even-order schemes and occurs in the vicinity of steep gradients. A special form of numerical diffusion is crosswind diffusion, which manifests itself when the flow field is not aligned with the computational mesh. Note the importance of this issue in DC-casting: the more or less parabolic profile of the mushy zone, and the circulating fluid flow in the liquid sump caused by buoyancy effects is a definite cause of the Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
200 nonalignment of flow field and mesh. An attempt to align the mesh has some positive effects, see (2), but is for 3D simulations a cumbersome task. At this point it is worth investigating how these sources of error affects predicted macrosegregation. As shown in several papers (2,3,4,5) the numerical diffusion in the solute conservation equation manifests itself as wiggles in the concentration profile. Although this may point to numerical dispersion at first sight, in (2) it was explained that the origin of wiggles is in the first cell in the upper part of the mushy zone. Moving downwards the scattering in the concentration is frozen. A better term for this phenomenon is false segregation. The location of minimum and maximum concentration can be traced back to the position of the steps in the staircase profile of the liquidus. The differencing scheme in these cases was upwind (2,5) and the power-law scheme (4). Note that the power-law scheme for Peclet numbers greater than 10, is equal to the upwind scheme. It is commonly known that upwind differencing introduces the greatest amount of artificial diffusion. In reference (4) an interesting comparison between a finite volume and finite element package is given. In the FE simulations, however, the streamline upwind Petrov Galerkin method of (6) is used. This scheme has been designed in such a way that artificial diffusion is only added in the streamwise direction, and thus eliminating crosswind diffusion. The results with that method were free of oscillations, whereas with the power-law scheme in some cases oscillations were present, see fig. 8 in that paper. In the remaining of this paper, we will look at some other schemes and will compare their behavior with that of upwind differencing.
3
Differencing Schemes
Out of the vast variety of differencing schemes available in literature, the following schemes were chosen: • upwind differencing • central differences • QUICK • QUICK-2D • higher-order upwind with Superbee flux limiter • skew upwind differencing scheme (SUDS) Each of these schemes is a representative of a particular class of schemes. Thus, upwind and central differencing are the most common used schemes of first and second order accuracy. The QUICK scheme of (7) is a third order modification of upwind differencing. QUICK-2D (8) is the local two-dimensional version of QUICK. The fifth scheme is only second order but has as special feature that the appearance of wiggles on the solution is prevented by the use of a non-linear flux limiter, in this case superbee of (9). Finally, the last scheme is like the QUICK-2D scheme also locally two-dimensional but has as special feature that the exact direction of flow is taken into account. It was especially designed by (10) to take into account the local direction of the flow at each cell face. The performance of these schemes was first tested with two benchmarks: the convection of a scalar profile over 180 degrees - the so-called Smith & Hutton problem (11) and laminar convection in a driven cavity. For both cases, it turned out that the two-dimensional QUICK scheme is the best option. Second best are SUDS for the Smith & Hutton problem, and central
201 differencing in case of the driven cavity. A known peculiarity of the Superbee flux limiter was found in case of the Smith & Hutton problem: i.e. the overcorrecting of numerical diffusion.
4
Numerical Procedures
The following set of equations was solved with the CFD-package CFX version 4.3. ∂ρ + ∇ ⋅ (ρ U ) = 0 ∂t
(1)
∂ (ρ U ) + ∇ ⋅ (ρ UU ) = − ∇p + ∇ ⋅ (µ l ∇ U ) + ρ 0 g − ρ 0 β T (T − T0 )g − µ l K (U − U s ) ∂t ∂f ∂ (ρ h ) + ∇ ⋅ (ρ Uh ) = ∇ ⋅ (λ∇ T ) − ρ L l − ρ L U s ∇ f l ∂t ∂t ∂ (ρ c ) + ∇ ⋅ (ρ Uc ) = ∇ ⋅ (ρ f l D l ∇ c ) + ∇ ⋅ [ρ f l D l ∇(c l − c )] − ∇ ⋅ [ρ (c l − c )(U − U s )] ∂t
(2) (3) (4)
The mushy zone is modeled as a porous region, with K the permeability calculated with the Kozeny-Carman relation. Liquid fraction (fl) and concentrations are calculated according to the lever rule. In order to investigate fully the effects of false segregation, no solidification shrinkage was taken into account. The calculated velocity field is therefore completely determined by the inlet and thermal buoyancy. A relative small ingot of 5 cm diameter and 10 cm length was taken as geometry. Metal entry was over the complete width of the ingot. The inlet velocity was 1.0 mm/s and as alloy Al-4.5wt% Cu. The material properties of this alloy were taken from (3). As primary cooling a heat transfer coefficient of 100 W/m2K was taken over the first 1.25 cm. For secondary cooling, we assumed a heat transfer coefficient of 500 W/m2K. The central difference scheme is unstable for the solute conservation equation, so in that case the upwind scheme was used. From the tested schemes, only QUICK-2D and SUDS are not standard in CFX 4.3. The program offers, however, the opportunity to include userdefined source terms. So, for these schemes, the upwind scheme was chosen as standard and the difference with the preferred scheme is included in the source term using the deferredcorrection method. The last term in equation (2) is included as a user-defined body force. The last term in eq.(3) and (4) are also convection terms and these were also included in the source term, with of course, the appropriate discretisation scheme. Besides the importance of the differencing schemes, the cell size of the mesh is also a crucial factor for numerical diffusion. We therefor performed simulations with each scheme on 4 different meshes: 10, 20, 40 and 80 cells in the radial direction. The cell-width in the axial direction was in all cases equal to that in the radial direction.
5
Macrosegregation Predictions
First, the effect of grid refinement is discussed. From figure 1a, we can conclude that all schemes converge to the same asymptotic concentration at the ingot center for extremely fine grids. With coarse meshes, however, the value can be either greater or smaller than the inlet composition, depending on the particular scheme.
202 All schemes predict a decrease in the concentration with decreasing cell size. It seems that predictions made by the QUICK, QUICK-2D, Superbee and SUDS scheme converge for small cell sizes towards each other, while the upwind (and central) scheme predict an ever increasing negative segregation. Simulations with an even finer mesh have to be performed in order to investigate whether all schemes finally convergence to the same asymptotic value. 5
4.6
4.4
a)
4.6
4.4
4.2
4.2
4 0.0001
Upwind Central Quick Quick-2D Suds Superbee
4.8
C_Wall [wt%]
4.8 C_Centre [wt%]
5
Upwind Central Quick Quick-2D Suds Superbee
0.001 cellsize [m]
4 0.0001
0.01
0.001 cellsize [m]
0.01
b)
Figure 1: Cu-concentration at ingot center (a) and edge (b) as function of the cell size.
Due to the easy conduction of heat by metals, cell Peclet number for the enthalpy equation are even for the coarsest grid, lower than 2.0. Numerical diffusion is than negligible compared to physical diffusion. Therefore, there is little difference in the temperature profile with the different schemes and meshes. The diffusivity in the momentum equations has an intermediate value. The result obtained with the finest mesh was more or less gridindependent, specifically for the radial momentum with the two QUICK schemes. From previous simulations of macrosegregation in literature, it is known that when only buoyancy-induced flow is considered, the amount of segregation is negligible. For the simulations in this work, this is only true with the two QUICK schemes and the Superbee scheme, see figure (1). This should come as no surprise, as both QUICK-schemes were tested as one of the best schemes for the Smith & Hutton problem. Except for near the outer edge, the predicted concentrations are close to the inlet concentration of 4.5 wt%, especially in the case of Superbee. With these two schemes, the predicted macrosegregation is quite large. Despite the obvious presence of false segregation, the profiles are surprisingly smooth. This has probably to do with the relatively small ingot in relation to the boundary conditions. These were such, that the mushy zone is fairly thick (about 5 cm), so it is described by a large number of grid cells. Only in case of SUDS, a strange long-wave oscillation is present. It is unknown what the cause of this oscillation is.
203 4.8
Upwind Central Quick Quick-2D Suds Superbee
4.7
C [wt%]
4.6
4.5
4.4
4.3
4.2
0
0.005
0.01
0.015
0.02
0.025
r [m]
Figure 2: Predicted macrosegregation profile with the finest mesh for all tested schemes.
6
Discussion and Conclusions
The performance of six differencing schemes in the calculation of macrosegregation of Cu in an Al-4.5wt%Cu alloy has been investigated. Conditions of casting were such that the true macrosegregation can be considered negligible. Any macrosegregation, predicted by the simulations, is therefore a direct consequence of numerical diffusion and is called false segregation. In terms of minimum amount of false segregation, the higher upwind scheme with the Superbee flux limiter has the best performance. It is however known that this scheme also reduces physical diffusion. The two-dimensional version of the QUICK scheme is therefore considered to be the most appropriate scheme of the six that were tested. It is emphasized here that even with the finest mesh, no grid-independent solution for the concentration field has been obtained. Given the size of the ingot and the large number of mesh cells, an even better scheme than QUICK-2D has to be found before reliable simulations of real size castings can be made. A possible candidate may be found in the finite element literature. The Streamline Upwind Petrov Galerkin is an often-used method in FE-packages, and has proved to give excellent results in terms of the absence of false segregation (11). Higher order (fourth, fifth) schemes are another possibility.
204
7
References
[1] P. J. Roache, Verification and validation in computational science and engineering, Hermosa publishers, 1998. [2] B. C. H. Venneker and L. Katgerman, Macrosegregation during DC casting of aluminium alloys: numerical issues and the effect of metal entry, MCWASP IX proceedings, Aug. 2000, 8p. [3] V. Reddy and C. Beckermann, Metall. Trans. B., v 28, 479-489, June 1997. [4] Ahmad et al., Metall. Trans. B., v 29, 617-630, Feb. 1998. [5] J. Vreeman and F. P. Incropera, Num. Heat Transf. B., v 36, 1-14, 1999. [6] N. Brooks and T. J. R. Hughes, Comp. Meth. Appl. Mech. Eng., v 32, 199-259, 1982. [7] P. Leonard, Comp. Meth. Appl. Mech. Eng., v 19, 59-89, 1979. [8] P. Leonard, Elliptic systems – finite difference methods IV, in Handbook of Numerical Heat Transfer, eds. W. J. Minkowycz et al., John Wiley & Sons, 1988. [9] L. Roe, Some contributions to the modeling of discontinuous flows, Lectures in Applied Mechanics, v 22, American Mathematical Society, 1985. [10] G. D. Raithby, Comp. Meth. Appl. Mech. Eng., v 9, 153-164, 1976. [11] M. Swierkosz et al., Numerical simulation of macrosegregation in calcoMOS, Empact report. March 2000, 64p.
Application of a New Hot Tearing Analysis to Horizontal Direct Chill Cast Magnesium Alloy AZ91
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
206
207
208
209
210
Micro- and Macrostructures
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
Nucleation Studies of Grain Refiner Particles in Al-Alloys P. Schumacher University of Oxford, Oxford, UK
1
Introduction
The observation of nucleation events in conventional solidification experiments is difficult, because of subsequent growth and impingement of crystals obscuring nucleation events. Nucleation mechanisms of α-Al on Al-Ti-B based grain refiner particles are still not very well understood and mechanisms of nucleation by either borides only or by a peritectic reaction only [1] were proposed early, but were based on indirect, post-solidification based observations. Nevertheless, in casting practice it appeared empirically that both borides and a peritectic reaction were required for sufficient grain refinement, which was empirically verified in synthetically added borides to Al melts [2]. The interpretation of the grain refiner efficacy to nucleate Al cannot be easily derived from the average grain size of as cast microstructure which is strongly affected by growth restriction effects of solute in a given alloy [1,3]. In particular, Ti can have potentially a double role as a strong growth restrictor and as a nucleation agent via a peritectic reaction with liquid and bulk Al3Ti to form Al above 0.1 at. % Ti. The difficulty of identifying nucleation mechanisms has been overcome recently by addition of conventional grain refiner particles into Al-based eutectic alloys which on rapid quenching from metallic glasses [4,5]. Nucleation of Al occurred in the undercooled melt but growth was halted when the atomic mobility rapidly decreased on cooling below the glass transition temperature. To distinguish between nucleation events on added grain refiner particles and those occurring in the glass without additions, the crystallization behaviour of the glasses (devitrification) was studied and was reported elsewhere [6,7]. The current work uses Al based metallic glasses not used previously to simulate effects of solute on the nucleation mechanism of α-Al on commercial Al-Ti-B based refiners in glass-forming model alloys. The amount of Ti and Zr is varied to study the effects of Ti as a nucleation agent and to study Zr poisoning.
2
Experimental Methods
Nominal metallic glass compositions of Al85Ni10Ce5 (in at. %) were produced by arcmelting under reduced, inert atmosphere from pure elements (99.95 at. %). The aluminium content of the arc-melted alloys was reduced to compensate for excess Al of subsequently added grain refiner, Ti and Zr additions. Grain refiner rods (supplied by London & Scandinavian Metallurgical Co. Ltd.) with a composition of Ti-B-Al 5:1 (wt. Ti %) and of stoichiometric Al-TiB2 (2.2 wt. Ti %) composition were added prior to melt-spinning. Charges were heated to 1200°C in boron nitride crucibles and held for varying times in an inert helium atmosphere before ejecting onto a rotating copper wheel with a speed of 40 ms-1. Resulting ribbons were Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
214 approximately 2 mm wide and 30 µm thick. The melt-spun ribbons were examined using transmission electron microscopy (TEM). Thin foils were prepared by electropolishing in 6% perchloric acid in ethanol at a temperature of approximately –30°C at 20 V. TEM was on a Philips 400T microscope at 120 kV.
3
Results and Discussion
3.1
Effects of Ti on the Nucleation Mechanism
In figure 1, two clustered hexagonal TiB2 particles with hypo-peritectic addition of 0.05 at. % Ti are shown embedded into a partial amorphous matrix of Al85Ni10Ce5. The hexagonal platelet shape of the particle tilted with its <00.1> zone axis nearly parallel to the electron beam can be seen while the second particle is tilted with its <11.0> parallel to the electron beam is viewed edge-on to the basal face. On both particles copious nucleation of Al has occurred but can only be seen on the second particle viewed with its basal face edge-on. This suggests that similar to previous observations [4,5] a distinct crystallographic orientation relationship is favored for nucleation of Al on boride particles.
Figure 1: (a) Faceted TiB2 particles (5:1) embedded in partial amorphous Al85Ni10Ce5. The hexagonal platelets are tilted close to the (1) <00.1> and (2) <11.0> zone axis and are separated by a thin layer (arrow). The layer nucleates copious Al on basal faces of the TiB2 only. (b) The inverted (negative) SAD pattern showing a 0-order streak indicative of the layer
Some nucleation of Al has occurred in the surrounding glassy matrix. The nucleation density of Al particles in the matrix is much lower than that on the boride particle indicating that the boride acts as a better nucleation site for Al. However, between the boride particles, marked in figure 1 with an arrow, and between the nucleated Al and the boride a layer is visible. This layer has been reported previously as a thin layer of Al3Ti in Al85Y8Ni5Co2 glasses [4,5] and is visible here in selected area diffraction (SAD) patterns as a weak streak.
215 Further identification of the layer as Al3Ti in the present case is difficult due to its limitation in thickness and in slight misorientations to the dominant orientation relationship between the boride, aluminide and Al with its close-packed planes being parallel. It is noteworthy, that the TiB2 particles do not nucleate Al directly but indirectly by a thin layer of Al3Ti, which appears to be preserved during the high melting temperatures in the metallic glass experiments. In contrast, no separate Al3Ti particles where found in the glassy matrix suggesting dissolution of these particle similar to that in casting practice. The layer does not only nucleate Al but appears to enhance cluster formation as suggested by its presence between boride particles.
Figure 2: A stoichiometric TiB2 particle embedded in glassy matrix (a) showing no nucleation of Al. and (b) a stoichiometric TiB2 particle with 0.3 at. % excess Ti showing a restored nucleation capacity for Al on basal faces
In figure 2(a) the effect of a lack of excess Ti on the nucleation mechanism of Al on boride particles is demonstrated. Stoichiometric addition of TiB2 particles did not result in copious nucleation of Al on basal faces of boride particles. This highlights that excess Ti is needed for the formation of a nucleating Al3Ti layer. In figure 2(b) stoichiometric boride particles at hyper-peritectic addition of 0.3 at. % excess Ti are embedded in an amorphous Al85Ni10Ce5 matrix. At hyper-peritectic addition levels the nucleation capacity of the boride particles is restored and nucleation of Al can be observed. The formation of a nucleating aluminide layer on hyper-peritectic addition of Ti suggests that the layer can be formed within the melt. However, for the given short melting times of 1 min the nucleation capacity of the boride particle with new layer is not restored to that of an effective boride particle of a 5:1 refiner rod seen in figure 1 suggesting that the layer may be present in the refiner addition. It is noteworthy, that the missing layer of Al3Ti on stoichiometric boride particles would only represent a few ppm of in commercial grain refinement added excess Ti with the remainder being available for growth restriction. This suggest a double role of excess Ti being responsible for the initial formation and/or stabilization of the Al3Ti nucleating layer and secondly for growth restriction of the nucleated Al crystals.
216 3.2
Effects of Zr on the Nucleation Mechanism
The effects of poisoning in Zr containing melts were simulated in Al85Ni10Ce5 glasses with Zr additions of 0.3 at. % Zr and 0.05 at. % excess Ti coming from 5:1 grain refining rod. Grain refiner addition and Zr additions where held at 1200°C for 1, 10 and 30 min before meltspinning. The processing temperature is somewhat higher than in conventional casting but is believed to be compensated by a lower atomic mobility in glass forming melts and shorter processing times than in conventional casting permitting a qualitative comparison.
Figure 3: (a) TiB2 particle embedded in a nominal Al85Ni10Ce5 glass with 0.3 at. % Zr addition held for 1 min at 1200°C showing copious nucleation of Al (arrow) and (b) TiB2 particle held for 30 min at 1200°C showing no nucleation of Al on basal faces
In figure 3(a) no significant difference in nucleation efficacy to that in figure 1 without Zr additions can be detected after holding the melt for 1 min before melt-spinning. Nucleation of Al on the TiB2 particle is copious on basal faces of the boride only. Faint streaks of the 0order spot in the SAD pattern, caused by the shape of the layer in reciprocal space along the <00.1> TiB 2 direction, indicate the presence of a layer on these basal faces. In contrast, in figure 3(b) after 30 min of holding at 1200°C before melt-spinning no nucleation of Al can be detected on boride particles and no obvious streaks are apparent in SAD indicating the absence of a layer on boride particles. At intermediate holding times at 10 min, a mixture of boride particles with and without copious nucleation of Al can be found. This indicates that the surface properties of boride particles are affected by a time dependent reaction not affecting all boride particles simultaneously. Interestingly, Al3Ti and Al3Zr have very similar crystal structures DO22 and DO23 respectively and have limited solid solubility in each other [8]. In contrast to experiments where Ta was present [4], the absence of a layer at longer holding times suggests that Al3Ti has not been replaced by a layer of Al3Zr. Work on Al87Ni10Zr3 glasses simulating high hyper-peritectic Zr concentrations have shown that TiB2 particles are readily transformed into ZrB2 particles which do not provide sufficient lattice matching to support an aluminide layer [9]. It appears that at lower concentrations this
217 reaction occurs slower and not simultaneously, suggesting in conventional grain refinement of Zr containing alloys a sufficiently long time gap at lower casting temperature for sufficient grain refinement. However, further work is required to elucidate transformation kinetics of the temperature and time dependent Zr-poisoning by boride transformation.
4
Conclusions
Similar to previous studies using an Al-Y-Ni-Co glass [4,5], separate bulk Al3Ti particles dissolved as in commercial grain refining practice at hypo-peritectic addition levels and do not play a role as a nucleant. Instead, only TiB2 particles, coated with a previously identified thin layer of Al3Ti and visible as a streak in SAD pattern, nucleate copious amounts of Al. The nucleated Al crystals are somewhat smaller due to the higher undercoolings achieved in metallic glasses while in commercial casting a boride particle of 1 µm in diameter would be expected to be occupied by one Al crystal alone at approximately 1°C of undercooling. Some excess Ti is needed to form a layer of Al3Ti on boride particles, amounts beyond that will be available for growth restriction. Addition of Zr results in a temperature and time dependent reaction to transform TiB2 into ZrB2 which is preferred over the substitution of Zr in Al3Ti or the formation of Al3(Zr,Ti). Further work is required to understand the detailed mechanism of Zr- poisoning. The nucleation mechanism observed on coated boride particles appears not to be dependent on the glass-forming model alloy chosen and is consistent with layers of Al3Ti observed in grain refiner rods analysised in detail by TEM [10,11].
5
Acknowledgements
The authors gratefully acknowledge financial support from the EPSRC in conjunction with London & Scandinavian Metallurgical Co. Ltd. and Alcan International Ltd.
6 [1] [2] [3] [4]
References
D. G. McCartney, Int. Mater. Rev. 1989, 34, 247 - 260. P.S. Mohanty and J.E. Gruzleski, Acta Metall. Mater. 1995, 43, 2001 - 2012. J.A. Spittle and S. Sadli, Mater. Sci Technol. 1995, 11, 533 - 537. P. Schumacher and A.L. Greer, Light Metals 1995 (Ed.: J. Evans), TMS, Warrendale, PA, 869 - 877. [5] P. Schumacher et al., Mater. Sci. Technol. 1998, 14, 394 - 404. [6] R.F. Cochrane, P. Schumacher et al., Mater. Sci. Eng. A 1991, 133, 367 - 370. [7] P. Schumacher and A.L. Greer, Mater. Sci. Eng. A 1997, 226, 794 - 797. [8] S. Tsurekawa and M.E. Fine, Scr. Metall. 1982, 16, 391 - 395. [9] P. Schumacher, P. Cizek and A.M. Bunn, Light Metals 2000 (Ed.: R.D. Peterson), TMS, Warrendale, PA, 839 - 844. [10] B. Mckay, P. Cizek and P. Schumacher, Light metals 2000 (Ed.: R.D. Peterson), TMS, Warrendale, PA, 839 - 844. [11] P. Cizek and P. Schumacher, this conference.
Effect of Solute Elements on the Grain Structures of Al-Ti-B and Al-Ti-C Grain-Refined Al Alloys A. Tronche1,2 and A. L. Greer1 1
University of Cambridge, Cambridge, UK Péchiney Centre de Recherches de Voreppe, Voreppe, France
2
1
Abstract
The influence of solute elements on the structure of inoculated Al alloys, and the best parameter to represent solute effects, are assessed in this work. Experimental grain sizes were measured on Al-Ti-C grain-refined samples under TP-1 conditions [1], and results obtained in other studies [2] of Al-Ti-B refiners were analyzed also. The evolution of the structure with the solute content is found to be independent of the nature of the nucleation centers. The parameter Q=m(k−1)C0 [3,4] is found to represent better the effect of the solutes than the parameter P=Q/k also used in this type of study [2]. Nevertheless, when the solute alters the nucleation process, like Zr in the case of Al-Ti-B refiners [5], the growth-restriction parameter does not allow to predict the structure of the final product. It was shown in the case of Al-TiC refiners that, for example, the nucleation centers, i.e. the TiC particles [6], are poisoned by Si. The measured data were compared to results computed using the “free-growth” model [7].
2
Introduction
Grain refinement of Al and its alloys is a widespread industrial technique, which is currently mainly achieved by addition of Al-Ti-B master alloys. The final structure of the cast product is known to depend on the nature and number of the nucleation sites, and on the growth restriction imposed by the solutes present in the melt. The latter effect was extensively studied in the case of Al-Ti-B refiners [2,3,4]. Several parameters have been used to represent the effect of the solute elements; the constitutional-supercooling parameter P=m(k1)C0/k [2] (m, k and C0 being respectively the liquidus slope, the partition coefficient and the solute content of the alloy considered), the parameter Q=kP and the parameter U=DelementQ/DAl which takes account of the diffusion coefficient of the solute, Delement [8]. Even if solute diffusion is an important process during the solid growth, coefficients of diffusion are difficult to assess and thus the parameter U is not considered in this work. The usefulness of the parameters in representing the measured evolution of the grain size is assessed for Al-Ti-B inoculated alloys, and for the newly developed Al-Ti-C grain refiners. The experimental evolutions of the grain size as a function of the growth restriction are compared to computed data. Recently, a model for grain refinement, based on that of Maxwell and Hellawell [3], was developed and tested for Al-Ti-B grain refiners [7]. This model, applied in this work, is valid
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
219 only to predict the grain size of equiaxed structures. Other models can be used to predict whether the structure is equiaxed or not [9].
3
Experimental Procedure
The influence of various solutes on the grain structures of Al alloys inoculated with Al3wt%Ti-0.15wt%C-1wt%Fe (produced by LSM Co Ltd) has been studied. The solute elements considered correspond to those considered by Spittle and Sadli [2] and represent the main elements found in commercial wrought alloys. The alloys studied are listed in Table 1. The alloys are prepared from commercial-purity Al. The solutes are added either using commercial master alloys or as the elements. Table 1: Chemical compositions of the inoculated alloys studied (2 kg tonne−1 Al-3wt%Ti0.15wt%C-1wt%Fe) and corresponding P and Q growth-restriction parameters, using the data reported in earlier articles [7]. The chemical compositions are measured by optical emission spectroscopy. Solute content: P Q (in wt% except Ti in ppm) (K) (K) Cr Cu Fe Mg Si Ti Zn Zr Mn 1 0 0 0.17 0.98 0.03 64 0 0 0 24.0 5.2 2 0 0 0.16 0 0.03 73 0 0 0.52 20.3 2.4 3 0.032 0 0.19 0 0.03 69 0 0.178 0 20 2.0 4 0 0 0.16 0 0.24 76 0 0 0 26.95 3.0 5 0 0 0.15 5.02 0.03 63 0 0 0 47.4 15.8 6 0 0 0.17 0 0.02 75 3 0 0 22.7 3.4 7 0 0 0.21 0 0.02 71 4.84 0 0 27.6 5.1 8 0.08 0 0.21 0 0.04 68 0 0 0 21.9 0.98 9 0 0 0.17 0 0.5 69 0 0 0 40.6 4.94 10 0 0 0.17 0 0.77 63 0 0 0 53.7 6.51 11 0 0 0.17 0 1.05 63 0 0 0 67.3 8.14 12 0 0 0.17 0 1.7 65 0 0 0 98.8 11,9 13 0 0 0.17 0 2.76 71 0 0 0 150.3 18.1 14 0 0 0.17 0 3.95 69 0 0 0 208.1 25.0 15 0 0 0.17 0 5.11 69 0 0 0 264.4 31.8 The grain size is assessed in TP-1 tests under standard conditions. The grain refiner is held in the melt at 730°C for 5 minutes. The mean linear intercept is measured on the TP-1 examination planes [1]. Calculations are carried out using the “free-growth” model. Any particle of size d gives birth to a grain when the solid cap on the particle can freely grow. At this stage the cap and the particles have the same diameter. In such a model the largest particles are active first since the nucleation undercooling related to these particles is lower. The efficiency of the grain refiner is fixed by the onset of recalescence. The size distribution of the nucleation centers is entered in the program and at each undercooling the number of activated particles is calculated. The temperature evolution depends on the heat released by the nucleation and the growth processes. The model and the calculation algorithm are described elsewhere [7]. The calculations are carried out assuming that the cooling rate at the TP-1 examination plane equals 3.5 K s−1 [1]. The main input of the program is the size distribution of the nucleation
220 sites. The size of the particles is measured on polished sections and transformed into a 3D distribution using standard stereological techniques. This had been done previously for the TiB2 particles [7], and was done in this work for the TiC clusters found to be acting as nucleation centers in the Al-Ti-C refined Al alloys [6]. The model predicts the grain size only in equiaxed structures. The type of structure (equiaxed or columnar) can be predicted prior to computations [9]. In the test used in this work, the temperature gradient roughly equals 1 K mm−1, a low value favoring equiaxed growth.
4
Experimental Results
The evolutions of the measured and computed grain sizes of the Al-Ti-B refined samples are reported in Figure 1 as a function of P=m(k−1)C0/k and Q=kP. 1000
1000
600 C om puted G rain Size
400
800 Grain Size ( µ m )
Grain Size ( µ m )
Measured Grain Size
Measured Grain Size
800
600
200
200 0
C om puted Grain Size
400
0
0
20 40 60 80 100 G row th-R estriction Param eter, P (K)
0
5
10
15
20
Growth-R estriction Param eter, Q (K)
Figure 1: Evolution of the grain size of Al-5wt%Ti-1wt%B inoculated alloys (2 kg tonne−1) as a function of the growth-restriction parameters P and Q
The grain size of Al-Ti-B refined alloys decreases when the growth restriction imposed by the solutes increases, and reaches a plateau at high P or Q values. The data are less scattered around the trend line when the parameter Q is used. The study was repeated under Al-Ti-C inoculation and the evolution of the grain size is plotted as a function of Q in Figure 2. Measured G rain Size
Grain Size ( µ m )
1000 800
8
15 14 10
600 400 200 0
9
2 3 1 4 7 6
13
11 12 5
Predicted
0 5 10 15 20 25 30 35 Grow th-Restric tion Param eter, Q (K)
Com puted G rain Size (µ m )
1000
1200
800 600 400 1
200 0
5 0
6 2
10
11 12 4 3 14 6 9 15 13 7 200 400 600 800 M easure d Grain Size (µ m )
1000
Figure 2: Evolution of the grain size of Al-3wt%Ti-0.15wt%C-1wt%Fe inoculated alloys (2 kg tonne−1) as a function of the growth-restriction parameter Q. Predicted and measured sizes are also compared directly. The numbers reported on the graphs correspond to the numbers in Table 1 and represent the alloys
221 As observed with Al-Ti-B master alloys, the grain size first decreases sharply with the growth-restriction parameter and then reaches a plateau. The computed match fairly the measured data for any alloy except the Si-containing ones. Unexpectedly, the grain size of these Al-Si alloys increases when the growth restriction increases (alloys 9 to 15). The evolution of the grain size of the Al-Si alloys is reported as a function of the Si content in Figure 3. 900
Grain Size ( µ m )
800 700 600 500 400 300 0
1
2
3
4
5
6
7
8
Si Content (wt% )
Figure 3: Evolution of the grain size of Al-3wt%Ti-0.15wt%C-1wt%Fe inoculated Al-Si alloys (2 kg tonne−1) as a function of the Si content
The evolution can be divided into two parts: for Si content < 3 wt%, the grain size increases slowly, while it increases quickly above this value.
5
Discussion
The evolution of the grain size with the growth restriction imposed by the solutes is independent of the representing parameter considered, P or Q, and is the same for the Al-Ti-B and Al-Ti-C grain refiners. The grain size first decreases sharply with the solute content, and then reaches a plateau. Under the experimental conditions (2 kg tonne−1 of Al-5wt%Ti1wt%B or Al-3wt%Ti-0.15wt%-C-1wt%Fe) the value of the grain size on the plateau is smaller for the B-containing master alloy. In this refiner, the higher particle volume fraction compared to the Al-Ti-C master alloy leads to a higher number of nucleation sites, despite the TiB2 particles being larger, and thus leads to the finer grain size. The trends of the measured and the computed grain sizes are clearer when the parameter Q is used to represent the effect of the solutes. In the phase diagram of the relevant solute at C0 solute content, P is the equilibrium freezing range. It represents the operating condition at the solid-liquid interface when the solid composition is the far-field composition. This case occurs under steady state planar-growth conditions, and is not related to the equiaxed growth problem. Q on the other hand represents the variation of the fraction solid fs with temperature at the liquidus. From the the Scheil-Gulliver approximation it follows that [10]: df s dT
= Tliq
1 1 = mC 0 (k − 1) Q
(1)
In the melt undergoing solidification, heat is released by the nucleating and growing solid. The rate of solid formation dfs/dt, and thus the rate of heat released, is according to Eq. (1)
222 related to the growth restriction factor Q: a higher Q reduces the absolute value of dfs/dT and in consequence reduces the rate of latent heat released. The recalescence is thus delayed allowing more grains to be nucleated. Thus under equiaxed growth, the number of grains which can nucleate, i.e. the final grain size, seems to be directly related to the parameter Q. As shown in Figure 1 and 2, it is possible from the value of Q to predict the grain size of the cast product, which is well modeled by the “free-growth” model as shown in Figure 2 for Al-Ti-C. The model fails to predict the grain size when the poisoning of the nucleation centers occurs, as it happens in Al-Ti-C refinement of Al-Si alloys. The evolution of the grain size of Al-Si alloys with the Si content reveals two stages. Microstructural observations, linked to thermodynamic considerations, reveal that for Si < 3 wt% the TiC clusters transform into particles rich in Al, Si & C. This dissolution/precipitation reaction is, like the same transformation of TiC into Al4C3 particles, sluggish [11]. Above 3 wt% Si, the particles transform into SiC. This reaction occurs quickly, leading to a complete disappearance of the nucleation sites, i.e. complete loss of the grain-refining effect.
6
Conclusions
Solute elements are of great importance in grain refinement. They restrict the solid growth since they need to partition between the solid and the liquid. The final grain size of a cast product largely depends on the growth restriction imposed by the solutes, which is well represented by the parameter Q under equiaxed growth. In addition, solute elements facilitate the columnar-to-equiaxed transition. Nevertheless, solute elements can also act during the nucleation stage either by destroying the nucleation centers (poisoning of TiB2 by Zr, of TiC by Si as shown in this study), or by enhancing the nucleation potency of the particles (formation of an Al3Ti layer on TiB2 particles) [12]. It was shown in this work that the effect of the solute elements can be well predicted by the “free-growth” model for both Al-Ti-B and Al-Ti-C refiners unless poisoning of the nucleation centers occurs.
7
Acknowledgements
AT acknowledges financial support from the LSM Co Ltd and Péchiney (CIFRE studentship), and useful discussions with A. Hardman and P. S. Cooper at LSM, with P. Jarry at Péchiney Centre de Recherches de Voreppe, and with M. Vandyoussefi at the University of Cambridge.
8
References
[1] Aluminum Association Standard Test Procedure for Aluminum Grain Refiners, The Aluminum Association, Washington D.C. 20006, 1987. [2] J. A. Spittle, S. Sadli, Mater. Sci. Technol., 1995, 11, 533 - 537. [3] Maxwell, A. Hellawell, Acta Metall., 1975, 23, 229 - 237.
223 [4] L. Backerud, M. Johnsson, Light Metals 1996, (Ed.: W. Hale), TMS, Warrendale, PA, 1996, 679 - 685. [5] J. A. Spittle and S. Sadli, Cast Metals, 1994, 247 - 253. [6] Tronche, A. L. Greer, submitted to Phil. Mag. Letters. [7] L. Greer, A. M. Bunn, A. Tronche, P. V. Evans, D. J. Bristow, Acta Mater., 2000, 48, 2823 - 2835. [8] P. Desnain, Y. Fautrelle, J. L. Meyer, J. P. Riquet, F. Durand, Acta Metall., 1990, 38, 1513 - 1523. [9] M. Vandyoussefi , A. L. Greer, these proceedings. [10] W. Kurz, D. J. Fisher, The Fundamentals of Solidification, 3rd edition, Trans Tech Publications, 1992, p. 285. [11] Tronche , A. L. Greer, Work in progress. [12] P. Schumacher, A. L. Greer, J. Worth, P. V. Evans, M. A. Kearns, P. Fisher, A. H. Green, Mater. Sci. Technol., 14, 1998, 394 - 404.
Grain Refinement Process in Aluminium Alloys Type AlZnMgZr 7RPDV]6WXF]\ VNLDQG0DU]HQD/HFK*UHJD Institute of Non-Ferrous Metals, /LJKW0HWDOV'LYLVLRQXO3LáVXGVNLHJR 6NDZLQD3RODQG
1
Abstract
The results of grain refining process in the aluminium AlZnMgZr alloys containing up to 0.20% wgt. of Zr are presented. In this experiences grain refiners type AlTiB and AlTiC have been used. Investigation have shown that grain refiners type AlTiC in more resistant to poissoning by Zr and under defined casting condition will be have efficient grain refiner than AlTiB. The metallographic studies have shown that intermetallic phases type Al3(ZrxTi1-x) have occurred into microstructure. This fact would be useful to explain of Zr poisson effect during grain refineme4nt process. The laboratory scale results have been confirmed on the industry plant.
2
Introduction
It is commonly known that the grain refining process has been used in preparation of liquid aluminium and its alloys for almost fifty years, i.e. since the first paper by A. Cibula, entitled: „The Mechanism of Grain Refinement of Sand Castings in Aluminium Alloys”, published in the Journal Institutes of Metals in 1949 [1]. Despite such a long practice and almost continuous studies of the process carried out by all research centres related to the aluminium casting, no uniform and watertight theory of the grain refinement in aluminum alloys has been presented yet. Many phenomena occurring in the industrial practice prove that a comprehensive control of this process is still a distant objective. A good example of this situation may be the process of grain refining in the alloys made on the basis of aluminum and containing boron in the form of AlB2, as the so called „excess boron” [2] and in the alloys, containing Zr or Cr as the alloy additive. This situation have also refered to common aluminium alloys such as siluminum The literature data [3-16] indicate the „poisoning effect” of Zr in the grain refining processes, carried out by adding to the liquid metal the controlled amounts of Ti and B, in the form of the inter-metallic compounds, Al3Ti and TiB2, using the AlTiB master alloys. The term „poisoning effect”, describing the phenomenon of perturbation of the grain refinement process by some elements present in the alloy is commonly used in the papers devoted to grain refinement in aluminium alloys. This term is not precisely defined however. Jones and Pearson [17] use this term when the size of grains obtained after the addition of grain refiner is considerably larger in the presence of some elements, e.g. Zr, than in their absence. This means that the „poisoning element” reduces the efficiency of the grain refinement process. The presence of a „poisoning element” may cause an earlier inhibition of
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
225 the grain refinement process, which may be observed as an increase of the grain size above the acceptable level after a short time of maturing of the metal bath. Another symptom of poisoning may be an incomplete change of the columnar structure to the equi-axial one, following the addition of the grain refiner. Nowadays, there are 4 hypotheses attempting to explain the poisoning of the grain refinement by Zr, in the presence of AlTiB as the grain refiner. The hypothesis put forward by Jones and Pearson [17] assumes that the active particles of TiB2 become covered with a layer of the inter-metallic compound ZrB2, which reduces their activity during the grain refinement. This hypothesis has not been confirmed, because ZrB2 has not been found yet. On the other hand, Abdel-Hamid [18] assumes that some portion of Ti in the inter-metallic compounds, TiAl3 and TiB2, which plays an active role in grain refinement, is replaced by such elements as Zr or Cr, forming the inactive tri-component compounds, passive in the grain refinement. According to Johnsson [19], the poisoning effect of Zr is related to Ti dissolution in the inter-metallic ZrAl3 compounds. A complex (Ti1-xZrx)Al3 is formed, thus reducing Ti activity in grain refinement. A. Arjuna [20] suggests that the poisoning effect on the grain refinement is caused by the formation of the complex compounds of ZrAl3, Ti and Fe, Al3Ti, containing Zr and Fe and Al3(Fe Ti Zr), which inactivates the particles potentially active in the grain refinement processes. Between 1997 and 1999, research was conducted at the Light Metal Department of the Non-Ferrous Metals Institute concerning the process of grain refining in high-strength aluminium alloys of the AlZnMg and AlCuMg type with a Zirconium addition up to 2% of the weight. [21] The research aimed on the one hand at analysing the mechanisms of the "poisoning effect" brought about by the introduction of Zirconium into aluminium alloys and its influence on the reduction or elimination of the refining processes arrived at by the use of AlTiB master alloys, and on the other at researching ways of neutralising the negative influences of the Zirconium addition on the grain refining process. The conducted analysis brought about positive results, which allow the following theses to be posited: • a Zirconium addition up to 2% of the weight of the molten metal does not bring about a grain refining effect in the solidification process of a given type of aluminium alloy; • the presence of Zirconium in aluminium alloys does interfere with grain refining processes with the use of classical AlTiB grain refiners; • the use of AlTiC master alloys brings about the stabilisation of grain refining processes in aluminium alloys with the addition of Zirconium; • the "poisoning effect" brought about by the presence of Zirconium consists in the creation of complex compounds Al3(ZrxTl1-x), whose presence was detected in the analysed alloys, and which lead to the distortion of the optimum Ti-B proportion which is indispensable for the proper conduction of the grain refining process. The results corroborate the Johnson hypothesis in this case. It means that the formation of complex compounds Al3 (Zr, Ti, Fe) reduces the number of active Al3Ti particles, which together with TiB2 compounds are necessary for the effective grain refinement. At the current state of our knowledge, in conformity with the nucleus theory of Cibula [1] it is assumed that due to the close lattice parameters of TiCx compounds to the α- Al phase, AlTiC is a direct
226 nucleus, as opposed to the TiAl3 phase, which is activated by the presence of TiB2. As the research data show, the necessary condition for a proper grain refinement using AlTiB is the presence of both latter phases in the appropriate proportions. These results of tests conducted in laboratory conditions and in semi-technical conditions provided a basis for the successive stage of the research, i.e. applied tests conducted in an industrial environment. The present paper will discuss the results of those and the conclusions reached in the process of applying the AlTiC master alloy as the grain refiner in the process of semi-continuous production of circular φ220 mm ingots of the AlZn4.5Mg1.3ZrCr alloy.
3
Test Conduction
In the existing industrial environment, the preparation and casting of aluminium alloy ingots with diameters exceeding φ200 mm is conducted in the following manner. The melting and alloying processes are conducted in an induction channel furnace, into which a grain refiner in the form of nuglets is introduced before the pouring-in is started, as required by the existing technologic procedures. After the metal is melted and the metal bath is confirmed to have reached the chemical composition required by the factory norm, the molten metal is poured into a resistance holding furnace.
induction furnace
holding furnace
filtr
semi-cont cast unit
Figure 1: Scheme of casting unit
Then the metal is transported from the holding furnace through a foam ceramic filter of the 40ppi type to a number of DC type mould. Ingots are cast according to the following parameters: • ingot diameter - φ220 mm • number of ingots cast – 8 • casting temperature – 707°C • casting speed – 60 mm/ min • water amount 48 cu. m./ h Typical macrostructure of AlZn4.5Mg1.3ZrCr ingots produced according to the above procedure is presented in Fig. 2.
227
Figure 2: Macrostructure of AlZn4.5Mg1.3ZrCr alloy ingot
Figure 3: A schema of sample selection for microstructure analysis
Figure 4 present the microstructures detected by Barker solution etching of samples selected from the particular macros according to the schema presented in Fig. 3.
surface
½ radius
Center Figure 4: Grain in ingot
In order to improve the quality of ingots cast from AlZn4.5Mg1.3ZrCr alloy, i.e. to reduce the grain size and eliminate twinned grain, it was decided to apply AlTiC master alloy as the grain refiner in the process of producing this alloy.
228 In order to optimise the grain refining process with the use of the AlTiC master alloy, it was decided to conduct the experiment in three different variations. Variation I assumed the application of the AlTi3C0.15 master alloy in the quantity sufficient for the introduction of 100 ppm Ti, which was melted into the metal bath in an induction melting furnace. In Variation II, the same amount of the grain refiner was introduced into the molten metal stream during the pouring of the metal bath from the melting furnace to the maturation furnace. In Variation III, the AlTi3C0.15 grain refiner was introduced into the molten metal stream, as in Variation II, but its amount was increased to 150 ppm Ti. The casting process parameters remained intact and reflected the existing technological procedure regarding etching of controlled melts. Examples of macrostructures of ingots cast during controlled melts in the three variations are presented in Fig. 5. Figures 6-8 present the micostructures of the samples separated from the macros (Fig. 5), and selected according to the schema in Fig. 3 The samples were etched by Barker solution and observed in polarised light in order to detect grain distribution and size in the analysed microstructure.
variant I
variant II
variant III Figure 5: Mactrostructure of AlZn4.4Mg1.3Zr0.17Cr0.15 alloy ingot with AlTiC as grain refiner
229
850 µm Surface
850 µm ½ radius
850 µm Center Figure 6: Grain in ingot - variant I
850 µm
Surface
850 µm
½ radius
850 µm center Figure 7: Grain in ingot – variant II
230
850 µm Surface
850 µm ½ radius
850 µm
center Figure 8: Grain in ingot – variant III
In the course of the holding of the molten metal and its casting, molten metal samples were taken according to the TP-1 test procedure adapted for the purpose. Examples of sample macrostructures from the TP-1 test and the macrostructure typical for the entire experiment are presented in Fig. 9.
Figure 9: Mactrostructure of Test TP1 samples
231 This stage of the analysis aimed at analysing the possible effects of grain refining reduction during the lengthy process of preparing and casting the molten metal.
4
Results
The results obtained during the analysis of AlZn4.5Mg1.3Zr0.17Cr0.15 ingot production, conducted in an industrial environment, reveals that the application the AlTiC type grain refiner in the process of producing alloys with a Zr addition assures the obtaining of properly refined grain throughout the length of the cast ingot. These results fully confirm the results obtained in laboratory conditions and in semi-technical conditions. The analysis also demonstrated that the existing procedure of introducing the grain refiner to the metal bath in the induction channel furnace is improper. This finding is demonstrated by the comparison of the macrostructure of the ingots analysed in Variations I and II of the experiment described here. The same amount of grain refiner is observed to be much more efficient when introduced into the molten metal stream during its pouring from the melting furnace to the holding furnace. This is demonstrated by the total elimination of the twin crystals growth zone in the ingots (the comparison of the microstructures presented in Fig. Fig. 6 and 7 above), which were produced from the molten metal produced according to the Variation II procedure. This finding was also confirmed for the preparation and casting processes of other alloys. It has to be emphasised, however, that the change concerns the place where the refiner is introduced, rather than its composition. It was also found that there is no rationale for the use of higher amounts of AlTiC master alloy to be introduced. The comparison of microstructures (Fig. 7 and Fig. 8) does not reveal considerable differences in grain distribution and size when 100 and 150 ppm Ti are revealed by the AlTi3C0.15 master alloy. The analysis of the samples taken with the use of the TP-1 Test procedure (Fig. 9) manifests that the grain refining effect obtained thanks to the use of the AlTiC master alloy is not reduced even when the time span between the introduction of the master alloy to the casting of the last part of the molten metal exceeds 200 minutes. In summary, we can conclude that the analysis of AlZn4.5Mg1.3Zr0.17Cr0.15 molten metal production in an industrial environment fully corroborate the efficiency of the application of the AlTiC type grain refiner for this type of alloys. The research also brought about a modification of the existing procedures of molten metal preparation and the casting of aluminium alloy ingots with the addition of Zirconium.
5
Conclusion
1. Grain refining in aluminium alloys with the addition of Zirconium should be conducted with the application of the AlTiC master alloy, which was confirmed by analyses in semitechnical and industrial conditions. 2. If the grain refining process is conducted properly as far as the place of application and the amount of the AlTiC master alloy is concerned, homogeneous fine-grain structure can
232 be reproduced regularly throughout the cross section of the aluminium alloy ingot with the addition of Zirconium, and the twin crystal zone can be eliminated.
6
References
[1] Cibula, The Mechanism of Grain Refinement of Sand Castings in Aluminium Alloys Journal Institutes of Metals, 1949 [2] - :R QLFND 7 6WXF]\ VNL 5R]GUDEQLDQLH ]LDUQD Z VWRSDFK Z\WZDU]DQ\FK QD ED]LH aluminium oSRG\Z V]RQHM ]DZDUWR FL ERUX 0DWHULDá\ .RQIHUHQF\MQH Ä$/80,1,80 98”, Zakopane 1998 [3] S. R. Thistlethwaite, Recent Developments in Grain Refiner Technology, 4th Australasian Asian Pacific and Conference Aluminium Cast House Technology, The Minerals, Metals & Materials Society, 1995 [4] M. Johnsson, Influence of Zr on the Grain Refinement of Aluminium Carl Hanser Verlag, München 1994 [5] )DUULRU '& %ULOOKDUW $ 3UDFWLFDO 0HWKRG IRU (YDOXDWLQJ *UDLQ 5HILQHPHQW 0DWHULDá\ firmy Kawecki Berylco [6] P.S. Mohanty, J.E. Gruzleski, Mechanism of Grain Refinement in Aluminium Acta metall. mater. Vol. 43, no 5 pp. 2001-2012, 1995 [7] P. Hoefs, W. Reif, Recent Developments in Grain Refining of Aluminium and Aluminium Alloys Solidification of Metals and Alloys, No 28, 1996 [8] R. Cook, P.S. Cooper ,Benefits of Master Alloy Melt Treatments in the Aluminium Foundry Industry Light Metals 1996 [9] R. By, A.P. Fielding, Recent Developments in Grain Refining Technology & Standardization Light Metals Age, June 1997 [10] J.A. Spittle, S.B. Sadli, The Influence of Zirconium and Chromium an the Grain Refining Efficiency of Al-Ti-B Inoculants Cast Metals, V. 7, 1994 pp 247-253 [11] M.A. Kearns, P.S. Cooper, Effects of Solute Interactions on Grain Refinement of Commercial Aluminium Alloys Light Metals Age, June 1997 [12] P.C. van Wiggen Al-Ti-B Grain, Refner - the Consistent Ingredient Light Metals Age, June 1997 [13] P. Hoefs, W. Reif, Development of an Improved AlTiC Master Alloy for the Grain Refinement of Aluminium Light Metals Age, June 1997 [14] M.A. Hadia, A.A. Ghaneya, Development and Evaluation of Al-Ti-C master Alloys as Grain Refiner for Aluminium Light Metals 1996 [15] M.A. Kearns S.R. Thistlethwaite,Recent Advances in Understanding the Mechanism of Aluminium Grain Refinement by TiBAl Master Alloys Light Metals 1996 [16] 36&RRSHU0DWHULDá\.RQIHUHQF\MQHPDU]HF 706&RQIHUHQFH6DQ$QWRQLR [17] G.P. Jones, J. Pearson, Metall Trans B 1976, 7B, 223-234 [18] A.A. Abdel-Hamid, Z. Metallkd 80, 1989, 566-569 [19] M. Johnsson, Z. Metallkd 85, 1994, 786-789 [20] Arjuna Rao, Materials Science and Technology, Sept. 1997, vol. 13 769-777 [21] 76WXF]\ VNL0/HFK*UHJD7KH*UDLQ5HILQLQJ3URFHVVLQ+LJK6WUHQJWK$OXPLQLXP Alloys Containing Addition of Zr, Proceedings of the International Conference "Light Alloys and Composites", 13-16 May 1999, Zakopane, Poland
Coupled Influence of Convection and Grain-refining on Macrosegregation of 1D Upwardly Solidified Al 4.5%Cu Ph. Jarry1, H. Combeau2 and G. Lesoult2 1
Pechiney Centre de Recherches de Voreppe, France LSG2M, Ecole des Mines de Nancy, France
2
1
Abstract
It has been shown in previous studies that even in the simple configuration of a 1D upward solidification set-up, inverse macro-segregation is influenced by convective phenomena. In order to better identify the respective contributions of dendrite settling, dendrite transport and thermo-solutal convection to inverse segregation in the vicinity of the chill, a comparison has been made between Al4,5%Cu ingots obtained with or without grain refinement, and solidified in a 1D upward solidification set-up either located in the laboratory at room temperature, or within a heated oven in order to inhibit convection. The macrosegregation results evidence the combined effects on the macro-segregation pattern, of dendrite mobility, as governed by the grain refinement and Ti concentration, and of convective behavior, as induced by the temperature around the mold.
2
Introduction
In order to get a better understanding of the parameters involved in inverse segregation buildup of 1-D solidified aluminum ingots, a few original experiments have been designed. Since it has been demonstrated [1] that the interaction of convection with equiaxed grains does bring a contribution to the concentration profile, it was natural to try to strongly vary the parameters governing grain mobility on the one hand, and convection on the other hand.
3
Experimental Procedure
In order to do so, a 1-D solidification set-up was used which allows to bottom-cool the liquid metal contained in an insulated cylindrical mold (50mm in diameter, 150mm in height), itself placed inside an oven maintained at a chosen temperature. The chill is a water-cooled block of copper and the cooling water is injected from outside the oven into the block. In order to act upon the formation and morphology of equiaxed grains, the grain refinement was varied between 0 and 3 kg/T of Al5%Ti1%B, and in some cases a hyper-peritectic quantity of Ti (0.2%) was introduced to favor the growth of Al3Ti at the liquidus and hence the settling of grains of solid solution of aluminum grown over those primary intermetallic particles. Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
234 Classical optical metallography was performed. The Cu concentration profiles were measured along the central vertical axis of the ingots on longitudinal polished sections by analytical scanning electron microscopy (ASEM) between 15 and 20kV accelerating voltage; for each position along the axis of the ingots, a 200µmx200µm square was scanned. The complete procedure is described in [1].
4
Results
4.1
Terminology
In the following, ingots1D-solidified in molds located in the laboratory at room temperature will be referred to as “RT-solidified”, whereas the ingots solidified in the 1D set-up placed within a heated oven will be referred to as “oven-solidified”. 4.2
Grain Structure and Sizes
The grain structures obtained are summed up in Table 1. Table 1. List of ingots studied with solidification conditions and type of as-cast structure Sample Grain refining T around Grain structure reference (with Al 5Ti 1B) mold during freezing # 18 3 kg/ton + 0.20%Ti 750° Equiaxed (400µm close to the chill) # 19 not inoculated 680°C Coarse-grained equiaxed (2500µm) # 20 3 kg/ton + 0.20%Ti 680°C Equiaxed (600µm close to the chill) #1 not inoculated 20°C Columnar then coarse-grained equiaxed #3 0,1kg/ton 20°C Coarse-grained equiaxed #5 0,3kg/ton 20°C Fine grained equiaxed #6 0,3kg/ton 20°C Fine grained equiaxed #7 3kg/ton 20°C Fine grained equiaxed #9 3kg/ton + 0.2% Ti 20°C Fine grained equiaxed 4.3
Copper Concentration Profiles
As previously described [1,2] a continuously decreasing Cu concentration over the length of the ingot (except sometimes at the upper end) is obtained in the case of the grain refined Al4,5Cu ingots solidified under usual conditions (environment at room temperature), see Figure 1a. A prolonged positive segregation is observed up to 50 to 75mm from the chill. A concentration plateau at (or near) the nominal Cu concentration is obtained in the well known case of columnar growth (ingot 1), or in the case of the very little grain-refined (0,1kg/ton) ingot #3 with coarse equiaxed grains. A striking feature is that a plateau is also obtained in ingots solidified within a heated oven regardless of the grain refinement, see Figure 1b. In this case the extension of the zone with positive segregation on the chill side is typically 25mm. The samples with hyperperitectic Ti follow the same trend except when solidified in a mold at room temperature: the Al4,5%0.2%Ti sample exhibits a more complex
235 concentration profile, see Figure 1a: after a very similar behavior up to 25mm, a positive plateau is obtained followed by a rather sharp drop in concentration. 0.2
0.1 0.05 0 0
25
50
75
100
125
ingot 19 : 0kg/T 680°C ingot 20 : 0.2Ti 3kg/T 680°C
0.15
150
-0.05 -0.1
relative segregation
0.15 relative segregation
0.2
ingot 6 : 0.3kg/T 20°C ingot 5 : 0.3kg/T 20°C ingot 7 : 3kg/T 20°C ingot 9 : 0.2Ti 3kg/T 20°C
ingot 18 : 0.2Ti 3kg/T 750°C ingot 1 : 0kg/T 20°C
0.1
ingot 3 : 0.1 kg/T 20°C 0.05 distance from the chill (mm)
0 0
25
50
75
100
125
150
-0.05 -0.1
-0.15
-0.15
distance from chill (mm)
Figure 1: Copper concentration profiles along the vertical axis of the ingots; a/ profiles obtained for equiaxed Al4.5%Cu samples in molds at room temperature; b/ profiles obtained for samples with no or little inoculation at room temperature or for “oven solidified” samples, whatever the inoculation.
4.4
Comparison of Macrosegregation Intensities
The compared intensities of macrosegregation in the different ingots are summed up in Figure 2, showing the average of absolute values of relative segregation in copper measured between 12mm and 137mm from the chill. These limits have been chosen in order to eliminate both the zone affected by chill inverse macrosegregation, where the concentrations measured are very scattered, due to the discrete detection of the presence of constituent phases, and a symmetrical zone at the top of the ingots affected by shrinkage porosity. This bar graph clearly shows that equiaxed samples solidified under usual conditions are twice as segregated as either non-inoculated samples or grain refined samples solidified in molds held at high temperature. ingot 6 : 0.3kg/T 20°C ingot 7 : 3kg/T 20°C ingot 5 : 0.3kg/T 20°C ingot 9 : 0.2Ti 3kg/T 20°C ingot 3 : 0.1 kg/T 20°C ingot 20 : 0.2Ti 3kg/T 680°C ingot 19 : 0kg/T 680°C ingot 18 : 0.2Ti 3kg/T 750°C ingot 1 : 0kg/T 20°C 0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
Figure 2: Average of the absolute values of relative segregation measured between 12 and 137mm from the chill for the different ingots. Striped bars illustrate the influence of convection on the macrosegregation behavior of the same highly grain refined alloy. The alloy with 0.1kg/T grain refiner exhibits an intermediate behavior, because it is equiaxed but with a coarse grained structure that reduced the mobility of dendrites.
236
5
Discussion
5.1
Driving Forces at Work
Once the well-known chill inverse segregation transient is finished, a steady state regime could ideally take place in which liquid migration towards the cold isotherms would exactly compensate for volumetric contraction, in such a way that the resulting mean Cu concentration would equal the nominal concentration. Now, any mechanism that can either bring more grains to the front, take more grains from the front, or modify the flowing regime of the eutectic-enriched liquid near the mushy zone will certainly alter the Cu concentration profile of the solidified ingots. A very likely possibility stemming from [1,2] is that thermosolutal convection movements suspend crystals in the undercooled liquid, removing them from the front or preventing them from settling onto the front, a mechanism likely to prolong solidification at a Cu concentration higher than the nominal one. This mechanism is very likely to occur in the case of grain refined ingots solidified under usual conditions and exhibiting a steadily decreasing Cu concentration starting from the chill. A good clue that this mechanism actually brings such a contribution, is that grain refined ingots solidified within an oven, where convection movements are inhibited, exhibit the concentration plateau usually known to form in columnar solidification. Potentially mobile equiaxed dendrites are no longer prevented from sitting on the solidification front. In the case of Al4.5Cu0.2Ti, dendrite transport is probably enhanced by the precipitation of Al37LDQGDVVRFLDWHGSULPDU\ $OXPLQXP 5.2
Dendrite Mobility
Dendrites respond to convective driving forces through their mobility, itself linked to grain refinement. Grain refinabiliy [4] can be understood through the notion of gradient of solid fraction fs near the liquidus which is equal to the product of dfs/dT at the liquidus temperature by the thermal gradient. The increase in solid fraction when temperature decreases under the liquidus can be derived from the Gulliver-Scheil law for linear approximations of the liquidus lines of the different binary alloys corresponding to each alloying element: df − s dT
= Tliquidus
∑
1 m i (k i − 1)C oi
where mi is the liquidus slope at the liquidus temperature in each binary alloy Al-i, ki the corresponding partition coefficient, and Coi the nominal concentration of element i. Dendrite mobility can thus be expressed by: df µ d = G th. s dT
liquidus
−1
=
∑ m (k i
i
− 1)C oi
G th
It is well known [3] that among the different alloying elements of aluminium, Titanium plays a very special role, linked with the high value of the product m(k-1) for Ti in Al: m(k1)§. The weaker the increase in solid fraction near the liquidus, the easier the nucleation and growth of equiaxed grains. Now, if both the grain refinability is high and the temperature gradient is low close to the front, nucleation and growth have room to occur in the undercooled region and convection movements can affect the suspended crystals.
237 Table 2: Compendium of mechanisms accounting for the observed profiles. Grain refined Non or little grain refined Mold at RT Mold in oven Mold at RT Mold in oven Convective driving force Yes Little or none Yes Little or none Dendrite mobility Suspension of dendrites
High
Settling of dendrites Segregation index*
Mediocre (high gradient)
Low (poor G R)
Low
None or little
No
No
Yes, at the end; Al3Ti may play a role
None or little
May play a role after C.E.T.
No
§
§
§
§
Yes, at the beginning
* Average of absolute values of relative segregation between 12 and 137mm
On the other hand, under too high temperature gradient conditions, no room is available for nucleation on a large number of sites, or for a large number of equiaxed grains to grow: fewer grains, thus larger ones, are formed, of impaired mobility. As a consequence, by using a mould held at high temperature, not only the driving force for natural convection is decreased or suppressed, but also the dendrite mobility is decreased because the temperature gradient is strongly increased at the front; it is true even for the highly refinable alloys with high Titanium concentration and strong inoculation. This framework allows qualitative interpretation of macrosegregation results, see Table 2.
6
Conclusion
The influence of dendrite movement on macrosegregation profiles in upwardly solidified ingots of Al4.5%Cu is confirmed. Dendrite mobility is boosted by grain refinement and Ti content but is inhibited by too high temperature gradients. When dendrite mobility and convection can have a coupled action, a prolonged positive segregation starting from the chill is observed, steadily decreasing over the length of the ingots, then turning into a negative segregation at around mid-height of the ingots. Suppression of dendrite mobility or of the convective driving force results into a macrosegregation plateau type profile typical of what is observed or calculated for columnar growth, even with grain refined samples. As similar phenomena operate within the far more complex context of vertical continuous casting, the presented results are believed to help provide a reference of mechanisms for qualitatively interpreting the macrosegregation patterns obtained in DC Casting.
238
7
Acknowledgments
The technical contribution of J.-M. Burtin and V. Chastagnier, Pechiney Centre de Recherches de Voreppe, is gratefully acknowledged. Part of this work was conducted within the framework of Brite Euram III Project BE-1112 (European Modeling Program on Aluminium Casting Technology).
8
References
[1] V. Albert, Ph.Jarry, H. Combeau, and G. Lesoult, “Influence of the grain structure on macrosegregation in Al-Cu ingots obtained by upward 1-D solidification,” presented at SP'97, Sheffield, 1997. [2] V. Albert, “PhD work: Macroségrégation et mouvement des cristaux équiaxes lors de la solidification d'alliages,” Ecole des Mines de Nancy, Institut National Polytechnique de Lorraine, 1998. [3] J. Moriceau, “Discussion des mécanismes d'affinage de l'aluminium par le titane et le bore,” Revue de l'aluminium, pp. 977-988, 1972. [4] P. Desnain, Y. Fautrelle, J. L. Meyer, J. P. Riquet, and F. Durand, “Prediction of equiaxed grain density in multicomponent alloys stirred electromagnetically,” Acta Metallurgica, vol. 38, pp. 1513-1523, 1990.
Tensile Behaviour of DC-cast AA5182 in Solid and Semi-solid State W.M. van Haaften, W.H. Kool, L. Katgerman Laboratory of Materials, Delft University of Technology, Rotterdamseweg 137, 2628 AL, Delft, The Netherlands
1
Abstract
Hot tearing is still one of the major problems during DC casting of aluminium alloys. It may result in cracks running through the entire ingot, which require costly remelting and cause loss of productivity. Numerical models are developed to simulate and improve the casting process but there is a lack of mechanical data of the alloys, especially at elevated temperatures. Therefore tensile tests were carried out with AA5182 at high temperature including the semisolid range. A mathematical description was found for the subsolidus temperatures and this description could be extended to above-solidus temperatures by assuming that the constitutive behaviour is determined by the solid network. This was done by considering not only the fraction liquid but also the percentage of grain boundary area covered by this liquid phase. Taking this approach, the mechanical behaviour of the alloys at semi-solid temperature can be described, at least for the low liquid fractions during which hot tearing occurs.
2
Introduction
Nowadays, a wide variety of aluminium alloys is being cast on a routine base by means of the DC casting process. However, in search for a better corrosion resistance or a higher strength for example, new compositions are being developed which may impose strict constraints on the process window. Not only the process parameters such as casting speed, casting temperature, cooling rate should be controlled accurately, also a profound knowledge of thermomechanical behaviour of the alloy during the casting process is required to produce a defect free slab or billet. This study focuses on this thermomechanical behaviour at high and even semi-solid temperatures. The main defect which occurs at these high temperatures during casting is hot tearing. Hot tears are cracks which initiate during the last stage of solidification, i.e. in the mushy zone, where solid and liquid co-exist. The hot tearing problem has since long been investigated [1-5] and many cracking criteria have been defined. From these studies it has become clear that the two key parameters are 1) inadequate interdendritic feeding and 2) stress and strain development in the dendritic network. The aim of this research is to determine the stresses that can be supported by the solid network at different liquid fractions and to find a mathematical description for the alloy behaviour in this semisolid state. As a starting point we will use the equation often applied to describe solid state creep:
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
240 Q (1) ε& = Aσ n exp − RT where ε& is the plastic strain rate, A is a constant, σ is the true stress, n is the stress exponent,
Q is the apparent creep activation energy, R is the universal gas constant and T is the temperature.
3
Experimental
The material investigated is an AA5182 alloy with the following composition: Mg: 3.6, Mn: 0.16, Si: 0.21, Fe: 0.26, Al: bal. (wt%). It is sampled from a rolling slab which was DC cast at an industrial research facility [6]. The slab did not receive any additional heat treatment after casting. The tensile tests are carried out with a Gleeble 3500 thermomechanical simulator. The specimens with the same shape as used earlier [7] are taken from the slices with their tensile direction parallel to the casting direction. They are secured between two water-cooled jaws of which one is moving. The specimen is heated at 50°C/s via Joule heating during which it is kept at zero force. The test temperatures are from 370°C to 580°C. The solidus temperature of the material is 510°C and the maximum liquid fraction, calculated with the Alstruc model is 0.07 [8,9]. True strain εp is measured with a dilatometer which monitors the diameter D of the specimen: εp = -2 ln(D/D0). Strain rates are from 3·10-5 to 3·10-3 s-1, which are comparable to values in the DC casting process. Most experiments were carried out in force control. Only at 560°C and 580°C the tests were carried out in stroke control as at these temperatures the measured force becomes too low for accurate control. The solid state creep results were fitted to Eq. 1. The specimens tested at semi-solid temperatures often fractured and their fracture surface was studied by SEM.
4
Results
The stress vs. strain rate data measured by the tensile tests are shown in Fig. 1. The strain rate increases with increasing stress and increasing temperature. At the lower temperatures steady state was reached quickly and the corresponding stress and strain rate were plotted in Fig. 1. In the semi-solid temperature range, steady state was not always achieved and in that case the peak stress with the corresponding strain rate was used. The data obtained at 370°C - 470°C were used to determine the parameters of the creep equation (Eq. 1, Table 1). The fracture surface of a specimen broken at 560°C, i.e. with a fL = 0.04, shows dendrite arms covered with a thin film, which was liquid at the time of fracture (Fig. 2). It also shows some rough regions due to solid state rupture. The fracture mechanism proposed [10] is that of fluid film separation and rupture of solid bridges. To describe the material behaviour in this semi-solid range the creep law is modified by taking into account the solid network only (see next section).
241
420
-1
Strain rate (s )
370
10
470
-3
500 510 520 540
10
-4
550 556 560 580
10
-5
1
10
Stress (MPa) Figure 1: Strain rate vs. stress at the temperatures investigated
Table 1: Parameters of the creep law (Eq. 1) for AA5182 in solid state. Parameter Value A 30 (Mpan s)-1 Q 120 kJ/mol n 3.3
Figure 2: SEM micrograph of fracture surface. Fracture took place at 560°C. fL = 0.04. SB1,2 = solid bridge
242 1.0 0.8
fLGB
0.6 0.4 0.2 0.0 0.00
0.05
0.10
fL Figure 3: Fraction of grain boundary area covered with a liquid film vs. total fraction liquid. Wetting angle θ = 0°
5
Discussion
To describe the thermomechanical behaviour of the alloy in semi-solid state it was assumed that the liquid cannot carry any load and that the load is transferred to the existing solid dendrite network. The same assumption was made previously by Drezet and Eggeler [11] and they modified the creep law by dividing the stress by the fraction solid (fS). In this way they explained the high apparent activation energies observed in mushy zone behaviour. However, this modification of the creep law cannot explain the sudden drop in strength as observed in our experiments when passing the solidus temperature. Therefore we propose a further modification of the creep law by taking into account the fraction of the grain boundary area which is covered by liquid. The resulting equation is: σ ε& = A 1 − f LGB
n
Q exp − RT
(2)
where fLGB is the fraction of grain boundary area covered with liquid. Assuming that the liquid is well wetting and that the liquid pockets have a tetrahedral symmetry the fLGB can be calculated from the fraction liquid fL if the liquid fraction is small [12]. The result is shown in Fig. 3 and indicates that the fraction of the grain boundary area covered by a liquid film rises steeply with increasing liquid fraction. In Fig. 4 the normalised stress is plotted vs. temperature for all experimental data. The normalised stress is the ratio of the measured stress and the stress predicted by Eq. 1 for the same temperature with the parameters listed in Table 1. Clearly, a steep decrease in stress level can be observed when entering the semi-solid range. Also shown are the predictions by the modified creep law. The thin line indicates the modification by [11] i.e. dividing the stress in the creep law by the fraction solid fS. The thick line is the prediction by Eq. 2 and follows the experimental data quite well. This indicates that the creep behaviour of this alloy can be extended to the semi-solid range by taking into account the fraction of grain boundary area covered with liquid. Deviations between experimental data and the model are due to the following reasons. Firstly, experimental inaccuracies may arise from the low forces and low strain rates involved.
243 Secondly, small variations in temperature lead to variations in liquid content and to large variations in the fraction of liquid grain boundaries, which have a strong effect on the strength of the material. Thirdly, creep data are extrapolated over a large temperature interval.
Normalised stress
Tsol 1
0.5
error bar 0 350
400
450
500
550
600
Temperature (°C)
Figure 4: Normalised stress vs. temperature. ∆ strain rate 10-3, ◊ strain rate 10-4, × strain rate 10-5. Thin line: creep law modified with fS, thick line: creep law modified with 1- fLGB (Eq. 2). Error bar holds for semi-solid temperatures where the uncertainty is large due to reasons explained in the text.
From Fig. 4 there are indications that in the semi-solid range the normalised stress is somewhat lower for lower strain rates. It is noted that at high strain rate the liquid will mainly be present as drops at triple junctions which leaves a mainly solid grain boundary area, whereas at low strain rates, there is more time for the liquid to spread over the grain boundary area. This is further promoted by the tensile stress state, which will enlarge the effect. Such an effect is not accounted for in the simple model presented and will lead to a stress dependent term in the calculation of the fraction of liquid grain boundaries. The Alstruc model was used to calculate the fraction liquid curve of the material after solidification. The cooling rate was taken as 3°C/s, representative for the specimen location in the ingot. Back diffusion of Mg into the grains will take place during cooling of the ingot to room temperature, thereby removing low melting eutectic phases. This was also taken into account. Reheating before the tensile test was so fast that further diffusion was neglected. However, the cooling rate during solidification strongly effects the liquid fraction curve and the solidus temperature. The prediction by Eq. 2 is thus also strongly dependent on the cooling rate. Taking a cooling rate of 1°C/s for example, the solidus temperature shifts to 550°C.
6
Conclusion
By taking into account the fraction of grain boundary area covered with liquid, the creep law can be used to describe the tensile behaviour of AA5182 during the last stage of solidification. The underlying assumption (i.e. a load bearing solid network with liquid in between) is supported by microstructural observations of the fracture surface which shows the presence of a thin liquid film and the remains of solid bridges.
244
7
Acknowledgements
This research was carried out as part of the EMPACT Brite-Euram project (BRPR-CT950112). Funding by the European Committee is gratefully acknowledged. The co-operation with Mr. T. Iveland (Hydro Aluminium, Norway) on the Gleeble 3500 was very fruitful. We thank Ms. A.L. Dons (Sintef, Norway) very much for the Alstruc calculations and VAW (Germany) for providing the material.
8
References
[1] W.S. Pellini, Foundry, 1952, 80, 125-199. [2] U. Feurer, Giessereiforschung, 1976, 28, 75-80. [3] T.W. Clyne, G.J. Davies in Proc. Conf. Solidification and Casting of Metals, Sheffield, UK, 1979, 275-278. [4] L. Katgerman, Journal of Metals, 1982, 34, 46-49. [5] M. Rappaz, J.-M. Drezet, M. Gremaud, Metall. Trans. A 1999, 30A, 449-455. [6] VAW, Germany. [7] W.M. van Haaften, B. Magnin, W.H. Kool, L. Katgerman in Light Metals 1999 (Ed.: C.E. Eckert), TMS, Warrendale, PA, USA, 1999, 829-833. [8] Private comm. with A.L. Dons (Sintef, Norway), E.K. Jensen (Elkem Research, Norway) and A. Hakonsen (Hydro Aluminium, Norway). [9] A.L. Dons, E.K. Jensen, Y. Langsrud, E. Trømborg, S. Brusethaug, Met. Trans. A, 1999, 30A, p. 2135-2146. [10] W.M. van Haaften, W.H. Kool, L. Katgerman, Materials Science Forum, 2000, 331-337, 265-270. [11] J.-M. Drezet, G. Eggeler, Scr. Metall. et Mater. 1994, 31, 757-762. [12] P.J. Wray, Acta Met. 1976, 24, 125-135.
The Columnar to Equiaxed Transition in Horizontal Direct Chill Cast Magnesium Alloy AZ91
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
246
247
248
249
250
Study of Heterogeneous Nucleation of α-Al on Grain Refiner Particles during Rapid Solidification P. Cizek, B.J. McKay and P. Schumacher University of Oxford, Oxford, UK
1
Introduction
In the Al casting industry, it is common practice to add grain refiner particles to the melt in order to promote nucleation of α-Al and, thus, to ensure grain refinement of the as-cast microstructure. Commercial Ti-B-Al grain refiner master alloys contain TiB2 and Al3Ti particles, which serve as potential sites for the heterogeneous nucleation of α-Al during casting. Attempts to relate the microstructure of these master alloys to their grain refining performance have proven inconclusive, as conventional grain refinement tests have a difficulty to identify a nucleation mechanism for α-Al conclusively [1]. Recent investigations involving Al-based metallic glasses [2], as slow-motion analogues to the undercooled Al melts, have revealed thin Al3Ti DO22 layers adsorbed on the surfaces of TiB2 particles. Nucleation of α-Al was found to occur epitaxially on these layers on the basal faces of boride particles, with a low-index orientation relationship (OR) [2]. The aim of the present work was to elucidate whether the Al3Ti layers originate directly from grain refiner master alloys and whether they undergo changes during re-melting of these alloys. Rapid quenching, using melt spinning, was used to study the nucleation behaviour of α-Al and Al3Ti on boride particles and the thermal stability of the aluminide layers. The resulting large undercoolings facilitated multiple nucleation events on individual boride particles, as a result of small critical nucleus sizes, in contrast to small undercoolings obtained during conventional solidification.
2
Experimental Methods
A commercial Ti-B-Al 5:1 wt.% grain refiner rod (supplied by London & Scandinavian Metallurgical Co. Ltd.) was used as an experimental material. Rapid solidification of the rod was achieved by melt spinning in an inert helium atmosphere. The charge was heated to 1050°C in a boron nitride crucible and held for 5 minutes before ejecting onto a copper wheel rotating at a speed of 20 ms-1. Resulting ribbon was approximately 3 mm wide and 30 µm thick. Both the as-received grain refiner rod and the melt-spun ribbon were examined using TEM. Thin foils were made by electropolishing using a solution of 25% nitric acid and 75% methanol at a temperature of approximately –30°C at 20 V. TEM investigation was performed using a Philips CM20 microscope operated at 200 kV.
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
252
3
Results and Discussion
3.1
As-Received Ti-B-Al Grain Refiner Rod
The microstructure of the grain refiner rod contained TiB2 and Al3Ti particles randomly distributed within the Al matrix. The boride particles exhibited a facetted, hexagonal platelet morphology with the {001} and {100} planes as their external faces (Fig. 1).
Figure 1: TiB2 particle, found in the grain refiner rod, embedded in an Al matrix: (a) TEM bright-field micrograph; (b) dark-field micrograph obtained from the streak located between the center and (001) spots which shows a thin layer (arrowed) coating the particle; (c) SAD pattern showing the streaks (arrowed) and indices of the corresponding diffraction spots (zone axis [100]TiB2, subscripts A and B indicate the Al and TiB2 crystal lattices respectively); (d) same as in (a) after re-tilting by several degrees to obtain the neighboring Al grains in contrast.
Their width-to-thickness ratio was approximately 4:1 with the width of the majority of the particles being less than one micron. Corresponding diffraction patterns (Fig. 1c) confirmed a TiB2 hexagonal crystal structure with parameters a = 0.303 nm and c = 0.323 nm [3]. Boride particles were embedded in a matrix composed of fine Al grains (Fig. 1d), which were largely separated by high-angle boundaries, with no apparent distinct OR observed between the borides and the neighboring Al grains (Fig. 1c). The above fine Al grains appeared to be a result of recrystallisation of the as-solidified α-Al matrix, which occurred during the
253 manufacturing process of the refiner rod. Randomly orientated fine recrystallised grains may have replaced the original as-solidified large grains, thus removing any possible pre-existing well-defined OR between the borides and the neighboring aluminium matrix. Selected-area diffraction (SAD) patterns obtained from the boride particles showed two sets of streaks perpendicular to the (001) and (010) planes (Fig. 1c). These streaks are indicative of a thin layer covering both their basal and non-basal faces. The presence of the layer, about 1 nm in thickness, was confirmed by dark-field imaging, in which the objective aperture was placed on the streak between the center and the (001) spot (Fig. 1b). This suggests that there is a strong similarity between the layers found in the grain refiner rod and those observed on TiB2 particles within the Al-based glasses [2] and identified as Al3Ti DO22 phase. This finding, indicating that the Al3Ti layer might be already present in the rod, is significant as it shows that the layer is not an artifact of the metallic glass technique. It also indicates that, in most instances, the aluminide layer does not have to be formed after adding the refiner rod to the melt, although it might be enhanced or depleted [4]. The Al3Ti particles in the rod were rather small, largely less than 2 µm in diameter, and displayed roughly equiaxed shapes (Fig. 2a). Corresponding diffraction patterns (Fig. 2b) confirmed the equilibrium Al3Ti DO22 crystal structure with the parameters a = 0.385 nm and c = 0.861 nm [5].
Figure 2: Al3Ti particle found in the grain refiner rod and related to one of the neighboring α-Al grains (labeled) by a near cube-to-cube OR: (a) TEM bright-field micrograph; (b) corresponding SAD pattern showing the equilibrium Al3Ti DO22 phase (zone axis [100]).
These particles were frequently related to Al grains by a well-defined, close to the cube-tocube, OR (Fig. 2b). Arnberg et al. [3] have identified three distinct morphologies for aluminide particles (block, flake and petal) in the as-solidified grain refiner master alloys. During casting, these morphologies were found to have a marked influence on the grainrefinement contact-time as a result of differing dissolution times. The roughly equiaxed Al3Ti morphology, observed in the present study, differs from the as-solidified shapes [3] and could be a product of recrystallisation which occurred during the rod manufacturing process. It is possible that the observed well-defined OR between the aluminides and the aluminium matrix was already present in the as-solidified material and was maintained during the recrystallisation process.
254 3.2
Melt-Spun Ribbon
The microstructure of the melt-spun ribbon was composed of fine equiaxed Al grains, largely less than 2 µm in diameter. A majority of these grains contained small petal-shaped particles, with diameters less than 0.5 µm, in their interiors (Fig. 3a). Diffraction analysis (Fig. 3b) indicated that the petal-shaped particles corresponded to metastable Al3Ti, having an ordered Ll2 crystal structure with the lattice parameter a = 0.400 nm [6], as expected for rapid quenching of the present hyper-peritectic Al-Ti alloy. The particles systematically displayed an epitaxial, near cube-to-cube OR with Al (Fig. 3b), which has also been reported in [6]. This suggests that the metastable Al3Ti Ll2 particles, after their formation from the melt, served as potent substrates for subsequent heterogeneous nucleation of α-Al, presumably due to the negligible lattice mismatch between the corresponding crystal lattices.
Figure 3: Petal-shaped metastable Al3Ti Ll2 particle, found in the melt-spun ribbon, located in an α-Al grain center: (a) TEM bright-field micrograph; (b) SAD pattern obtained from the entire grain area, showing a near cube-to-cube OR between the respective crystal lattices (subscripts A and P with the diffraction spot indices indicate the Al and Al3Ti Ll2 lattices respectively).
TEM analysis showed that the thin, presumably Al3Ti DO22 [2], layer covering the TiB2 particles was inherited from the refiner rod and remained essentially unaffected by the high temperature during melt spinning. As shown in Fig. 4, the boride particles were systematically observed to nucleate α-Al, as well as the metastable Al3Ti Ll2 phase, each with the same lowindex OR, found previously using the metallic glass technique [2] In this OR, the {001}, {112}, and {111} planes and the <100>, <110> or <201>, and <110> directions corresponding to the TiB2, Al3Ti DO22 layer and Al (or metastable Al3Ti Ll2 structure), respectively, are approximately parallel to each other [2]. Several α-Al grains were frequently found to nucleate on individual boride particles and they displayed either one or the other of the two possible orientation variants of the above OR (Fig, 4). Thus, borides covered with a thin aluminide layer appeared to serve as potent substrates for heterogeneous nucleation of α-Al (as well as the structurally similar, metastable Al3Ti Ll2 phase), which is in agreement with the findings of the metallic glass technique [2]. This appears to be consistent with observations in conventional casting practice showing that excess Ti (beyond TiB2 stoichiometry), which facilitates the formation of the aluminide layers, is necessary for an effective grain refinement [2,4].
255
Figure 4: TiB2 particle, found in the melt-spun ribbon, nucleating several α-Al grains and metastable Al3Ti Ll2 particles with the low-index OR found in [2]: (a) TEM bright-field micrograph (A1 and A2 indicate α-Al grains corresponding to the two possible variants of the observed OR); (b) large-area SAD pattern; (c) schematic showing indices of the corresponding diffraction spots (zone axes [100]TiB2 and [011]Al or [0-1-1]Al, subscripts A1 and A2 are the same as in (a), subscript B indicates the TiB2 crystal lattice).
4
Conclusions
A detailed study of a commercial Ti-B-Al grain refiner rod, both in an as-received state and after rapid solidification using melt spinning, was undertaken in the present work. The microstructure of the as-received rod appeared to be dominated by recrystallisation rather than solidification processes. TiB2 particles were covered with thin Al3Ti DO22 layers that were already present in the as-received rod and appeared to remain unaffected by the high melt spinning temperature. These particles proved to be very potent sites for heterogeneous nucleation of α-Al during quenching, competing successfully with the metastable Al3Ti Ll2 particles having a negligible lattice mismatch with α-Al.
5
Acknowledgements
The authors gratefully acknowledge financial support from the EPSRC in conjunction with London & Scandinavian Metallurgical Co. Ltd. and Alcan International Ltd.
256
6 [1] [2] [3] [4]
References
D. G. McCartney, Int. Mater. Rev. 1989, 34, 247 – 260. P. Schumacher et al., Mater. Sci. Technol. 1998, 14, 394 – 404. L. Arnberg, L. Bäckerud, H. Klang, Met. Technol. 1982, 9, 7 – 13. 4. A. M. Bunn et al. in Solidification Processing 97 (Ed.: J. Beach and H. Jones), University of Sheffield, Sheffield, UK, 1997, 264 – 267. [5] S. Hashimoto, K. F. Kobayashi, S. Miura, Z. Metallkd. 1983, 74, 787 – 792. [6] W. T. Kim et al., Int. J. Rapid Solidif. 1992, 7, 245 - 254.
Effect of Instability of TiC Particles on the Grain-Refining Behavior of Al-Ti-C Inoculants in Aluminum Alloys M. Vandyoussefi and A. L. Greer University of Cambridge, Cambridge, UK
1
Abstract
According to the Al-Ti-C phase diagram, TiC is only a stable phase in Al-melts rich in Ti and C, and should convert to Al4C3 at the normal addition levels of Al-Ti-C refiners to melts. The effects of melt composition on TiC stability and on grain-refiner performance are examined using TP-1 refining tests on super-purity Al and using thermodynamic modeling. Effective grain refinement with little fading of performance on holding the refiner in the melt is observed only for melt compositions within the liquid + TiC phase field. In contrast when Al4C3 is stable, the refining is poor even at short holding times, and fades considerably at longer holding times. This change in grain-refining behavior is related to the decomposition of TiC.
2
Introduction
In spite of wide use of the conventional Al-Ti-B grain refiners [1-2], their disadvantages i.e. poisoning [3] and agglomeration of the boride particles [4], have led to interest in the production of Al-Ti-C based refiners [5-8]. Although the Al-Ti-C refiners have overcome these limitations, their effectiveness is lost on holding in the melt, especially at higher temperatures. This could be related to the intrinsic thermodynamic instability of the TiC phase in melts poor in titanium [9-11]. TiC particles could convert at their surfaces to, e.g., Ti2AlC or Al4C3, or dissolve completely and cause the precipitation of Al4C3 [9-10]. Consequently, nucleation potency would decrease through loss of the lattice matching of TiC and ccp-Al [10-11]. The formation and decomposition of Al4C3 have been experimentally confirmed [12-13], but under conditions different from normal grain refinement practice. Both nucleation and solutal restriction of crystal growth control grain refinement [14-17]. In aluminum melts, titanium plays a more important role in the restriction of crystal growth than other solutes. Usually, it enters into melts through addition of grain refiners which have more Ti than required to form TiB2 or TiC. The stability of the TiC phase depends on the dissolved titanium content in the melt. Thus the performance of Al-Ti-C refiners should be analyzed based on the titanium content. The aim of the present study is to show the effect of the stability of TiC particles on the grain-refining performance, through thermodynamic modeling and experiments.
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
258
3
Experimental Procedure
The TP-1 test [18] was applied to evaluate the performance of the Al-Ti-C refiners. Superpurity (SP) aluminum (99.999%) was chosen to highlight the role of dissolved titanium from the refiner itself, avoiding the interference by other solutes. To investigate the effects of phase stability on grain refinement, different phase fields in the Al-Ti-C phase diagram were sampled by selecting two refiners added to various levels. The Ti contents of refiner 1 and refiner 2 are respectively 4.91 and 1.82 wt.% and C contents are 0.14 and 0.17 wt.%. The melts were held at 700±5Û& EHIRUH DQG DIWHU DGGLWLRQ RI UHILQHU 7DEOH VXPPDUL]HV WKH experimental conditions and the chemical compositions of the alloys. Concentrated Tucker’s reagent (25% H2O, 45% HCl, 15% HNO3, 15% HF) was used to reveal the grain structure. Areas of 2 x 2 cm were cut from the centers of TP-1 sections and after polishing were anodized in Barker’s reagent (98% H2O, 2% HBF4). The grain size was measured in polarized-light microscopy on these samples. Table 1: Experimental conditions for the TP-1 tests on super-purity Al. Refiner 1 is used throughout, except for the S45 series (45 ppt addition of refiner 2). The shortest holding time in the melt is 5 min and the longest time is 180 or 360 min. The chemical compositions are measured by optical emission spectroscopy. The compositions of non-refined alloys are all very similar and are given only for the S1 series. The carbon contents are calculated from the known refiner addition levels. Q is the growth-restriction factor which is used to describe the effect of different solutes on grain size.† Sample series
Addition Holding time level (ppt) [min] Ti S1-0 0 — 0 S1-1 to S1-5 1 5, 30, 60, 180, 360 0.005 S5-1 to S5-4 5 5, 30, 60, 180 0.027 S17-1 to S17-4 17 5, 30, 60, 180 0.094 S30-1 to S30-4 30 5, 30, 60, 180 0.17 S45-1 to S45-4 45 5, 30, 60, 180, 360 0.08
Solute content (wt.%) C 0 0.0001 0.0007 0.0023 0.0041 0.0073
Si 0.0006 0.0004 0.003 0.085 0.0031 0.0049
Fe 0.0004 0.0009 0.0015 <0.01 0.0056 0.0077
Q V 0 0.0001 0.0008 — 0.0051 0.0026
[K] 0 0.67 3.35 11.39 20.1 8.1
† Q = m (k – 1) C0 , where m, k, and C0 are respectively the liquidus slope, the partition coefficient, and the solute content [14]. Measured solute contents are used in the calculations with m and k values from a standard reference [19].
4
Experimental Results
Figure 1 shows the macrostructures found on TP-1 test sections for the S1, S5, S17 and S45 series of samples (Table 1). With 1 ppt refiner addition (S1 series, Fig. 1a) the refinement is insignificant and appears to fade with holding time. With 5 ppt refiner addition (S5 series, Fig. 1b) the grain structure is refined at 5 min holding time, but at longer times, becomes nonuniform with elongated grains. Complete grain refinement was observed only at the higher refiner addition levels as shown in Fig. 1c-d for the S17 and S45 series.
259
Figure 1: Macrographs of TP-1 test sections of super-purity Al samples, showing the effects of holding time on grain-refining performance for: (a) S1, (b) S5, (c) S17 and (d) S45 series
Figure 2 shows the effect of Ti content and holding time on grain size. At low addition level (5 ppt), there is significant fading. In contrast at high addition levels, there is essentially no fading (30 ppt) or only slight fading (17 ppt). In the only series with refiner 2 (45 ppt), the grain-refining performance fades, but more slowly than for refiner 1at 5 ppt addition.
5
Discussion
Figure 3 shows the 700Û&LVRWKHUPDOVHFWLRQRIWKH$OULFKFRUQHURIWKH Al-Ti-C equilibrium phase diagram, as calculated using Thermocalc [20] and the database by Bälzner [21]. The liquid and liquid + Al3Ti phase regions are not shown as they are present only at very low carbon concentrations (below 10-6 wt.% C). As shown in Fig. 3b, TiC is stable in both refiners, together with Al3Ti formed due to the excess titanium. At 1 to 5 ppt addition levels, TiC is not stable because of low titanium content and instead Al4C3 should form. Also, Al3Ti is unstable and dissolves easily [22-23]. For 17 and 30 ppt addition levels of refiner 1, TiC is stable and therefore the equilibrium (liquid + TiC) can be applied. Also TiC can coexist with Al4C3 as shown in Fig. 3a for the case for 45 ppt addition level of refiner 2.
260
Figure 2: Variation of grain size with Ti concentration for various holding times
Figure 3: An isothermal section at 700Û& RI WKH $O7L& SKDVH GLDJUDP DW D ORZHU DQG E KLJKHU 7L DQG & concentrations. The compositions of melts inoculated with 1, 5, 7 or 30 ppt addition of refiner 1, and 45 ppt addition of refiner 2 are indicated in (a) and the compositions of two refiners in (b)
Figure 4: Variation of grain size with Q for 5 and 180 minute holding time. As the total growth-restriction factor is insufficient for grain refinement with 1 ppt addition, it is not shown
As shown in Fig. 3a, in samples with 1 ppt and 5 ppt addition levels, the stable carbide particles are Al4C3. However, some TiC particles may be preserved at short holding times, as the kinetics of transformation from TiC to Al4C3 may be slow. Therefore the performance of the refiner should be degraded, as Al4C3 is not a favored nucleant for ccp-Al. The slight fading with 17 ppt addition (Figs. 1-2) could be related e.g. to agglomeration or particle coarsening in the melt. Higher addition levels provide more inoculant particles but, there is a saturation level beyond which more refiner particles do not lead to finer grains [17]. Thus the
261 effect of variation of the number of particles on grain size is weaker at high addition levels, which may explain the slightly different fading behavior of samples S17 and S30. As the TiC converts, dissolved titanium enters into the melt. Although this increases the overall growth-restriction factor Q, its fractional increase is negligible. Figure 4 shows the effect of Q on the grain size. With 5 ppt addition, the growth-restriction factor is low which allows the occurrence of a strong fading effect. The effect of Q on grain size tends to saturate at higher values [17]. As shown in Figs. 2 and 3, the grain sizes are very similar in samples with 17 ppt and 30 ppt additions. This is consistent with the above prediction, as the difference in growth restriction between the two alloys is not substantial. The refiner addition levels in melts with unstable TiC were much lower than in those with stable TiC. Thus the interpretation of the grain-refining results could be biased by the differing populations of nucleant particles. As shown in Fig. 3a, the composition of the sample with 45 ppt addition of refiner 2 is in the liquid + Al4C3 + TiC phase field. Therefore when grain refiner is added to the melt, some of the TiC particles transform to Al4C3 on holding. Thus, refinement can be degraded even with a very high addition level and a high population of nucleants. This corresponds roughly with a break in refining behavior between the samples with unstable TiC (5 and 45 ppt) and stable TiC (17 and 30 ppt) as shown in Fig. 2.
6
Conclusions
By addition of different refiners at different levels, melt compositions can be altered such that, according to thermodynamic calculations, the TiC particles are stable or unstable in the melt under typical processing conditions. The stability or otherwise of TiC results in distinctly different refining behavior. When it is not stable, the refiner performance is weak even for short holding times and fades strongly with increasing holding time. The fading occurs progressively, but is more marked at lower solute levels where growth restriction is not adequate to promote refinement. It is concluded that the effectiveness of Al-Ti-C grain refiners at normal addition levels is affected by the instability of TiC particles, but that the TiC present in the refiners does not decompose instantly on addition to the melt. The refiner effectiveness is likely to be degraded by increased holding time, lower titanium content, or increased carbon content in the melt. Therefore, in industrial practice, the refiner is added to the melt at the last moment before casting (launder addition) in order to prevent the fading due to increased holding time.
7
Acknowledgements
This work was supported by the Engineering & Physical Sciences Research Council (UK). The authors acknowledge helpful discussions with project partners: Dr P. Schumacher and Dr K. A. Q. O’Reilly (University of Oxford); Dr R. G. Hamerton and Dr M. W. Meredith, J. Worth (Alcan); Dr P. S. Cooper and Dr A. Hardman (LSM); and A. Tronche (University of Cambridge).
262
8
References
[1] D. G. McCartney, Int. Mater. Rev., 1989, 34, 247 - 260. [2] P. Schumacher, A. L. Greer, J. Worth, P. V. Evans, M. A. Kearns, P. Fisher, A. H. Green, Mater. Sci. Technol., 1998, 14, 394 - 404. [3] A.M. Bunn, P. Schumacher, M. A. Kearns, C. B. Boothroyd, A. L. Greer, Mater. Sci. Technol., 1999, 15, 1115 - 1123. [4] G. P. Jones, J. Pearson, Metall. Trans. B, 1976, 7B, 223 - 234. [5] Banerji, W. Reif, Metall. Trans. A, 1986, 17A, 2127 - 2137. [6] M. A. Hadia, A. A. Ghaneya, A. Niazi, Light Metals 1996, (Ed.: W. Hale), 729-736, 1996, Warrendale, PA, TMS. [7] P. Hoefs, W. Reif, A. H. Green, P. C. Van Wiggen, W. Schneider, D. Brandner, Light Metals 1997, (Ed.: R. Huglen), 777 - 784, 1997, Warrendale, PA, TMS. [8] J. Whitehead, S. A. Danilak, D. A. Granger, Light Metals 1997, (Ed.: R. Huglen), 785 793, 1997, Warrendale, PA, TMS. [9] M. Vandyoussefi, J. Worth, A.L. Greer, Mater. Sci. Technol., in press. [10] P. S. Mohanty, J. E. Gruzleski, Scripta Metall. Mater., 1994, 31, 179 - 184. [11] D. Mayes, D. G. McCartney, G. J. Tatlock, Mater. Sci. Eng. A, 1994, A188, 283 - 290. [12] Jarfors, H. Fredriksson, L. Froyen, Mater. Sci. Eng. A, 1991, A135, 119 - 123. [13] N. Frage, N. Frumin, L. Levin, M. Polak, M. P. Dariel, Metall. Mater. Trans. A, 1998, 29A, 1341 - 1345. [14] Maxwell, A. Hellawell, Acta Metall., 1975, 23, 229 - 237. [15] M. Easton, D. St John, Metall. Mater. Trans. A, 1999, 30A, 1613 - 1623. [16] J. A. Spittle, S. B. Sadli, Mater. Sci. Technol., 1995, 11, 533 - 537. [17] L. Greer, A. M. Bunn, A. Tronche, P. V. Evans, D. J. Bristow, Acta Mater. 2000, 48, 2823 - 2835. [18] Standard test procedure for aluminum alloy grain refiners TP-1, The Aluminum Association, Washington DC, USA, 1987. [19] T. B. Massalski (ed.), Binary alloy phase diagrams, Vol. 1, 1990, ASM International. [20] Sundman, B. Jansson, J.-O. Andersson, Calphad, 1985, 9, 153 - 190. [21] Bälzner, Ph.D. thesis, University of Stuttgart, Germany, 1994. [22] T. W. Clyne, M. H. Robert, Met. Technol., 1980, 7, 177 - 185. [23] J. Asbjornsson, T. L. Sigfusson, D. G. McCartney, T. Gudmundsson, D. Bristow, Light Metals 1999, (Ed.: C. E. Eckert), 705 - 710, 1999, Warrendale, PA, TMS.
Grain Refiners for Thin Strip Twin Roll Casting Ray Cook London & Scandinavian Metallurgical Co Limited, Rotherham, UK
1
Abstract
In recent years there have been many attempts to improve the productivity and applicability of Twin Roll Casters for aluminium sheet products. Several developmental strip casters were manufactured around the world to investigate Thin Strip Twin Roll Casting, and from these results we see the installation of equipment capable of casting gauges down to 2mm. Productivity benefits have not yet been as expected, due in part to the difficulties in achieving desirable microstructures with the fast cooling rates and deformation characteristics inherent with this type of caster. The importance of selecting the correct grain refiner to optimise the microstructural properties of Twin Roll Cast aluminium alloys at all gauges and casting conditions is reviewed here.
2
Introduction
The Twin Roll Casting process for manufacturing aluminium alloy sheet was developed in the 1950’s by Hunter. From initially casting low solute containing alloys (e.g. foilstock) the range of castable alloys was steadily increased up to the 1980’s. The limitations to casting these alloys were either process related (heatlines or load/torque requirements) or structure/property related (centreline channel segregate formation [1]). For products where microstructural control was critical, casting at 10mm followed by hot rolling would be employed. To avoid hot rolling passes, and thus significantly reduce processing and equipment costs, the optimum gauge was 6mm where final processing would be by cold rolling. Mathematical modelling of the process in the 1980’s predicted significantly higher productivities were achievable if the casting gauge was reduced below 6mm [2, 3], and this was later confirmed by trials using laboratory casting equipment [4, 5]. The result of this and similar investigations [6] in other laboratories was the construction in the 1990’s of a number of commercial Twin Roll Casters capable of producing strip at gauges of 3 to 4mm typically and in some alloys down to 1mm [7].
3
Typical As-Cast Microstructures
3.1
Segregate Structures
In conventional Twin Roll Casting (6 to 10mm gauge) as the casting speed (and thus productivity) is increased there is a tendency for centreline channel segregates to form. If the Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
264 casting speed is increased further then at some point the melt is not fully solidified on exiting the caster, resulting in flow of liquid aluminium through the roll bite to form a heatline. For some applications (e.g. thick foil) centreline channel segregates are acceptable defects in the strip, however for critical applications a fine and uniform microstructure is required, as seen in Figure 1a. In Thin Strip Twin Roll Casting there are many other microstructural features which occur [8 - 10]. These have been represented on schematic segregation limit diagrams which are alloy specific [8]. At faster casting speeds at thicker gauges centreline channel segregates may form, but as the gauge is reduced the channel segregates begin to disperse away from the centreline, until at a critical setting of gauge and casting speed a disordered segregate structure containing surface segregates is formed. This critical setting varies depending upon the alloy composition, but is usually characterized by a visible change in the cast strip which begins to buckle. Mathematical modelling of the casting conditions when buckling occurs has shown it to be caused by a recirculation effect in the semi-solid region of the liquid sump [8].
Longitudinal
Transverse
(a) 1m/min, 0.01% Ti
Longitudinal
Transverse
(b) 1.3m/min, 0.01% Ti
Figure 1: Effect of casting speed on the structure of AA8111, 6mm gauge, grain refined with 5/1 TiBAl [13]
3.2
Grain Structures
As with segregate structures the grain structure of cast strips vary significantly with gauge and casting speed. At conventional gauges and speeds a uniform structure of fine columnar grains elongated in the solidification direction is found, Figure 1a. The grains are generally finer on the surface and slightly coarser in the centre due to differences in the solidification rate determined from dendrite arm spacing measurements [10]. As the casting speed is increased at conventional gauges the grain structure coarsens and becomes more equiaxed in the centre of the strip forming a banded microstructure as shown in Figure 1b. This change in grain structure becomes more apparent at thin gauges, Figure 2, where as the casting speed is increased further the equiaxed grain structure extends to almost the full strip thickness.
265
Longitudinal
Transverse
(a) 1.5m/min, 0.01% Ti
Longitudinal
Transverse
(b) 2m/min, 0.01% Ti
Longitudinal
Transverse
(c) 2.5 m/min, 0.01% Ti
Figure 2: Effect of casting speed on the structure of AA8111, 4mm gauge, grain refined with 5/1 TiBAl [13]
3.3
Grain Refinement
In twin roll casting, as with most other casting processes, it is necessary to control the solidifying microstructure in order to reduce casting defects, improve down stream processability and optimise final mechanical and physical properties. For the majority of alloys this entails the addition to the molten alloy of a grain refiner. The primary action of the grain refiner is to provide numerous nucleation sites for the solidifying metal, thus achieving a fine and uniform microstructure throughout the casting. In aluminium alloys the most regularly used and most potent grain refiner is based on the Al-Ti-B ternary system and comprises 5% titanium, 1% boron and the balance aluminium (5/1 TiBAl). The microstructure of 5/1 TiBAl consists of fine (typically 0.5 to 2µm) titanium diboride particles, TiB2, dispersed in an aluminium matrix containing coarser (typically 30 to 50µm) titanium aluminide, TiAl3. On addition of the grain refiner to molten aluminium the aluminium matrix melts dispersing TiAl3 and TiB2 into the melt. The TiAl3 rapidly dissolves leaving the TiB2 particles, which do not dissolve, as the nuclei required for grain refinement. The TiB2 particles are denser than molten aluminium however, 4.3g/cm3 compared to 2.3g/cm3 respectively, so that over time in a quiescent melt they will tend to settle out. There is also a tendency for the TiB2 particles to agglomerate by inter-particle contact, exacerbated by the presence of halide salts (Cl, Fl, etc.), thus increasing the settling rate. Because the settled TiB2 particles will no longer be active nuclei the degree of grain refinement decreases over time or fades, though this can generally be recovered by stirring the settled TiB2 back into the melt [11]. 3.3.1 Process Considerations In twin roll casting applications the grain refiner is most commonly added continuously to the molten metal stream between the melting/holding furnace and the caster in the form of a nominally 10mm diameter coiled rod. One important factor to consider is the selection of the correct addition point for this rod in the casting line. In a typical casting line consisting of a melting/holding furnace, degassing unit, filter and then casting tip it is usual practice to add grain refiner either after the furnace or before the filter. Adding the refiner in the melt stream exiting the furnace has the advantage of increased turbulence to maximise mixing into the melt stream, however the disadvantage is that depending on the size of the degassing unit, filter and launders it can take up to 1 hour from addition to final casting. During this time the probability of TiB2 agglomerate formation increases, and with the slow flow rate of the melt along the launders the TiB2 has the tendency to settle out. Addition before the filter can
266 overcome this to a large extent though, depending on the filter type used, long residence times may occur. Addition after the filter can minimise residence times and thus agglomeration/settling can be reduced, however at this stage the metal stream is quiescent and thus it can be difficult to achieve adequate mixing into the melt. The selection of the addition point is usually a compromise between the degree of grain refinement required and the efficiency of mixing in the launders. 3.3.2 Selection of TiBAl Grain Refiners Although 5/1 TiBAl is the most potent and most widely used grain refiner for aluminium alloys, there are cases where alternative compositions are favoured even though potency is generally reduced [12]. As mentioned above it is possible for TiB2 particles to agglomerate and settle in the launders. If this occurs after the filter then there is the possibility of large boride particulates being carried through the casting tip and into the cast strip. For many applications this would not cause any significant degradation in strip properties. However in thin foilstock products, which may be rolled down to 5µm gauge, boride agglomerates can lead to high pinhole counts. To reduce the possibility of agglomeration it is necessary to reduce the total number of boride particles in the melt. Although this can be achieved by reducing the total grain refiner addition level it may be achieved more effectively by using a grain refiner with more dilute borides, for example 5/0.2 TiBAl where the boron level is only 0.2%. Because of the lower level of TiB2 particles this refiner is not as potent as 5/1 TiBAl and higher quantities may be required to achieve equivalent grain refinement. However the total boron content added will be lower, as shown for example in Figure 3. 0.01
3/1 TiBAl Lines of constant grain size
B A d d ition %
5/1 TiBAl
0.005
3/0.2 TiBAl 5/0.2 TiBAl 0 0
0.02
0.04
Ti A dd itio n %
Figure 3: Effect of alternative grain refiner compositions on final melt chemical analysis (note: actual values depend greatly upon the alloy chemistry and casting/processing conditions used, thus the above is an indication only of the possible effect of alternative grain refiners)
267 Another problem which can occur is the build up of titanium in the melt due to recycling of scrap material. If a low total titanium level is required to meet the alloy specification and the remelted alloy already contains significant titanium, the amount of grain refiner which can be added may be limited. If this addition does not give the desired degree of grain refinement it is possible to add a grain refiner containing only 3% Ti such as 3/1 TiBAl. As with the low boron refiners this is not as potent as 5/1 TiBAl, however to achieve equivalent grain refinement the total titanium addition will be lower as shown for example in Figure 3. It is also possible to minimise boride agglomeration and minimise titanium additions by use of 3/0.2 TiBAl.
(a) 0.01% Ti
(b) 0.02% Ti
5/1 TiBAl
TiCAl315
(c) 0.01% Ti
(d) 0.02% Ti
Figure 4: Effect of grain refiner type and addition level on the structure of AA8111, 4mm gauge, 2.5m/min, transverse sections [13]
3.3.3 Alternative Grain Refiners In addition to the traditional TiBAl grain refiners there has been considerable development in recent years of products based on titanium carbide. These TiCAl alloys were originally developed to overcome zirconium poisoning of TiBAl alloys, and have been shown to be more effective in Thin Strip Twin Roll Casting [13]. Figure 4 shows how use of TiCAl315™ (3% Ti, 0.15% C, balance Al) at lower Ti addition levels in AA8111 cast at 4mm gauge leads to the formation of a finer grain structure than that achieved with 5/1 TiBAl. In addition carbides are less prone to agglomeration than borides, and thus they could offer benefits in reducing pin hole defects in thin foil. Industrial trials are ongoing using TiCAl at several Twin Roll Casting facilities.
4
Conclusions
It has been shown that to overcome some of the problems associated with grain refining Conventional and Thin Strip Twin Roll Casting there are a range of alternative TiBAl alloy chemistries which although having differing potencies may offer process advantages. In Thin Strip Twin Roll Casting it has been found that TiCAl315™ can be more effective than 5/1 TiBAl at lower addition levels, and their use is currently being evaluated in production at both conventional and thin gauges in various alloys.
268
5
References
[1] Jin, L. R. Morris, J. D. Hunt, JOM June 1982, 70 - 75, [2] M. J. Bagshaw, J. D. Hunt, R. M. Jordan, Proc. Third Confr. Modelling of Casting and Welding Processes, AIME, 1986. [3] D. J. Browne, The Measurement of Heat Transfer Coefficients in Roll Casting, M.Sc. Thesis, University of Oxford, 1989. [4] D. V. Edmonds, et.al., Proc. Int. Symp. on Extraction, Refining and Fabrication of Light Metals, CANNMET, 1991, 18 – 22. [5] M. Yun, et.al., Cast Metals, 4 (2), 1991, 108 – 111. [6] Anon., Aluminium Today, 1995 August/September, 21 – 22. [7] I. Nussbaum, Light Metal Age, 1996 December, 8 – 19; 1997 February, 34 – 39. [8] P. M. Thomas, P. G. Grocock, Alumitech ’97, Altlanta, 1997. [9] S. Ertan, et.al., Light Metals 2000 (Ed: R. D. Peterson), TMS, 2000, 667 – 672. [10] O. Daaland, et.al, Light Metals 1997 (Ed: R. Huglen), TMS, 1997, 745 – 752. [11] W. Schneider, et.al, Light Metals 1998 (Ed: B. Welch), TMS, 1998, 953 – 961. [12] P. S. Cooper, P. Fisher, Light Metals 1994 (Ed: U. Mannweiler), TMS, 1994, 997 - 984. [13] M. Yun, et.al., Light Metals 2000 (Ed: R. D. Peterson), TMS, 2000, 857 - 862.
Characterisation and Optimisation of Thixoforming Feedstock Material S. Engler1, D. Hartmann2, I. Niedick3 1
Gießereiinstitut, RWTH Aachen, Germany EFU Gesellschaft für Ur-/Umformtechnik mbH, Simmerath, Germany 3 Volkswagen AG, Braunschweig, Germany (former EFU GmbH) 2
1
Abstract
The standard alloy for serial production by the Thixoforming process is alloy A356. It shows good corrosion resistance as well as acceptable mechanical properties after heat treatment. Due to the increased demand on mechanical properties of thixoformed components this alloy could not fulfil all requirements. The objective of this work is to develop thixoformable alloys with higher strength (compared to A356) and self-hardening alloys. KEYWORDS: alloy development, alloys with higher strength, self-hardening alloys, DCcasting
2
Introduction
Thixoformed components produced in AA 6082 are showing good mechanical properties [1, 2, 3]. On the other hand this alloy is relatively difficult to process by thixoforming due to its narrow process window and the pronounced susceptibility to hot cracking. Therefore the modified alloy AlMgxSix with increased magnesium and silicon content was investigated. With regard to crash behaviour the required mechanical properties for thin walled spaceframe-knots are YS = 120 – 150 MPa, UTS > 180 MPa, elongation > 15 %. Standard production is high pressure die casting with heat treatment of the components which results in distortion and therefore expensive dressing. The self-hardening alloy AlMg5Si2Mn (Magsimal 59) was chosen as a promising candidate and two modifications (AlMg3,5Si1,4Mn; AlMg2Si0,8Mn) were included to study the influence of silicon and magnesium.
3
Feedstock Production
Feedstock material for the investigations was produced by EFU’s vertical continuous DCcaster with electromagnetic stirrer (Fig. 1). The main components of the DC-caster are a tilting induction furnace (capacity: approx. 250 kg aluminium) and a vertical hydraulically driven casting machine. The furnace is electrically powered by medium frequency current Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
270 (500 Hz). The stirring method is circumferential. All experiments were performed with a degassing step that was carried out in the furnace bath before casting. The used impeller works with Argon as refining gas. The treatment was performed in two subsequent periods (2 x 15 min, each 12l Ar/min). After degassing the furnace is also used as pouring furnace and the melt is continuously poured into the launder from where it flows into the water cooled DC-casting mould. The mould design used is a Hot-Top system which allows the additional use of electromagnetic mould stirring. From one batch (250 kg) of molten metal up to four billets (length 3,5 m) could be cast subsequently.
Figure 1: Vertical continuous DC-caster with electromagnetic stirrer
Generally the requirements for thixoforming feedstock material and the influence of casting speed and stirring intensity may be summarized as follows: 3.1
Microstructure
The reheating process will cause the rosette-like structures still present in the as-cast billet to become fully globular, thereby creating the material structure typical for the process. The best flow behaviour for forming is achieved when the microstructure consists of very fine and very round globular grains. In addition the amount of liquid phases entrapped within the grains should be very low, because it could not participate to the deformation of the material and therefore is to be considered as solid. The grain size and the entrapped liquid decreases with increasing stirring intensity. The casting speed has no influence on the entrapped liquid and a slight one on the grain size; decreasing grain size with increasing casting speed. The form factor (quotient of circumference of a circle with the same area than the grain and circumference of the grain) increases with increasing stirring intensity and is not influenced by the casting speed.
271 3.2
Macrosegregation
Macrosegregation in the feedstock material could lead to parts with uneven mechanical properties. Near the surface the highest concentration of the alloying element silicon could be measured while the casting speed is high or the stirring intensity is high. The marked degree of the so called reversed block-segregation in the centre of the material increases with increasing casting speed and increasing stirring intensity. 3.3
Melt Loss
It is generally noted that various feedstock types lose different amounts of liquid phase during the inductive heating process [4]. Due to economically reasons it is important to ensure that the melt loss is <10 %. Additionally it has to be avoided that the melt loss is transferred inside the part. Due to its eutectic composition this would result in a part with inhomogeneous chemical composition and therefore with uneven mechanical properties. The melt loss is strongly dependent on the casting speed (decreased casting speed ⇒ decreased melt loss) and slightly on the stirring intensity (decreased stirring intensity ⇒ decreased melt loss). 3.4
Viscosity
The viscosity of the feedstock material during the forming processis is of key importance for its flow characteristic, which in turn has a major impact on the die filling behaviour [4]. It was evaluated with a straightforward “capillary” viscosimeter. Low values for the viscosity could be examined with high casting speed or with high stirring intensity. Within a number of DC-casting experiments the process parameters (casting speed, stirring intensity…) were varied to investigate the influence of them on forgoing criteria. Feedstock material was produced with the best casting parameters for further investigation.
4
Alloys with Higher Strength
The alloy AA6082 was found to be thixoformable, but it showed a pronounced susceptibility to hot cracking, which reduces the application to relatively simple geometries. Fig. 2 shows the typical microstructure of the original AA6082 wrought alloy under T6 heat treatment condition, taken from a thixocast connecting rod in the region of the outer core. The picture also shows a hot crack. According to the low eutectic phases content, there are only a few amounts of these phases surrounding the α-phase. Literary enquiries showed the influence of the main alloying elements (silicon and magnesium) on this hot cracking tendency [5]. The more silicon is added the more hot cracking susceptibility can be reduced. The pure metal (100 % aluminium) shows no susceptibility for hot cracks because there is only one phase building up the microstructure during solidification. A slight addition of other elements like silicon or magnesium increases the amount of low melting phases, like the ternary eutectic in case of AA6082, only to very low levels. Therefore the hot cracking susceptibility rises because of the alloys inability to compensate the materials shrinking during the final steps of solidification. If there are enough eutectic phases available, the hot cracking susceptibility decreases as observed for alloys with high silicon or magnesium content.
272
Figure 2: Hot crack in a principal part (connecting rod); alloy AA6082
Therefore the development work was focused on the alloy modification AlMgxSix with increased magnesium and silicon content. Thermochemical calculations at equilibrium for the distribution of fraction liquid over temperature proving a wider process window for alloy AlMgxSix compared to AA6082. The connecting rod (Fig. 2) has been used to study the thixoforming behaviour of the alloy AlMgxSix. It is a typical simple EFU test and demonstrator part for thixoforming process. Three different cross sections were realised with different die-inlets (rectangular cross section, C-profile, I-profile). Due to this fact different wall thicknesses were possible. Furthermore the connecting rod has typical difficulties for the thixoforming process. At the ingate there is a change of 90° in the direction of the „flow-path“ and at the end of the „flow-path“ the semi-solid metal has to reweld after flowing around the cores. YS; UTS (MPa) 400
YS UTS A5
350
elongation (%) 20
15
300 250 200
10
150 100
5
50 0
0 AlMgySiy A356 [6]
A356 [7]
A356 [8]
A356 [9]
Figure 3: Mechanical properties of alloy AlMgxSix and A356, after T6 heat-treatment
273 The alloy AlMgxSix has a sufficient thixoformability concerning its flow behaviour and its ability of rewelding flow fronts. No hot crackings could be observed during the tests. Tensile test specimens were dissected from the connecting rod. The specimens were T6 heat-treated. The parameters for heat treatment were the following: 530°Cx2h / water quenched / 160°Cx6h. The yield stress of the alloy AlMgxSix is approx. 310 MPa the ultimate tensile strength is approx. 370 MPa. Both values are therefore approx. 50 MPa above of the values given in the literature (Fig. 3). The elongation of the alloy is 7,5 % and consequently sufficient.
5
Self-hardening Alloys
In addition to the mechanical property targets for the space-frame-knots (YS = 120 – 150 MPa; UTS > 180 MPa; elongation > 15 %) which should be achieved without heat-treatment the feedstock material must have excellent flow and die filling behaviour, because this type of components involves long flow distances with very thin walls. Therefore the self-hardening alloy AlMg5Si2Mn was examined (Table 1). Table 1: chemical composition of alloy AlMg5Si2Mn (Magsimal 59) Si Fe Cu Mn Mg Zn Ti others Min. 1,8 0,5 5,0 0,1 Max. 2,5 0,13 0,05 0,8 5,5 0,08 0,2 0,06 The reheated microstructure (Fig. 4) consists of fine globular primary phase surrounded by eutectic phases. The α-Phase is less globular then known from alloy A356.
100 µm
100 µm
as cast reheated Figure 4: Microstructure of alloy AlMg5Si2Mn before and after reheating
The thixoformability was tested with various test parts, representing thin and thick walled cross sections. The alloy AlMg5Si2Mn showed excellent flow and die filling behaviour and is even better than of alloy A356 [10]. The mechanical properties measured in a thin walled plate (thickness = 3 mm) and in a door handle (wall thickness 5 – 20 mm) are a little bit to high concerning the yield stress, sufficient concerning the ultimate tensile strength and slightly too low concerning the elongation (Fig. 5).
274
YS; UTS (MPa) 500 required properties 450 properties of the plate 400 properties of the door handle 350 297 311 300 250 120 - 177 160 > 180 200 150 150 100 50 0 YS UTS
elongation (%) 20 15 18 13 > 15 16 14 12 10 8 6 4 2 0 elongation
Figure 5: Mechanical properties of alloy AlMg5Si2Mn
The content of eutectic phases after reheating of the alloy modifications AlMg3,5Si1,4Mn and AlMg2Si0,8Mn is less than for alloy AlMg5Si2Mn, due to the decreased content of silicon and magnesium. The thixoformability and the mechanical properties were tested with the connecting rod (Fig. 2). The flow and die filling behaviour is unsatisfactory, while the material showed non-rewelded flow fronts at the cores. The yield stress of the alloy modifications was reduced as required, but the elongation in the as cast condition could not be improved.
6
Conclusion
Alloy AlMgxSix has a sufficient thixoformability concerning its flow behaviour and its ability of rewelding flow fronts. The alloy has no hot cracking tendency. The mechanical properties measured in tensile test specimens dissected from the demonstrator part after T6-heat treatment were significantly improved compared to alloy A356). The best agreement with the required properties for thin walled space-frame-knots is alloy AlMg5Si2Mn. Also this alloy showed excellent die filling capacity (judged even better than alloy A356 [10]).
275
7
References
[1] G. Chiarmetta, 4th Int. Conference on Semi-Solid Processing of Alloys and Components, Sheffield (England), (1996), p. 204 – 207 [2] H. E. Pitts, H. V. Atkinson, 5th Int. Conference on Semi-Solid Processing of Alloys and Components, Golden (Colorado), (1998), p. 97 - 104 [3] G. Hirt, R. Cremer, T. Witulski, VDI-Berichte Nr. 1324, (1997), p. 55 – 73 [4] F. Niedermaier, J. Langgartner, G. Hirt, I. Niedick, 5th Int. Conference on Semi-Solid Processing of Alloys and Components, Golden (Colorado), (1998), p. 407 – 414 [5] S. Nikitin, R. Ellerbrok, S. Engler, Gießerei 76, Nr. 9, (1989), p. 297 – 299 [6] M. P. Kenney, J. A. Courtis, R. D. Evans, G. M. Farrior, C. P. Kyonka, A. A. Koch, Metals Handbook, 9th edition, ASM International, Volume 15: Casting, (1988), p. 327 – 338 [7] B. Wendinger, 4th Int. Conference on Semi-Solid Processing of Alloys and Components, Sheffield (England), (1996), p. 239 – 241 [8] N.N., Thixocasting, Aluminium Pechiney, (1994) [9] J.-P. Gabathuler, R. Jaccard, R. Röllin, Ch. Ditzler, , VDI-Berichte Nr. 1235, (1995), p. 81 – 105 [10] Niedick (ed.) ”Thixotec”-Abschlußbericht, ISBN 3-89653-651-6, p. 12-27
Experimental Study of Linear Shrinkage during Solidification of Binary and Commercial Aluminum Alloys Dmitri Eskine and Laurens Katgerman Netherlands Institute for Metals Research, Delft, The Netherlands
1
Introduction
The process of direct-chill (DC) casting is a very common way to produce ingots and billets for further deformation. Although this technology is used in aluminum industry since 1950s, there are still many problems related to the process. Among these problems, there are the formation of structure and properties of the mushy zone and, the most important for industry, the causes of defects. Hot tearing, porosity and macrosegregation are the major defects occurring during casting. Obviously, the flow of liquid, the diffusion of solute elements, the structure formation, and the development of strength in the mushy zone, being in complex interaction, may result in the formation of defects. Hot tearing or hot cracking is one of the most common problems encountered in DC casting of aluminum alloys. The main cause of this defect is that the stresses and strains built up during solidification are too high as compared to the actual strength of the semi-solid material. This type of defects occurs in the lower part of the solidification range, close to the solidus, when the solid fraction is more than 0.9 [1]. At this moment, the mushy zone is definitely coherent, but the liquid film still exists between most of grains. The term coherency (or coherency temperature) should be used with caution. Usually, the coherency marks the formation of a continuous dendritic network, when the material starts to develop the strength [2, 3]. At temperatures above the coherency point, the grains are free to move with respect to each other and do not transfer any forces. Moreover, before the coherency point is reached the liquid phase can easily flow between grains and, therefore, the melt feeding and the redistribution of solute elements occur without much difficulties. Another meaning of the coherency point designates it as a point at which the measured temperature dependence of the viscosity of the liquid-solid mixture changes the slope. Depending on the alloying system and, what is very important, on the technique of measurement, the coherency point may occur in the broad range of temperatures (or solid fractions). Different authors assign this point to the solid fraction in the range from 0.1 to 0.5%. The question is when hot tearing really occurs, and what are the driving forces for hot tearing. To answer the first question, the terms of the effective solidification range [2, 4] or the vulnerable part of the solidification interval [5] were introduced in 1940s–1950s. The upper boundary of this range is the point where the solidification shrinkage starts [2]; the lower boundary is the solidus (equilibrium or nonequilibrium, depending on the solidification conditions). As for the driving forces, it is well adopted that the collapse of a dendritic network is caused by the developing stresses, which result from the shrinkage and driving pressure for liquid flow. The necessary condition for hot tearing is the existence of thin, continuous, interdendritic liquid film alongside the low permeability of the mushy zone. This Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
277 condition is usually fulfilled at large volume fractions of solid, 0.9 to 0.99 [6]. Therefore, there is a clear gap between the coherency point and the temperature below which hot tearing occurs. Novikov [2] suggested to determine the upper boundary of the effective solidification range by measuring the linear shrinkage. Hence, the upper temperature of this solidification range is the temperature, at which the linear shrinkage starts. To avoid the confusion, we need to specify the term “linear shrinkage”. There are different terms in which the word “shrinkage” is used. The casting shrinkage is a technological parameter describing the difference in dimensions or volume between the original cavity of the mold and the final casting cooled to the room temperature. This shrinkage, by definition, includes the volume differences between liquid and solid phases, the shrinkage occurring in the solidification range, and the contraction of the completely solid casting. The solidification shrinkage is the shrinkage (usually volume shrinkage) occurring in the solidification range, from 100% liquid to 100% solid. The linear solidification shrinkage is the horizontal change in linear dimensions of a casting during solidification [2]. Above the temperature of the linear shrinkage onset, the alloy is fluid as there is not continuous (between the walls of the mold) network of dendrites. In this stage of solidification, the thermal shrinkage cannot manifest itself as the horizontal contraction of the casting. All volumetric changes appear as the decreasing level of the melt in the permanent mold or does not appear at all during DC casting, due to the continuous supply of the melt to the mold. However, the linear shrinkage appears, and can be measured, when the fluidity of the alloy drastically drops, and the rigid skeleton of the solid phase forms. Starting from that moment, the alloy acquires the capability to retain its shape, and the thermal shrinkage of the solid phase displays itself as the linear contraction in the horizontal direction. One can also call this parameter “a one-dimensional solidification shrinkage”, keeping in mind that it is measured in the horizontal dimension of the casting. A special technique was developed to measure the linear shrinkage (and pre-shrinkage expansion) upon solidification [2, 7, 8]. Several designs of an experimental setup were suggested, all sharing the following features: graphite mold (providing low friction and high thermal conductivity) with one moving wall; water-cooled base (for high cooling rates comparable with those upon chill casting); and simultaneous temperature and displacement measurements [2]. Despite numerous published studies by Novikov et al., e.g. [2, 7–9], there are only few data available on the contraction behavior of solidifying wrought alloys, which are mainly produced by DC casting. The aim of this paper is to describe the experimental technique used for measuring the linear shrinkage upon solidification of binary and commercial wrought alloys; to discuss the factors influencing the measured property; and to present the experimental results obtained using the developed technique.
2
Experimental
The experimental setup used in measurements of the linear shrinkage upon solidification comprised the following parts: a graphite mold (Fig. 1) with one moving wall, a water-cooled bronze base; a K-type thermocouple (0.25 mm wires); and a linear displacement sensor (a Schaevitz DC-DC LVDT). To attach the solidifying metal to the moving wall on the one
278 side and to the permanent wall on the other side, we used a metallic rod with a thread embedded into the moving head and a T-shaped cavity on the other side of the mold. The cross-section of the T-shaped cavity was thinner than that of the main cavity, which allowed the melt to solidify faster. The metallic rod fixed in the moving head was frozen in the sample during experiments. The cross-section of the main cavity was 25 × 25 mm with a gauge length of 100 mm. The data was acquired using a Keithley KPCI-1801HC interface card and a LabVIEW software.
Figure 1: Experimental mold for measuring linear contraction behavior in the solidification range
The dimensions of the mold were chosen in agreement with Novikov et al. [8], who showed that these dimensions made the measured property not scale-dependent. The linear shrinkage was determined as follows: ε = {(ls + ∆lexp – lf)/ls} × 100%, where ls is the initial length of the cavity (100 mm); lf is the length of the sample at a temperature of solidus, and ∆lexp is the pre-shrinkage expansion. The linear displacement was measured accurate to 6 µm or 0.006% with 3 to 5 samples measured for each point. The distance between the thermocouple tip and the bottom of the mold was about 1.5 mm. Otherwise, there were problems with filling the gap. Evidently, the measured property should depend on the structure parameters; chemical composition; and the force applied to the moving wall of the mold. In order to examine the effects of these factors we (i) varied the melt temperature from 700 to 800 oC (grain size) and the cooling rate from 7 to 15 K/s (dendritic parameter); (ii) used alloys of different alloying systems (Table 1); and (iii) changed the friction force applied to the moving part from 0.11 to 0.83 N. In most cases, the liquidus and solidus temperatures could be derived from the cooling curve. Note, however, that we determined the linear shrinkage in the entire solidification range which, at the used cooling rates, extends down to the lowest possible eutectic temperature. The liquidus and solidus temperatures (both equilibrium and nonequilibrium) are given in Table 1. After acquiring the primary data, temperature and displacement against time (Fig. 2a), we processed the cooling curve in order to obtain information about critical temperatures. After that, we reconstructed the data to receive the direct dependence of displacement on
279
o
700 600 500 400 300 200 100 0 -100 -200 -300 -400
0.00
1
Displacement, %
Temperature, C Dispacement, µm
temperature (Fig. 2b). From this dependence the linear pre-shrinkage expansion, the linear solidification shrinkage and the temperature of its onset can be extracted.
2
-0.40
-0.80 0
10
20
Time, s
30
400
40
a)
500
600 o
Temperature, C
700
b)
Figure 2: Examples of data obtained during experiments: a, primary data, 1 for temperature and 2 for displacement (an Al-4% Cu alloy; melt temperature 750 oC; cooling rate 12 K/s; friction force 0.164 N) and b, temperature dependence of displacement (an Al–4% Cu alloy; melt temperature 750 oC; cooling rate 11 K/s; friction force 0.12 N)
Table 1: Nominal chemical composition and critical temperatures (tl – liquidus; tes and tnes – equilibrium and nonequilibrium solidus, respectively) for examined alloys. Alloy Chemical composition, wt% tl, oC tes, tnes, o o C C Cu Mg Zn Mn Fe Si Al Al–Cu 1050 2024(a) 5182(a) 6082(a) 7075(a) 1.
3
4.0 4.4 0.1 0.1 1.6
1.5 4.5 0.8 2.5
5.6
0.6 0.35 0.7 0.2
0.15 0.3 0.3 0.2 0.3 0.3
0.1 0.2 0.3 0.2 1.0 0.3
bal. bal. bal. bal. bal. bal.
652 657 638 638 649 635
582 646 502 577 566 532
548 502 ~530 557 477
Contained also Ti and Cr.
Results and Discussion
An Al–4% Cu alloy was chosen for examination because of the well-known phase diagram, structural features and some reference data on its contraction behavior. The results are summarized in Fig. 3, where the combined effects of structure and applied force are given as a 3-D plot along with contours of the linear solidification shrinkage. The structure factor used in these plots is a combined function of the grain size (D) and the dendritic parameter (d): d/D. The examination of structure showed that the actual grain size varied from 120 to 220 µm, the structure being mostly equiaxed across the sample. For this type of macrostructure, the finer is the grain and the coarser is the dendrite, the larger is linear solidification shrinkage.
280 In further experiments with commercial alloys, the friction force was 145 mN. The linear solidification friction measured at considerably low friction forces is about 0.2%. This value agrees well with that previously reported (0.25%) [2]. Korol’kov [9] reported that the measured linear solidification shrinkage was always higher than the calculated values. This is in favor of our results. The important parameter, that can be derived from our experiments, is the temperature of the linear shrinkage onset (tso). This temperature is about 560 oC, being independent of the friction force and structure. This temperature, being between the temperatures of the equilibrium solidus and nonequilibrium solidus, shows that the continuous dendritic network is formed at very high solid fractions. Moreover, the amount of nonequilibrium eutectic liquid should be taken into account when considering the solidification at high cooling rates. The results on the linear solidification shrinkage for several commercial alloys are given in Table 2. Structure factor (x 1000)
5.0 4.5
0.17
4.0
0.15 0.13
3.5
0.11
3.0
0.09
2.5
0.07
2.0
0.05
1.5
0.03
1.0 0.0
0.01
0.2
0.4
0.6
0.8
1.0
Friction force, N Figure 3: The effect of structure and friction force on the linear solidification shrinkage of an Al–4% Cu alloy
Obviously, commercial alloys behave differently upon solidification, with respect to an alloying system. Commercial aluminum (1050) shows almost no shrinkage in the solidification range, which is no surprise. However, Al–Mg and Al–Mg–Si alloys (5182 and 6082, respectively) also demonstrate negligible linear solidification shrinkage and start contraction only at the solidus temperature or even below. The largest linear solidification shrinkage is observed in a 7075 (Al–Zn–Mg–Cu) alloy, the temperature of shrinkage onset being between the equilibrium and nonequilibrium solidus. Table 2: Temperature of shrinkage onset (tso), pre-shrinkage expansion (+∆l) linear solidification shrinkage (εs) and the total casting shrinkage (εc) of commercial alloys (melt temperature 720–730 oC; cooling rates 12–18 K/s). Alloy tso, oC +∆l, % εs, % εc, % 1050 657 0 0.03 2.3 2024 515 0.035 0.2 1.5 5182 528 0.08 0 1.5 6082 563 0.025 0.02 1.5 7075 525 0.03 0.27 1.5
281
4
Conclusions
1. A technique to adequately measure the linear solidification shrinkage is developed and used for binary and commercial aluminum alloys. 2. The linear solidification shrinkage depends on the chemical composition, structure, and applied force. 3. The finer the grain structure and the coarser is the internal dendrite structure, the larger the linear solidification shrinkage. 4. The temperature of the shrinkage onset depends on the chemical composition.
5
References
[1] J. Campbell, Castings, Butterworth-Heinemann, Oxford, 1991. [2] I.I. Novikov, Goryachelomkost tsvetnykh metallov i splavov (Hot Tearing of Nonferrous Metals and Alloys), Nauka, Moscow, 1966 (in Russian). [3] A.K. Dahle, L. Arnberg, Acta Mater. 1997, 45, 547–559. [4] A.A. Bochvar, Izvestiya Akad Nauk SSSR, OTN 1942 (9) 31 (in Russian). [5] W.S. Pellini, Foundry 1952, 125–133; 192–199. [6] T.W. Clyne, G.J. Davies, Br. Foundrymen 1981, 74, 65–73. [7] A.N. Yakubovich, I.I. Novikov, G.A. Korol’kov, Russian Castings Production 1969, (10), 472–474. [8] I.I. Novikov, G.A. Korol’kov, A.N. Yakubovich, Russian Castings Production, 1971, (8), 333–334. [9] G.A. Korol’kov, Liteinoe Proizvod. 1986, (1), 6–7 (on Russian).
The Influence of the Cooling Rate on the Type of the Intermetallic Phases in the Aluminium Alloys of the 3XXX (AlMnMgSi) Group Tomasz Stuczyn´ski and Marzena Lech-Grega Institute of Non-Ferrous Metals, /LJKW0HWDOV'LYLVLRQXO3LáVXGVNLHJR 6NDZLQD3RODQG
1
Abstract
The results of investigations of the role of cooling rate, within the range of 0.6 to 28°C/s, and of chemical composition on the type of intermetallic phases in the aluminium alloys of the 3xxx (AlMnMgSi) type are presented. The procedure introduced and published by the Alcoa Research Centre was used in the studies. It was shown that the fraction of phases of the Al6(FeMn) and a-AlFeMnSi type varies in the investigated alloy depending on the cooling rate and the Fe/Si ratio. It was found that the AlFeMnSi phase, required from the point of view of the application of the studied alloy (beverage cans) predominates at the Si content twice the Fe content. On the other hand the a-AlFeMnSi phase is the only one ferricmanganese phase at the cooling rate of 2.6°C/s. The results obtained on the laboratory scale have been confirmed on the pilot plant scale, during casting the ingot from the studied alloy, using a semi-continuous method.
2
Introduction
A characteristic feature of the process of ingot casting of aluminium and its alloys by a semicontinuous method is that each elementary volume of the cast metal is solidified with different rate, the magnitude of which depends on the employed parameters of the casting process, i.e., the temperature of the melt, casting speed, the amount of cooling water and on the position in relation to the mould walls and the axis of the ingot being casted. It means that this phenomenon is critical in the case of casting of large ingots of rectangular cross-section used as the charge for making rolled products. The studies on the thermal processes occurring upon the casting of the melt into an ingot have shown that the cooling rate expressed indirectly through the intensity of the temperature drop of the given elementary volume of the cast metal by the semi-continuous method into classical crystallizers of the DC type ranges between 0.5 to 20°C/sec. A typical distribution of the cooling rates in an ingot is presented in Fig. 1. In the plot depicting the distribution of cooling rates in the ingot, the four zones can be distinguished: Zone I: the area of the highest cooling rate placed at the surface of the ingot. Responsible for the kinetics of the cooling process in this area is the intensity of carrying away of heat through the walls of the mould forming the cast ingot. Zone II: the area of reduced cooling rate placed between the end of the area of the primary cooling impact - through the mould walls - and the beginning of the area of the secondary cooling caused by the direct cooling water spray flowing out of the mould. Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
283
Figure 1: Typical distribution of cooling rates in an ingot
Zone III: the area of increased cooling rate placed in the area of the impact of the secondary cooling. Zone IV: includes the middle part of the ingot. The size of this area depends on the intensity of carrying away heat along the axis of the cast ingot and for the most part it is the function of the casting speed. The occurrence of these four zones in the distribution of cooling rates is reflected in the pattern of the solidification front in the casted ingot (Fig. 2).
Figure 2: Pattern of solidification front
There exists a known relationship between the cooling rate Vk and the casting speed VL expressed by the equation: r r vK = vL cos α where: α is the angle between the normal to the solidification front line and a line parallel to the direction of casting. In the foundry practice it was found that in many cases, even the changes in the casting speed of 50% do not affect the distribution of intermetallic compounds, the example of which
284 can be the result obtained on the 5182 alloy. Nevertheless there exists a group of alloys in which the occurrence of differentiated conditions in the solidification process in the ingot, as manifested by a different cooling rate of different elements of volume dx⋅dy⋅dz, gives rise to significant differentiation in the structure of the alloy at the level of its microstructure. This group of alloys includes mainly the alloys of the series 1xxx and 3xxx according to ASTM designation. In ingots made of these alloys, a significant differentiation occurs in the form and type of intermetallic compounds based on Al, Fe, Mn and Si depending on the position of the studied area in relation to the side walls of the ingot, i.e., on local parameters of the solidification process. Since the studies on the temperature field in ingots cast by the semi-continuous method are difficult because of the apparatus problems and a lack of a mathematical 3D model of the process of ingot solidification, in the investigations on the determination of the distribution of cooling rates, model studies are used. In the presented studies, the research procedure for the determination of the impact of the cooling rate and the composition of intermetallic compounds was employed as introduced and published by the Alcoa Research Centre. [1]
3
Experimental
Three variations of 3104 alloy (AlMnMg) were selected for the studies. The chemical composition of the studied alloys is presented in Table 1. Table 1: Chemical composition of studied alloys. Alloy No. Si Fe Cu Mn Mg Zn 1 0.27 0.29 0.24 1.23 1.18 0.22 2 0.28 0.69 0.24 1.23 1.21 0.24 3 0.57 0.28 0.24 1.25 1.20 0.23
Cr 0.08 0.11 0.10
Ti 0.04 0.05 0.04
The alloys were prepared in a resistance crucible furnace of crucible volume of ca. 8 kg of the melt. The weight of the melt used in the tests was equal 5 kg. The charge was A1E grade aluminium produced by Aluminium Plant "Konin" whereas alloy additions were introduced in the form of master alloys produced AlTab type of by LSM. On this stage of investigation the grain refiner was not. The temperature of the melt prior to the casting was constant and equal to 700°C. The melt was cast into a permanent mould the schematic diagram of which is presented in Fig. 3. Prior to proper tests, preliminary studies were carried out in order to determine the intensity of the cooling of the cast metal in fixed places positioned in relation to the copper plate constituting the main chill of the permanent mould. A computer-aided thermal analysis stand was employed in the studies. Selected values of cooling intensities are presented in Fig. 3 depicting the schematic diagram of the construction of the experimental permanent mould. It was found that the employed research procedure allows to determine the impact of the cooling rate in the range of 0.5 to 28°C/sec on the form and shape of the components of the structure of the studied alloy. It means the entire range of the cooling rates occurring in the rolling ingots cast by the semi-continuous method into the DC type mould is included.
285
Figure 3: Schematic diagram of test set-up [1]
The research methods were employed which allowed to determine the morphology of the phases, the grain size of α-solution and the identification of phase components. The observation of the microstructure was carried out by a Steresscan 420 Scanning Microscope and microanalysis by a LINK ISIS 300 Microanalyser.
4
Results
The samples cast into the permanent mould shown schematically in Fig. 3. were cut and from the areas in which the cooling rate was determined, samples for metallographic studies were prepared. After metallographic microsections had been prepared, the samples were etched in Barker’s reagent in order to disclose the grain size and thus to determine the effect of the cooling rate on this structure describing parameter. In Fig. 4. a typical relationship between the grain size and the cooling rate is depicted (without grain refiner). Next, the same samples were polished once again and new metallographic microsections were made for the studies allowing to identify the single phases occurring at the boundaries of the solid solution α. As a result of the microscopic observations and the studies of the chemical composition of intermetallic phases it was found that in the studied alloys occur the following metallic phases: Al.(FeMn), α-Al.(FeMn)Si and Mg2Si. Comparing the number of the identified phases, their relative amount (%) in the studied microareas was estimated. The results are summarised in Table 2.
286
28oC/sek
1.6oC/sek
5.0oC/sek
2.6oC/sek
0.9oC/sek
Figure 4: Relationship between the grain size and the cooling rate (50x)
In Figs. 5 to 7, typical structures of the studied alloy 3104 as a function of the Fe and Si content for given cooling rates are illustrated.
Phase α−Al(FeMn)Si a, e, f, g, h
Al 6 (FeMn) b, c, d, i, k, l
Figure 5: Alloy No. 1, cooling rate 0.9°C/sec, Fe/Si = 1
Phase α−Al(FeMn)Si b
Al 6 (FeMn) a, b, c, d
Figure 6: Alloy No. 2, cooling rate 0.9°C/sec, Fe/Si = 2.46
287
Phase α−Al(FeMn)Si a, c, d, e, f, g,h
Al 6 (FeMn) k
Figure 7: Alloy No. 3, cooling rate 0.9°C/sec, Fe/Si = 0.49
Table 2: The identified phases relative amount in the microarea of studied alloys Alloy Phase Cooling rate [°C/sec] 28 5.0 2.6 1.6 0.9 1 α-Al(FeMn)Si 20 30 40 30 50 Fe/Si =1 2 Fe/Si = 2.5 3 Fe/Si=0.49
5
Al6(FeMn) α-Al(FeMn)Si
80 5
70 20
60 20
70 20
50 5
Al6(FeMn) α-Al(FeMn)Si
95 60
80 80
80 100
80 80
95 90
Al6(FeMn)
40
20
-
20
10
Discussion
Analysing the obtained results, it should be remembered above all that the model tests carried out, based on the Alcoa procedure, include the entire range of the changes of the cooling rates occurring in practical conditions during the casting of the studied aluminium alloy of the AlMnMg type into rolling ingots by the semi-continuous method. The above statement makes good reason to put a thesis that the observed correlations between the cooling rate, the chemical composition, particularly the ratio of Fe to Si content, and the type of intermetallic compounds in the structure of the studied alloy and the grain size are reflected in real conditions in industrial practice. The disclosing of the dependence of the grain size on the cooling rate indicates in practice the importance of employing the grain refinement treatment, the correct carrying out of which can result in the decrease of this dependence. As a result of the application of the grain
288 refinement, a homogeneous distribution of the grain size in the entire cross-section of the ingot should be obtained, irrespective of the position, thus of the cooling rate. The results indicate that the percent fraction of the given intermetallic phases mainly depends on the ratio of iron to silicon content within the chemical composition according to the standard in force. This finding offers the possibility of the conscious control of the structure formation of a given alloy taking into account its application and further processes occurring during the working and heat treatment. It was proved by the experimental cast with the objective to form a rolling ingot of the 3104 alloy, with the predominating fraction of the α-Al.(FeMn)Si phase in its structure. According to the research studies, the chemical composition presented in Table 3 was selected. Table 3: Chemical composition of slab ingot 300 × 125 mm of 3104 alloy. No. Chemical composition [wt%] Fe Si Mg Mn Cu Ti 2 Casting 18 0.25 0.51 1.13 1.20 0.28 0.016
Zn 0.02
After the casting in the Institute of Non-Ferrous Metals - Light Metals Division Skawina at the production-experimental stand of semi-continuous casting, the ingot was subjected to metallographic studies. A typical picture of the obtained structure together with the disclosed intermetallic compounds is shown in Fig. 8.
a b Al(FeMn)Si Al 6(MnFe)
c Al(FeMn)Si
d Al(FeMn)Si /Al 6(MnFe)
e Al(FeMn)Si/ Al 6(MnFe
Figure 8: Identification of compounds in ingot cast in the Institute of Non-Ferrous Metals - Light Metals Division
As can be seen, according to the expectations, a structure was obtained in which αAl.(FeMn)Si constitutes the predominating intermetallic phase which was the aim of the experiment carried out. Summarising, it should be noted that the presented studies created the basis for the formulation of definite technological requirements allowing to consciously control the formation of the structure of the studied alloys of 3104 type cast into rolling ingots in industrial conditions.
289
6
Conclusions
Summarising the obtained results it can be noted that: 1. To a large extent, the type of intermetallic compounds is determined by the ratio of iron to silicon content. 2. In the case of the ratio of Fe/Si = 2, the predominating intermetallic compound in the structure of the 3104 alloy is Al6(MnFe) phase, whereas if this ratio is equal 0.5, the predominating compound becomes the α-Al.(FeMn)Si phase. 3. In the studied range of cooling rates (0.5 to 28°C/sec), the cooling rate affects mainly the grain size in the structure of the 3104 alloys and its impact on the type of intermetallic compounds is insignificant, but higher cooling rate is conductives to growth of intermetallic compound type Al6(MnFe). 4. The obtained results on model studies carried out according to the procedure developed by Alcoa were confirmed in industrial conditions upon the studies of the structure of rolling ingots cast by a semi-continuous system.
7
References
[1] P.N.Anyalebechi, Light Metals 1991, p. 821-850
Suppliers Session – Aluminium
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
Horizontal Direct Chilled (HDC) Casting Technology for Aluminium F. Niedermair Hertwich Engineering GmbH, Braunau/Inn, Austria
1
The Universal Caster
HDC casting has well earned its place in modern Al-casthouses, and is still gaining momentum. Hertwich Engineering has successfully commissioned 37 Horizontal Continuous Casting Plants world wide to date. Todays generation of HDC-casting machines is one of the most versatile pieces of equipment, which may be employed to produce any of the following: • Foundry ingot • T-bar • Busbar • Extrusion billet • SSM-feedstock etc.
2
Foundry Ingot and T-bar
Over the past few years especially the mass producers of remelt product have discovered the Hertwich Horizontal Casters to fulfil their demanding needs in terms of product quality and process control. Large scale production of high quality foundry ingot has been shifted from ingot belts to HDC. T-bar casting on Vertical Direct Chilled Casting Machines (VDCs) is loosing ground to the over the years developed Horizontal Direct Chilled casting process from Hertwich Engineering. Figure 1 shows T-bars produced on the Hertwich HDC casting machine.
Figure 1: T-bars produced on Hertwich HDC Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
294 Traditionally T-bars were mainly produced on Vertical Direct Chilled casting machines (VDCs). The VDC process has the following drawbacks compared to the HDC process: • Higher costs of the VDC, especially due to higher cast house necessity of a overhead crane and foundation for the casting pit. • The semi-continuous character of VDC-Casting results in lower productivity. A great amount of set-up work per drop is required, which demands more labour. Whereas with the Hertwich HDC, continuous production runs of 3 to 20 days are common. For T-bar production only one operator per shift is needed. • On VDC plants sawing is not integrated in the process, so that an additional sawing station plus operator is required. Horizontal casting employs a automatic flying saw, which cuts the T-bars to length without disturbing the casting process. • The fully continuous HDC process is ideal for automation. This advantage has been fully exploited by Hertwich Engineering. The downstream equipment is fully in line with the casting process and no additional personnel is required. Sawing, weighing, hard stamping, ink marking, labelling, stacking and strapping is carried out fully automatic. Figure 2 shows casting of foundry ingots, Figure 3 shows automatic marking, stacking, strapping and weighing of foundry ingots.
Figure 2: Casting of foundry ingots
Sows, pigs and ingots were traditionally produced employing the open mould technology. Although this technology was improved over the past years, dross formation and inclusions are unavoidable. The HDC process, however, is absolutely free of any dross formation. It results in savings due to avoided metal losses and in inclusion-free products. On the HDC the metal flows smoothly, protected by an undisturbed oxide layer via launder and tundish to the closed mould. Thus leaving no chance for oxides and other impurities getting into the product. In contrast to that stands the open mould technology. Due to cascading, turbulence occurs when filling the mould. So a relatively big unprotected surface area is offered to the atmosphere for oxidation. The dross formation is mainly ruled by the metal temperature,
295 pouring height and pouring rate. Values achieved during production of pure aluminium sows are shown in table 1.
Figure 3: Foundry ingot automatic marking, stacking, strapping and weighing
Table 1: Values achieved during production of pure aluminium sows. Pouring height [m] Temperature [°C] Dross formation [kg per ton of poured metal] Approx. 0,2 to 0,3 700 –770 0,2 - 0,4 > 800 0,3 - 0,6 Approx. 0,6 to 1,0 approx. 750 2,5 – 4 approx. 850 to 900 5 - 7 The horizontally cast T-bar and foundry ingot are chilled at least ten times faster than sows and pigs. This ensures a fine and uniform grain structure as well as a uniform analysis throughout the cast product. A further step ahead in the production of remelt products in terms of quality is the combination of the HDC process together with an Inline Degasser and ceramic foam filter. Both items can be delivered by Hertwich Engineering to obtain foundry ingot and T-bars free from porosity and inclusions.
296
3
Extrusion Billet and SSM-feedstock
Especially for extrusion mills and “thixoforming die casters”, a Compact Type Remelt Plant, which includes a Horizontal Direct Chilled casting machine, offers commercial and technological advantages for in-house recycling. The capacity of these plants covers a range of 2.000 – 20.000 tpy. Figure 3 shows the schematic of a Compact Type Remelt Plant.
Figure 4: Schematic of compact type remelt plant
In-house generated extrusion or thixoforming scrap can be charged by means of a charging machine into the Two Chamber Melting and Casting Furnace. The stationary furnace consist of a melting and a holding chamber. Applying the submersion melting process permits to remelt profile scrap at a melting loss of lower than 0,5%. Primary metal and clean scrap from the market may be remelt as well. Contaminated scrap, like painted profiles, can be processed in a recently developed 3 chamber furnace. This furnace evaporates and combusts the hydrocarbons from the paint prior to melting. Thus avoiding additional metal losses, increasing the thermal efficiency and destroying harmful compounds like dioxins etc. Through a tap hole in the holding chamber, the metal flows via an Inline Degasser and CFF to the Horizontal Direct Chilled Casting machine. Extrusion billets up to 10“ and SSM feedstock up to 6“ are cast as single or multiple strands. For extrusion billets, the logs are directly fed into Hertwich Continuous Homogenizing Furnace for heat treatment. SSM feedstock is produced on the HDC, by fitting a electromagnetic stirrer around the mould, as shown in figure 4. The solidifying metal is stirred in a helicoidal manner. This avoids growth of fringe crystals and leads to a fine grained and uniform globular structure throughout each slug. The HDC Rheocaster produces high quality SSM feedstock, with reproducible thixotropic properties to an attractive low price.
297
Figure 5: HDC Rheocaster casting 4” SSM feedstock
4
Casting of Forging Stock
Forging stock of 25mm to 125mm diameter can be directly produced in 5 to 30 strands casting process. For many forged products forging stock being free from extrusion grain texture is of great advantage. For many applications the cast bars will require a scalping process. However, HDC casting of forging stock plus scalping is in any case more cost effective than billet casting and extruding.
5
Plant Description and Capacities of the Universal Caster
The mechanical structure of the plant embraces the following major parts: Casting conveyor, flying saw and saw run out system (figure 6).
Figure 6: Major assembly groups of a HDC
298 The Horizontal Caster is the key machine in the effective HERTWICH Compact Type Remelt Plant (figure 7)
Figure 7: Layout of a compact type remelt plant for production of extrusion billets from clean and contaminated scrap
In the primary aluminium field, initially a HDC plant is often bought for producing busbar for potline construction. In phase two the busbar caster is then typically turned into a T-bar or foundry ingot caster to produce high quality remelt products for foundries. The caster can produce up to 13 tonnes per hour T-bar or 8 t/h foundry ingot. A wide range of alloys can be produced, for instance from pure aluminium to 11% silicon and 5% magnesium. Each product type follows its own exit downstream the flying saw (Figure 8).
Figure 8: Typical layout of the Universal Caster
Once all these exit systems have been installed, a product change can be undertaken within one shift, by changing to a different tundish and mould as well as loading the new applicable cast recipe on the PC. Even a changeover to SSM feedstock would be possible by adding a stirrer.
299 The HERTWICH HDC plants are highly automated. They require only one to two operators per shift. Over the past years the plants were improved consistently and feature now automatic cast starts and stops as well as automatic tundish adjustment. The plant is controlled by the Hertwich PCPLC system, which offers an error manager system as well as menu-type casting recipes. Besides, all important plant parameters are monitored and controlled.
6
Conclusion
The Universal Caster from Hertwich Engineering has become a familiar sight in cast houses and extrusion mills. Its versatility, the low investment costs involved, the high quality products produced and requiring few personnel makes this plant unique.
Automatic “Bleed Out” Detection and Plug Off in VDC Billet Casting Manfred Lück Wagstaff Inc., Spokane WA, USA
1
Casting of Extrusion Billets
The Vertical Direct Chill (VDC) semi continuous casting of extrusion billets employs a water-cooled mold through which molten aluminium is solidified and afterwards cooled by the direct impingement of circular water jets or a water curtain. The solidification of the billet shell, providing physical stability for the billet exiting the mold, takes place within the mold bore (indirect cooling). In this area the shell created has to withstand mechanical forces caused be friction, contraction and shrinkage, including the danger of billet surface damage. The advantages of minimal billet shell zones to the extrusion process led to the development of the hot top casting technology, including the use of shorter mold bores resulting in less time for indirect billet cooling. Consequently the achievable shell thickness decreases, and mold bore friction was minimized by installing continuous oil lubrication systems, using graphite as a mold bore material or a combination of both principals. Modern billet casting systems, which today are ‘state of the art’ technology, in addition to this implant an air cushion between the mold bore and the solidifying billet shell, avoiding any direct metal surface/mold bore contact and so again reduce the heat extraction within this area. Consequently the billet’s shell thickness is reduced further, making it more vulnerable to any mechanical damage which can cause an opening of the billet surface and a run out of molten metal. Such defect, known as a “bleed out” , can be a hazardous situation with the high potential of molten metal/water reactions. The start of a cast includes a higher risk for “bleed outs”, which can be related to variations from established casting practice parameters or to the casting equipment not operating correctly. The practice of preventive maintenance and thorough training of operators have combined to significantly reduce, but not fully eliminate , the possibility of “bleed outs”. Localizing the bleeding position can be challenging and the common practice to manually plug of a bleeding billet by hand with a stop off cone brings an operator close to molten metal. To enhance operator safety and to take a further step into the direction of fully automated billet casting Wagstaff developed a technology called Wagstaff StopCastTM Automatic Bleed Out Detection System (further referred to as StopCast)which includes several new patents. Three major goals were identified at the start of the development project in 1997: • A “bleed out” within a casting table needs to be positively identified as quick as possible. • The “bleed out” identification needs to be reliable. • The “bleed out” identified must be plugged off promptly. The new system was dedicated to become an option to Wagstaff’s MaxiCastTM Billet Casting Systems of which more than 550 – mostly equipped with the Wagstaff AirSlipTM Air Casting Process – are operated worldwide today. During the development within Wagstaff’s research department it turned out to be necessary introducing a new trough design, providing Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
301 metal to each mold individually, to make the system work reliable. The StopCast system itself consists of individual dams, located at the entry of each mold, which are raised and lowered by a guided air cylinder. At the start of a cast, the dams are in the up position. With an identified bleed out the air cylinder is activated, rotates the dam to a position above the mold entry, lowers the dam and so shuts off the metal flow to the affected mold. Each dam position is connected to the “bleed out” detection system which is integrated into conventional AirSlip billet moulds.(Although not demanded yet MaxiCast moulds can be equipped with this feature too). The mold has a groove machined into its alignment lugs around the mold’s circumference where a sacrificial nylon tube is placed to act as a detection circuit providing 360° of bleed out detection. The pressurized tube ruptures when contacted by molten metal and the releasing air pressure activates the related StopCast dam via a pilot valve. A control logic assures a lowered dam cannot be opened again until the cast is finished. StopCast was first installed at Alcoa Lafayette, Indiana, USA in June 1998, and has been developed there in a production environment . A second system was commissioned at Aluminium Ranshofen (AMAG) in April 2000 and is in constant operation since, a third system for this plant will be commissioned in December 2000.
2
Summary
A „bleed-out“ is a condition where molten metal runs out of an opening of the billet surface during VDC casting. A „bleed-out“ can be a hazardous situation with the potential of water/metal reactions. When a „bleed-out“ occurs, the bleeding position must be positively identified and plugged off promptly. The use of modern billet casting systems, the practice of preventive maintenance, and thorough training of operators have combined to significantly reduce, but not fully eliminate, the possibility of bleed-outs. Localizing the bleeding position can be challenging and the common practice to manually plug off a bleeding billet by hand with a stop off cone brings an operator close to molten metal. As an advanced step in VDC billet casting a system for automatic bleed-out detection and plug off is presented. Experimental results and field results from the first European application in a production environment are reported.
3
Acknowledgements
Wagstaff Inc. wishes to acknowledge and thank the Aluminium Ranshofen Hüttengießereigesellschaft m.b.H., Austria, especially Mr. Carl van Gils and Mr. Helmut Suppan for strongly supporting this work and contributing to this presentation.
The AIRSOL VEIL® Technology Package for Aluminium Billet Casting Gerd W. Bulian VAW Aluminium Technologie GmbH, Bonn, Germany
Manfred Langen VAW aluminium AG, Bonn, Germany
1
Summary
Mould systems working with air are the dominant technology for casting aluminium billets. To achieve high billet quality, it is important to have a reliable air feed to the mould and a uniform distribution of the melt temperature within a multiple casting unit. The AIRSOL VEIL® Technology Package, consisting of moulds, designed with the aid of pressure-loss calculations and a computerised air-feeding system, guarantees uniform air distribution over the mould circumference as well as optimum feeding of the necessary air-flow rate during casting. Also incorporated in the mould system unit is an optimised metal distribution system (launder) and an advanced design of starter blocks. The launder system, optimised by means of computer modelling, reduces the temperature variation across the casting unit to less than 10° Celsius. The Technology Package is rounded off by a starter block design which avoids the formation of starting cracks in the billets. This and the uniform melt temperature distribution lead to a significant reduction in the cut-off length of the billet butt.
2
The AIRSOL VEIL® Technology Package
The first production unit for extrusion billets and forging ingots came on stream in 1985 [1]. It soon became clear, however, that an air-regulating system operated by conventional flow adjustment alone was insufficient. On the one hand, the correct backpressure in the system must be at the exactly same level as the metallostatic pressure formed by the melt height in the mould. On the other hand, the required airflow rates should remain within a defined range depending on the mould diameter. [2] Conventional multiple casting systems operate with constant airflow rates and the same amount of air in every mould. As noted previously, this should be controlled individually to provide optimum billet quality. The AIRSOL VEIL® Technology Package consists of a specially designed mould, a computerised air feeding system, an optimised launder system and a specially designed starter block. The different parts of the technology will be described in more detail in the following sections of this paper.
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
303
3
The AIRSOL VEIL® Mould
The AIRSOL VEIL® mould is an advanced VAW Hot Top Mould with an optimised indirect and direct cooling system . This type of mould was introduced into VAW casthouses between 1971 and 1974. An advantage of the mould is that there is no dependence on costly special spare parts, such as graphite rings. This greatly simplifies the maintenance procedures. Figure 1 shows two current generation AIRSOL VEIL® moulds, one round and one rectangular. These moulds display a major modification compared with the Hot Top System, i.e. the addition of an air and oil distribution ring. This is shown in Figures 2 and 3.
Figure 1: AIRSOL VEIL® mould
Figure 2: Configuration of rings
304 The channels in the oil and air distribution ring were designed using pressure-loss calculations [3,4]. To ensure uniform air and oil supply, all channel sizes have been carefully optimised. Another advantage is the fact, that no mineral oil is used as a lubricant but only vegetable oils such as rape seed oil, which does not generate any cracked, cancerogenic vapour during casting. Oil residues in the water can be effectively skimmed off by a hydrocyclone system in the water basin. Top of Plate
Bottom of Plate
Air distribution
Lubricant distribution
Figure 3: Air and oil distribution plate
4
Computerised Air-Feeding System
Besides the mould system, the main reason for the high billet quality is the sophisticated aircontrol system, which allows individual pressure control of the air supply to each mould [5]. Using this system, the respective air pressure is always equal to the metallostatic pressure in every mould. In conventional casting systems with an air cushion, the airflow can become too high, which creates overpressure. The resulting surplus air escapes from the mould by bubbling through the melt surface and creating oxides in the melt. If the air flow is lower than the metallostatic pressure, the heat flow through the mould wall is higher; this results in a thicker surface segregation layer. In AIRSOL VEIL® moulds, these quality-impairing effects are eliminated by means of the computerised air-control system, which regulates each individual mould. Figure 4 shows an air-control cabinet for up to 48 moulds. This system is based on a process control computer, which allows the required pressure of the air-volume flow rate to all moulds to be set simultaneously and independently during the continuous casting process. Depending on the selected casting phase, the relevant control-system variable is either the required air-volume rate to the mould (start-up phase with a constant, high air-volume rate, end phase with a constant low air-volume rate) or the pending air pressure in the mould (constant air pressure during the casting phase). The desired values for the maximum or minimum air-volume flow rates, as well as for the air pressure, can be set by means of the relevant input menu on the operator control panel.
305
Figure 4: Air-control cabinet
A special feature of this system lies in the fact that the control-loop parameters are adjustable via the operator interface. This facilitates the individual adjustment of the control loop to the specific requirements, which can vary according to the composition of the alloy and the layout of the compressed-air system. In this way, the pressure loss in the tubing between the control cabinet and the moulds is adjusted to zero by the system.
Figure 5: Visualisation of process parameters
306 Figure 5 shows the main control menu during casting. The screen displays 4 moulds with their parameters simultaneously. This is the optimum number since more would make the screen unintelligible.
Figure 6: Mould menu during the stationary pressure-control phase. Channel #21 is disturbed
The display, which normally appears on the screen during casting, is shown in Figure 6. This figure shows the mould parameters during the stationary pressure-control phase of the cast. In this case, the condition indicator reports a disturbance in Channel #21 (red alarm light). Detailed information about the operating parameters of the affected channel can be obtained through the Control menu.
5
Launder System
The launder system was designed using numerical modelling of the starting or filling time as well as for the casting period [6,7,8]. The optimisation resulted in very even filling of all moulds and also gives a very small variance in the temperature of the metal entering each individual mould. Figure 7 shows the simulation of the metal velocity in the launder. Figure 8 shows the filling of a 48-strand casting unit before and after optimisation. The filling behaviour of the new launder design is considerably improved. Figure 9 shows the photo of a launder system of a 72-strand 7" casting unit. The temperature difference between the coldest and warmest region of this unit is less than 10 °C.
307
Figure 7: Metal velocity of a 48-strand launder
Original
Improved
Figure 8: Filling of a 48-strand launder system before and after optimisation
308
Figure 9: Launder system of a 72-strand casting unit (K 7<6$)
6
Starter Block Design
The design of the current starter block (see Figure 10) was developed with the aid of computer simulation of the cooling and solidification of the starting period during casting. The main advantage of this design is the reduced cracking behaviour of the billet during the starting phase. This resulted in a significant reduction in the cut-off length of the billet butt and improved pit recovery.
Figure 10: Starter block
309
7
Conclusion
The VAW AIRSOL VEIL System is a two-phase casting system which feeds air and oil into the mould to reduce heat extraction through the mould wall. This results in a much smaller segregation thickness and a smoother billet surface. Together with the optimised launder system and the new starter block design a remarkable improvement in productivity is thus achieved. To date VAW has built more than 70 AIRSOL VEIL® units. These units are in continuous operation in casthouses throughout the world.
8
References
[1] W. Schneider, E.Lossack, Improvement of billet quality by use of a hot top mould with a two phase lubrication, Light Metals 1987, (Ed.: R.D. Zabreznik) The Metallurgical Society of AIME, 763 – 768 [2] G.W. Bulian, H.D. Peters, Higher quality billet through pressure controlled air feeding for the AIRSOL VEIL® system, Arabal (1993) VI-4, 1 – 7 [3] W. Schneider, Improved technology of the AIRSOL VEIL® billet casting system, Light Metals 1994, (Ed.: U. Mannweiler), The Minerals, Metals & Materials Society, 985 – 989 [4] W. Schneider, Rechnergestütztes Kokillensystem für das Stranggießen von Pressbarren aus Aluminium, Metall, 1995, 9, 589 – 595 [5] G.W. Bulian, VAW's AIRSOL VEIL® casting system – A tried and tested casting technique with a future, Aluminium, 2000, 76, 298 – 299 [6] G.-U. Grün, I. Eick, D. Vogelsang, Optimal design of a distribution pan for level pour casting, Light Metals 1995, (Ed.: J. Evans), The Minerals, Metals & Materials Society, 1061 – 1069 [7] W. Schneider, G.-U. Grün, The AIRSOL VEIL® billet casting system - Improvements, Mathematical modelling, cast house experiences -, proceedings of the 4th Australasian Asian Pacific Course and Conference Aluminium Casthouse Technology, (Ed.: M. Nilmani), The Minerals, Metals & Materials Society, 1995 297 – 315 [8] G.-U. Grün, W. Schneider, Numerical modelling of fluid flow phenomena in the launder – Integrated tool within casting unit development -, Light Metals 1998, (Ed.: B. Welch), The Minerals, Metals & Materials Society, 1175 - 1182
The Manufacturing, Design and Use of Combo Bag Distributors in Sheet Ingot Casting S.P. Tremblay Pyrotek High-temperature Industrial Products Inc., Chicoutimi, Canada
R. Green Pyrotek Engineering Materials Ltd., Netherton, West Midlands, Great Britain
1
Introduction
In the last decade, the aluminum industry has mainly focused on improving metal quality by working on furnaces and casting practices and especially on in-line treatment units[1,2,3,4]. These developments have modified the function of fiberglass bags widely used for molten metal distribution and macro-filtration of refractory particles or large oxide films. The use of these fiberglass fabric bags often called “Combo bags” for DC sheet ingot casting is not recent. These bags were developed in the seventies when the automatic DC casting systems were introduced. The large bags used so far and called “Channel bags” were interfering with the non-contact level probe located above the starting block. The use of these combo bags solved that problem. However, the manner that the molten metal is delivered to the ingot head is important for the solidification structures of the final cast ingot. For instance, poor bag design or malfunction of the feed system can with certain aluminum alloys result in the formation of coarse-grained regions. The molten metal delivery is not only influenced by the bag geometry but also by its construction such as the type of fiberglass yarn, the weaving and the fiberglass coating used. This paper will review the important parameters in fiberglass fabric manufacturing and metal distribution design to fulfill today’s high quality standards of cast aluminum DC ingots.
2
Fiberglass Fabric Manufacturing
In the aluminum industry, “E” fiberglass is mainly used as raw material for fiberglass yarn. Usually, there are two base yarns used in contact with molten aluminum. The small yarn is composed of 408, 9 microns, filaments. The big yarn is identical in composition and morphology except that it is composed of 816 filaments. These yarns are provided from the supplier with a starch coating to facilitate subsequent operations and with a twist of 30 turns/meter. The number of yarns used during this operation will have an important impact on the chemical resistance of the final woven cloth. For instance, a 4-strand yarn will expose less surface to molten aluminum than a 3-stand yarn. This is even more important because considerable fiber strength reduction is observed as a function of temperature. At casting temperature (~ 700 C), a glass fiber will retain about 15% of its original strength[5,6]. The 4Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
311 strand yarn is recommended for long casting time (>2hours) or for distribution in high magnesium alloys. There are mainly 2 types of open-weave cloths used in contact with molten aluminum. The plain weave style[7] where the warp (lengthwise) and filling (cross wise) yarns cross alternately. The plain weave material is the most common and it is mainly used for dense cloths or fine mesh filtration cloths. The Leno weave style[7] where the warp yarns are twisted around one another locking the filing yarns in place. The Leno style results in greater dimensional stability of the final woven cloth and it is mostly used for large mesh cloths. This latter is mainly used in the manufacturing of combo bags. Finally, the fabric is coated to improve chemical locking, fabric rigidity and resistance to molten aluminum attack. Without coating, the fiberglass fabric would be quickly reduced by molten aluminum. Pyrotek offers 3 different finishes for DC casting applications and they are: Polyvinyl alcohol or PVA, ceramic finish and carbohydrate based finish (F-100). Table 1 shows the different coating properties with respect to smoke generation, toxicity, rigidity and chemical resistance. Table 1: Differences between finishes. Finish type Smoke Toxicity Rigidity PVA High High High Ceramic Low Low Low F-100 No smoke Non toxic High
Chemical resistance High High High
PVA finish is the most commonly used coating for fiberglass fabric applications in contact with molten aluminum. However, in contact with molten aluminum, it generates fumes and toxic materials. The F-100 finish is totally non smoking and non-toxic. It has a comparable rigidity to PVA without the associated health hazards.
3
Combo Bag Design and its Influence on Molten Metal Distribution
The Figure 1 shows a typical combo bag. A combo bag is a complex bag made of several fiberglass pieces sewn together. Some pieces are made of dense fiberglass fabrics while some are made of open-weave fabric. Each of its components will be reviewed and the influence on each component on the metal flow will be detailed.
Figure 1: Combo bag parts
312 The main objective of using a combo bag is to distribute the molten metal into the ingot head which determines the temperature profile around the mold. This latter influences the metallographic structure of the solidified aluminum alloy. Second objectives are to retain oxides within the bag, minimize turbulence and collect any large particles. The combo bag is the last tool in contact with molten aluminum and it must always be within tight tolerances to ensure a consistent molten metal distribution reproducibility cast after cast. All the molten metal distribution patterns shown here have been taken from previous water modeling experiments. There are always a schematic top and side mold view showing metal distribution with the help of small arrows. There is no sump simulation in these Figures because it is very difficult to represent the sump effect in a water model. The main body of a combo bag (1) is an open-weave fabric bag. In conjunction with the outer shell, it controls the molten metal distribution, the spread and speed of metal in the mold for the best temperature profile. There is a difference in flow pattern between a large weave fiberglass fabric and a small weave material used to make the main body. Generally, a finer mesh material increases resistance to the metal passage and slows down the exiting metal speed. The flow is usually more open and it is also less in surface than with a large weave fabric. The outer shell (2) is the solid weave fabric sewn around the main body. Its primary function is to control the direction of the metal outflows, to hold back oxides and to reduce casting turbulence. In general, the shorter the outer shell, the more open the flow. The end patches (3) are made of lightweight solid weave material. They work together to keep oxides within the bag. They work also to direct the flow below the surface of metal. The flow is not noticeably changed by these end patches. However, to take advantages of them, the metal level has to be at least 12-15 mm higher than the bottom of the end patches.
Figure 2: Metal flow without bottom patch
Figure 3: Surface flow pattern in a water model
Figure 4: Distribution of oxide films in an ingot
The bottom and reinforcement patches (4) act together to direct the metal flow sideways and to protect the fabric of the main body from burn-through. They also prevent the combo bag of deforming too much during the cast. Finally, they stop vertical diffusion. The Figure 2 shows the metal distribution where there is no bottom patch. The pronounced curvature shoots the metal towards the surface where turbulence is generated. Then, the metal bounces back under the surface. Deformation under the spout is useful but is has to be limited. If there is no deformation under the spout, the metal will rapidly bounce back to the surface causing turbulence within the bag. Turbulence is always associated with oxide films and inclusions. The Figure 3 shows a typical flow pattern in steady casting regime to illustrate the affect of turbulence on final ingot quality. The exiting metal facilitates oxide film generation which
313 may accumulate in the rotation zone. It is then slowly pushed into the stationary zone. The result on quality is illustrated in Figure 4 showing oxide film distribution in an ingot. Most of the oxide films are found in the middle of the ingot corresponding to the stationary zone shown in the previous Figure. The rest is evenly distributed in the ends. As expected, the oxide films are mainly present at the beginning of the cast and decrease with casting time. Typical oxide films found in DC sheet ingot are chloride, TiB2 and MgO. The side windows (5) control the metal outflow to fill the mold and reduce hot spots on the side of the ingot. There is more flow directed toward the rolling faces. The bottom holes (5) on a 330 mm combo bag have no other function than that of draining the bag at the end of the cast. The metal speed into the bag is high enough that it is like if there were no bottom holes. When the length is increased from 330 mm to 450 mm, the highest surface tension of this longer bag decreases the metal speed. Then, a part of the flow is distributed into these bottom holes as already presented[8]. The only function of the positioning tabs (6) is to correctly position the combo bag around the spout. A slightly non-centered combo will shoot more metal on one side that the other, affecting the temperature distribution around the mold. It is also important to correctly level the combo bag to avoid an uneven flow of metal. The combo bag fork system or fixture has to be robust and functional. Pyrotek has several designs which can adapted to the needs of the customer to ensure a perfect combo bag alignment cast after cast. The influence of the length is illustrated in Figure 5a and 5b between a 330 mm and a 450 mm bag. As explained before, a longer bag decreases the exiting metal speed because of its greater fabric surface tension. However, the bag being closer to the short mold end, the flow is more concentrated on this latter surface than a 330 mm bag.
Figure 5a: Flow pattern of a 330mm combo bag
Figure 5b: Flow pattern of a 450mm combo bag
A spout sock is often used in conjunction with the combo and it consists in a small bag made of a large weave fiberglass fabric which is placed around the spout. This spout sock acts as an additional resistance to the metal flow decreasing the metal velocity and increasing the diffusion surface. A skim dam can also be used in conjunction with the combo to improve metal quality. It is usually used with alloys containing magnesium because they have a tendency to generate more oxide or skim. A skim dam is a machined piece of refractory board usually a N-14, N-17 or B-3 board or it could also be a cast refractory part which is surrounding the combo. Its usual form is an elongated ring. The design and the shape of the skim dam have to be determined by the flow and the type of alloy cast. If it is properly positioned, the skim dam is a very effective tool to keep the dross or generated oxide from the metal transfer between the spout and the combo bag inside. A well-positioned skim dam having a depth between 12 and 19 mm will marginally slow down and spread the metal flow.
314 As previously discussed, it is important to have a uniform temperature distribution around the mold as well as the smallest temperature gradient possible. The Figure 6 shows that with a good combo bag design it is possible to obtain a 7 C gradient compared to a 27 C gradient for a large channel bag used in the same casting conditions. The main purpose of using a combo bag is to improve the final quality of a cast ingot by improving the temperature distribution throughout the ingot heat, by reducing turbulence and holding back metal oxides from entering in the mold.
Figure 6: Impact of design on metal temperature distribution
For best results, the combo bag needs to be matched to the alloy and casting conditions at each customer. Component parts can be changed individually or jointly to give the best results during casting. Water modeling can be used to visualize the effects of these changes on the flow without affecting the production. Pyrotek has a water model available for combo bag design changes along with the technical resources to help the customer to meet today’s highest quality standards of cast aluminum ingots. Pyrotek is working on 2 combo bag development projects. The first one consists of a new manufacturing technique which will lead to a significant decrease of the production costs. The second project is the development of a re-usable combo bag. Preliminary tests have shown an even better molten metal distribution than a regular sewn combo bag. Tests are going on and they will be reported next year. In both of these technologies, patents have been filed.
4
Conclusion
A well-designed combo bag gives low turbulence, minimizes oxides film generation and temperature gradient. There is no perfect combo bag available and each combo has to be matched to the alloy and casting conditions to give best casting results. A better understanding of the fiberglass fabric manufacturing leads to stiffer and long lasting combo bags. Pyrotek has developed a totally non-toxic and non-smoking fiberglass finish to avoid any potential health hazard associated with the use of fiberglass in contact with molten aluminum. Finally, water modeling allows for quick and safe evaluation of different distributor design and set-up and technical resources are available to help the customer to improve its molten metal distribution to meet today’s high quality standard of DC sheet ingots.
315
5 [1] [2] [3] [4] [5] [6] [7] [8]
References Dore, J.C. Yardwood, Light Metal, 1977, 2, 171 - 189 R. Mutharasan, D. Apelian, C. Romanovski, Journal of Metals, Dec. 1981. 12 - 18 G. Dubé & P. Waite, RASELN’91, Japan, 1991 C. Dupuis, R. Dumont, Light Metal, 1993, 997 W.F. Thomas, Chemical Abstract, 1960, 4 - 18 Owens-corning Fiberglass Corporation, Chemical Abstract, 1960, p. 271 Bay Mills Limited, Technical Brochure, 1982, 16 C. Brochu, R. Dault, J. Dery and S.P. Tremblay, Light Metal, 1996, 839 - 844
Recent Quality and Efficiency Improvements Through Advances in In-Line Refining Technology Volker Dopp and Scott Simmons Pyrotek, Inc., Tarrytown, New York, USA
1
Abstract
The first spinning nozzle, in line refining systems were developed by Union Carbide under the SNIF® brand name and introduced to aluminum cast houses in 1974. This patented technology has been widely accepted and employed in all types of cast houses throughout the world. The results have been improved metal quality, reduced system processing time, reduced emissions, and lower overall operating costs. As the quality standards for aluminum products continue to increase, in-line refining is becoming a necessity. These refining systems are customized for the process requirements of individual casting lines. Recent technical advances in in-line refining systems have included: more efficient nozzle design, improved reaction chamber configurations, flexible system configurations, improved chamber access for cleaning, and inerted and sealed reaction chambers. This paper describes some of the subsequent advances and design changes that have improved performance and simplified operation. Case studies are included which demonstrate the expected improvements.
2
Introduction
SNIF® (Spinning Nozzle Inert Flotation) is a recognized leader in aluminum refining technology, with systems operating in all types of cast houses around the world. SNIF’s success is attributable to its effectiveness in degassing molten aluminum, in controlling alkali metals, in removing non-wetted, nonmetallic particles prior to casting, and to its ease of operation and maintenance. Since the first system was installed in 1974, SNIF systems have been continually improved to further enhance performance and to simplify operation and maintenance. Today, a variety of SNIF systems are available to fit most metal casting flow rates and pit layouts.
3
SNIF Aluminum Refining Process Principles
SNIF systems operate by injecting a process gas, typically argon, into the molten aluminum through one or more patented two piece spinning nozzle assemblies (Figure 1).
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
317
Figure 1: SNIF in-line refining principles
The spinning action of the nozzle creates a large amount of small gas bubbles which are thoroughly distributed throughout the molten aluminum, creating a ”well-mixed reactor.” As the bubbles rise, the dissolved hydrogen is de-sorbed into the rising bubbles and floated out of the molten aluminum. The SNIF system head space is inerted to inhibit hydrogen reabsorption and oxidation.
Figure 2: SNIF SHEER furnace construction
When a small amount of chlorine (less than 0.5% of the process gas flow) is added to the process gas, non-wetted, nonmetallic inclusions are separated from the aluminum, attached to the rising bubbles, and are floated to the surface. Chlorine will also react with the dissolved
318 alkali metals (e.g., Na, Li, Ca, and K) to form salts, which if solid, are also floated to the surface. The nonmetallic particles and salts are then removed from the SNIF system by skimming.
4
SNIF SHEER and “R” Refining Furnace Features
The SNIF nozzle is a two-piece design consisting of a stationary sleeve (the stator) and a rotating shaft/rotor assembly. The process gas travels in the space between the stator and the shaft/rotor and is discharged at the base of the stator. As the gas exits, it is sheared into small bubbles by the spinning rotor. The stator reduces the undesirable tendency of the melt to vortex at high nozzle speeds. It also acts as a stationary seal to control the atmosphere above the melt surface. SNIF SHEER technology was introduced in 1993. A bottom rib is added to the bottom of each stage, and the spinning nozzle has been redesigned (Figure 2). The bottom rib stabilizes the metal flow patterns, while the nozzle directs the bubbles down and disperses them evenly throughout the melt. The SHEER nozzle and rib work together to reduce melt agitation, reduce surface splashing, and increase the process efficiency of the refining system, compared to our standard (non-SHEER) R systems. All new SNIF Systems employ SNIF SHEER technology unless specified otherwise by the end user The SNIF R furnace is a refractory-lined vessel. Three furnace walls are constructed of an inner layer of cast refractory and multiple layers of insulating boards. The fourth wall of the furnace consists of one or more graphite heating element blocks and multiple layers of insulation boards. Multi-stage furnaces have one graphite block for each stage. Silicon carbide plates are used to divide the furnace into refining stages. Furnace heating is provided by electric resistance heating elements encased in steel tubes, and inserted into the graphite heater block. The steel tubes; (a) shield the heating elements from the furnace atmosphere, and (b) seal the heater block atmosphere from ambient air. Air is prevented from entering the heater block zone from the refining stage by a refractory protection plate extending from the top of the furnace down to just below the melt idle line. The remaining free surfaces of the heater block are enclosed and purged with nitrogen to prevent oxidation of the graphite.
5
SNIF “P” Replaceable Furnace Cartridge Lining System
When the SNIF refining furnace heater block and/or lining requires replacement, the entire casting pit is usually taken out of service. To minimize this downtime, the first SNIF Psystem was developed and put into commercial service in 1994. ”P” systems are identical to their counterpart ”R” systems except that they incorporate a patented, quickly and easily replaceable pre-fired refractory lining cartridge. The cartridge consists of multiple layers of dense, metal face refractory and highly insulating castable and refractory boards. All refractory is non-wetting and is pre-cured before shipping to eliminate all chemically combined water. The cartridge is enclosed in foil to prevent moisture absorption while in storage and to minimize contact with the insulation during handling.
319 At the end of its service life, the system cover is removed, and the old cartridge simply lifted out of the steel shell. A new cartridge is then installed into the existing steel shell and the cover is replaced. The system is then heated to operating temperature and filled with molten aluminum. Refractory replacement can be accomplished in a few hours, significantly reducing down time. Older SNIF R-systems can be converted into P-systems when the refractory lining needs to be replaced. The P-system cartridge can be equipped with optional immersion heaters or the standard heater block with removable heating elements.
6
Inert and Sealed Reaction Chambers and Cover Lifting Devices
During processing, an in-line refining system creates a large amount of inert gas bubbles, which break the bath surface and cause splashing or exposure of the metal surface to the chamber atmosphere. If the refining chamber atmosphere is exposed to the external atmosphere, hydrogen re-adsorption can occur, making the system less efficient. Furthermore, exposure to oxygen in the atmosphere will lead to the formation of a large amount of oxides in the form of dross. The re-entrainment, due to surface turbulence, of this dross can lead to the formation of inclusions in the metal. In addition to the minimization of bath turbulence as achieved with the SNIF SHEER system, effectively sealing the refining chamber is crucial to the efficient operation of an in-line refining system.
Figure 3: Rear mast cover lifter with graphite block heating system
Figure 4: Four post cover lifter with immersion element heating system
SNIF systems can be equipped with hydraulically lifting covers of a one piece or multisectioned design. The hydraulic covers can either be equipped with a self-contained single
320 mast hydraulic lifter (Figure 3) or a four-post lifter (Figure 4) with separate hydraulic power pack. When the cover is raised, the entire bath surface is exposed to facilitate cleaning. In the closed position, the cover provides an excellent system seal to maintain an inert atmosphere in the refining chambers. Systems using tightly sealed one piece lifting covers in conjunction with the patented two piece nozzle (for cover sealing), trough airlocks and/or under pour baffles experience longer nozzle life and less dross formation. Airlocks and under pour baffles allow the metal to flow freely in or out of the system while preventing air infiltration, the major cause of excessive dross formation and graphite oxidation. As a final benefit, the well inerted and sealed systems permit the elimination of cover gas, since the chambers are sealed to the extent that only the idle gas flow is required to keep them inert. Several SNIF customers using the well inerted sealed systems have eliminated the use of cover gas without any detrimental effect on dross formation, nozzle life, or metal quality. 6.1
Case Study - Benefits of the Inerted and Sealed Chamber Design
The first SNIF system to take advantage of a well inerted and sealed system was installed in Europe near the end of 1992. It was evident from the increased nozzle life and the reduced dross formation that elimination of the oxygen from the reaction chambers was very beneficial to the overall system performance. A sealed SNIF system was installed at Kaiser Trentwood in March of 1996. As a SNIF user for over 18 years, Kaiser has had ample plant experience for comparing the new inerted and sealed system to their other, older SNIF systems. Kaiser found that their dross generation decreased dramatically while nozzle life increased dramatically.1
Figure 5: SNIF RAC System with immersion heating and swivel cover
Previously Kaiser would de-dross after every drop (on average, six times a day, every day of the week) removing approximately 100 kg of dross each time they skimmed the bath surface. In addition, they would have to scrape or chisel a heavy buildup off the chamber walls and cover. On the inerted and sealed system, they needed only to lightly scrape the wall and cover
321 once or twice a week, because there is usually little or no buildup (about 10 kg). Nozzle life at Kaiser Trentwood increased from an average 12.5 days with their older unsealed systems, to 5 months with the new sealed system. In other parts of the world, SNIF customers with well inerted and sealed systems also have nozzles that last from 6-12 months. The cover is typically raised one to three times a week for cleaning depending on alloy, operating conditions, and plant practice.
7
SNIF Rapid Alloy Change (RAC) System
The SNIF RAC System is designed for casting lines where alloys are frequently changed. The RAC System employs a hydraulically powered swivel mast to move the cover to the side for maintenance. A hydraulic furnace tilt apparatus can be added to easily empty the molten metal either through the inlet/outlet ports, or backwards into a sow. The refining furnace is heated with either a heater block or immersion elements. The first SNIF SHEER P-140HB RAC System, employing graphite block heating was installed at ALCOA Europe in Kerkrade, the Netherlands. The first of three (3) SNIF SHEER P60 RAC Systems equipped with ceramic immersion elements (Figure 5) has recently been commissioned at VAW in Stade, Germany.
8
Summary
A variety of SNIF systems are available to meet all cast house requirements. Equipment enhancements have resulted in improved process performance, decreased refractory replacement down time, increased nozzle life, and simplified operation.
9
References
[1] R. Lally, ”Practical Experiences in Metal Quality Improvement” (Paper presented at the Foseco Light Metals Seminar, Owensboro, Kentucky, 3-4 June 1997).
Suppliers Session – Copper
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
Horizontal Continuous Casting of Copper Alloy Billets Michael Brey SMS Meer GmbH, Demag Technica, Veitshöchheim, Germany
1
Introduction
The steadily growing demand for copper and copper alloy products and the increasing cost pressure on metal manufacturers in a global competition demand for • an increase of productivity and plant availability • optimization of product quality and quality assurance • a cut of production costs. These demands represent a particular challenge to the casting process, usually the first step in the production chain. In order to meet or even exceed the above-mentioned demands it is obligatory to apply equipment, which incorporates the most economic and efficient technology. Although the horizontal continuous casting process is well established for the production of brass billets, it offers sufficient chances for further development. These chances are the drive for the research and development activities of Demag Technica. Demag Technica, a subsidiary of SMS Meer, is an ISO 9000 certified manufacturer of continuous casting equipment predominately for the copper and copper alloy industry. The history of Demag Technica can be traced back to companies Krupp, Technica Guss and Knoevenagel, whose know-how on continuous casting is today concentrated under one roof. Within more than 35 years in this business Demag Technica supplied over 500 horizontal and vertical continuous casting plants including more than 70 horizontal continuous casters for copper alloy billets. Nowadays horizontal continuous casting plants are capable of producing billets up to 400 mm diameter in single or multiple strand operation. It is realized however that the tendency, particularly for leaded brasses, goes towards casting larger billet diameters and multiple strand operation. With the example of two recently supplied horizontal continuous casting plants it is outlined how the developments of Demag Technica contribute to the raise of competitiveness of its customers.
2
Horizontal Continuous Casting Plant Technology
Due to the history of Demag Technica two different technologies are available for the horizontal continuous casting of copper and copper alloy billets. Figure 1. shows the layout of a horizontal billet caster applying the so-called COMPUTOCAST technology, which was developed by Technica Guss in the early sixties. This casting line substantially consists of an induction heated holding furnace (1) with the attached moulds (2), a small secondary cooling unit (3) for protecting the downstream Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
326 equipment against the heat influence of the billets, a withdrawal machine (4), a process control cabinet with visualization as well as an in-line saw (5), which allows cutting the billets directly to extrusion length. The roller conveyor (6) behind the in-line saw transfers the cast billets either to a storage table or with an additional transportation system directly to the extrusion press. The COMPUTOCAST technology is characterized by a cyclic withdrawal movement of the strands. The schematic on Figure 1. makes evident that each withdrawal cycle consists of one forward stroke, which is usually followed by one or two waiting times and backward strokes.
Figure 1: Plant layout and withdrawal schematic of COMPUTOCAST system
The HOM-TEC technology, a development of company Krupp, is derived from the vertical continuous casting process. Industrial application of the HOM-TEC technology started in 1980s. Figure 2. reveals that the layout of a HOM-TEC casting plant basically equals that of the COMPUTOCAST system. The major difference between the two processes is the type of withdrawal movement. It is shown on schematic (Figure 2.) that the strands are withdrawn in a continuous movement with constant speed. The holding furnace together with the continuous casting moulds executes an oscillating movement. This led to the name HOMTEC, which stands for Horizontal Oscillating Mould Technology. The withdrawal method and the intensive secondary cooling (Pos. 3 of the HOM-TEC layout) close to the moulds were taken over from the vertical continuous casting.
Figure 2: Plant layout and withdrawal schematic of HOM-TEC system
327 In the following the attention is focused only on some developments, which represent a major contribution to an increase of productivity, optimization of quality and quality assurance as well as a cut of production cost. 2.1
Pressure Controlled Holding Furnace
It is well known that a constant metallostatic pressure above the continuous casting moulds is beneficial for producing high quality castings. Figure 3. shows the schematic of a traditional holding furnace consisting of the slim upper case (1) and the inductor (2), which is bolted to the bottom of the upper case below the opening for the moulds (3). On such kind of furnace the metallostatic pressure above the moulds depend upon the metal filling level. Due to a batchwise transfer of melt from the induction melting furnace the metal level changes during casting. This leads to fluctuations of metallostatic pressure in the mould and strong metal turbulence during refilling, which has a significant influence on the strand solidification, quality and operational safety.
Figure 3: Schematic of traditional holding furnace
The pressure controlled holding furnace overcomes this restriction of the traditional furnace due to its unique three-chamber design. At the same time it offers further benefits, which will be described in the following. Figure 4. shows the schematic of a pressure controlled holding furnace, which consists of the filling chamber, the pressure chamber and the casting chamber. The vertically arranged inductor forms a separate unit between the pressure chamber and the casting chamber. Before start-up of the casting process liquid metal is poured into the filling chamber until the maximum metal level in the three chambers is reached. During casting the furnace empties and in order to maintain a constant melt level above the moulds, pressurized gas is injected into the pressure chamber. The three chambers behave like “communicating tubes”. Gas pressure increases the further the furnace empties. During refilling the gas pressure is released correspondingly. That way the metallostatic pressure above the moulds is kept constant regardless of the filling level inside the pressure controlled holding furnace. Besides the operation of the pressure controlled holding furnace is characterized by very low production of slag as the melt surface is well covered against the atmosphere by the gas tight lid of the pressure chamber. Solid impurities are retained inside the pressure chamber.
328 Therefore inclusions in the cast product are significantly reduced. Turbulence in the melt, which may occur during refilling, does not affect the solidification process due to the large distance between the moulds and the filling chamber. The metal content in the pressure controlled holding furnace can be monitored with a weighing device integrated in the furnace base frame. This patented method allows automation of melt transfer from the upstream melting furnace. At the end of a production run, when the melt level has reached the minimum in the three chambers, gas pressure is released and the melt level drops below the opening for the coolers. Afterwards the billets can be withdrawn from the coolers avoiding long tail ends, which need to be scrapped. The pressure controlled holding furnace provides two drainage plugs on its backside. The upper plug allows quick emergency emptying of the furnace to a metal level below the opening for the moulds. The lower plug allows complete emptying e.g. for carrying out changes of alloy or maintenance work. Traditional holding furnaces without drainage plugs e.g. need to be ladled out until the metal level is below the opening for the moulds. Only then billets can be withdrawn from the moulds in order to allow complete emptying of the upper case and inductor by tilting the entire furnace. Pressure controlled holding furnaces are already applied to six horizontal continuous casting plants of Demag Technica and have proven successful for many years in operation.
Figure 4: Schematic of pressure controlled holding furnace
2.2
“Stationary” Mould Arrangement
The pressure controlled holding furnace features a so-called “stationary” mould arrangement. Stationary arrangement means here that the moulds are mounted to a steel support, which is located in front of the base frame of the holding furnace (Figure 5.). In contrast to conventional coolers, which are bolted to the furnace front plate (see Figure 4.), the stationary moulds are merely fixed to the steel support and pressed against the furnace front plate. A flexible sealing tightens the interface between the stationary coolers and the furnace. The benefit of the stationary arrangement is that thermal and mechanical deformations of the furnace front plate do not affect the alignment of the moulds. Such an unaffected
329 alignment assures the optimum performance of the moulds as well as an extended service lifetime of the graphite dies. A further benefit of the stationary mould arrangement is the easy and quick mould change and alignment. Conventional moulds can be aligned only after being bolted to the holding furnace. For this the entire holding furnace has to be adjusted laterally and in height. This work can only be executed only during standstill of the casting plant. The stationary moulds however can be aligned outside the casting line with the so-called adapter, which carries the mould package. Therefore the time for standstill can be considerably decreased. The prepared mould/adapter assembly is merely put onto the steel support in front of the holding furnace. Once this assembly is mounted to the steel support no further preparation is required.
Figure 5: “Stationary” mould arrangement on pressure controlled holding furnace
The idea of stationary mould arrangement was taken over from the above-described HOMTEC casting process, where it was applied as a standard. Due to the good experience with this particular mould arrangement it is made available for all horizontal continuous casting of Demag Technica today. 2.3
Automatic Cooling Water Supply
The amount of heat extracted from the solidifying strand by the indirect cooling of the moulds is determined by the mathematical product of cooling water flow rate and temperature difference of cooling water between the mould outlet and inlet. During operation the amount of heat to be extracted from the strands experiences fluctuations caused by e.g. periodic superheating of the melt inside the holding furnace or changes of casting speed. On manually adjustable cooling water distributions water flow to each individual mould has to be checked frequently in order to assure constant cooling conditions, because uneven cooling results in uneven product quality. The higher the amount of strands the higher the attention required by the operating personnel.
330 An advanced cooling water distribution system allows automatic regulation of cooling water flow inside the moulds. The cooling water temperature on the outlet of the moulds serves as set value for the actuation of motorized flow control valves. In case this temperature varies from the pre-selected set value, the motorized valves open or close automatically in order to adapt the water flow correspondingly. Each strand is regulated individually. This leads to a production of castings with constant product quality. At the same time the danger of human errors is reduced, as permanent attention of the operating personnel is not required any more. The cooling water distribution cabinet incorporates the motorized flow control valves, temperature measuring devices and flowmeters. This cabinet will be supplied as a completely assembled unit, which must be merely connected to the mains and the interconnecting piping. Thereby installation work on site is significantly simplified. The cooling water temperatures and flow rates are monitored on the process visualization of the central control panel. 2.4
Process Control
Today the requirements on process control systems for continuous casting plants exceed the pure actuation of the withdrawal movement and the remaining equipment. • Increasing labor costs call for automated casting processes accompanied by high operational safety, • Process data has to be archived since international quality standards as e.g. ISO 9000 demand for reproducibility and transparency of the casting process, • Process parameters should be available in form of recallable programs in order to facilitate and accelerate frequent change of alloys and/or cast cross sections, • Beside these technical aspects the costs for control equipment represents another important requirement. Horizontal continuous casting plants by Demag Technica are equipped with PC based process control systems. Each plant component as e.g. withdrawal machine, in-line saw and billet transport is controlled by a separate PLC system. The corresponding control parameters are transferred from the operator panel on the PC via bus connection to the PLCs. The operator panel provides a process visualization, which indicates the most important process parameters on several colored displays (Figure 6.). Thus the entire casting process can be monitored from this panel. The industrial PC offers capacity for storing a sufficient number of casting programs. These casting programs can also be saved on floppy disk or printed out as hardcopies. After start-up of the casting process, either with a stored program or with newly input parameters, the casting plant can be switched over to automatic operation. Optical and acoustical signals indicate if parameters, such as temperatures, flows or pressures leave their pre-selected set values. Exceeding of pre-selected “shut-off values” induce an emergency stop of the casting process in order to avoid accidents. A remote maintenance system allows diagnostics of the PC as well as the PLCs via telephone network. This ensures immediate service without cost intensive delegation of service engineers. Beside the above-mentioned features a number of optional programs and functions can be incorporated in the process control system as the control software as well as the corresponding ancillary equipment are of modular design. This offers the possibility to select only these features and functions, which are actually required.
331
Figure 6: Control cabinet and process diagram
2.5
Electromagnetic Stirring
A permanently growing number of industrial applications require special brasses with superior mechanical and chemical properties. In order to obtain such properties conventional brasses consisting of copper and zinc (CuZn) are alloyed with aluminum (Al), iron (Fe), nickel (Ni), silicon (Si) and/or manganese (Mn). Some of these high alloyed brasses however tend to strong segregation during continuous casting. This can affect further processing and thus the quality of the final products. The application of electromagnetic stirring contributes to the solution of this problem as it leads to a homogeneous and fine grain structure, which is basically free of internal defects like pores and cracks. The induction coil inside the stirring unit produces a rotating electromagnetic field, which causes a rotating movement of the melt when the cast billets pass through the ring shaped unit. Copper alloys usually solidify very quickly due to their good thermal conductivity. Even billets of big diameters are almost completely solidified after they have left the coolers. Therefore the electromagnetic stirring unit is placed coaxially with the cooler as close to the furnace as possible.
Figure 7: Microstructure of a non-stirred and a stirred billet
332 Figure 7 shows two microstructures of continuously cast 220 mm diameter brass billets. The first section represents a non-stirred billet, whose microstructure is characterized by coarse grains and the typical dendritic grain growth. The second section shows the microstructure of a billet, which was cast with electromagnetic stirring. This microstructure reveals that the grain size is very uniform on the entire cross section. The alloying elements are fine dispersed and well distributed in the billet. Furthermore this billet is basically free from internal defects.
3
Conclusions
The above-described developments represent the current state of the art of the continuous casting equipment by Demag Technica. As every product of Demag Technica however is tailored to the customers’ requirements, the process technology experiences a permanent further development. Finally it can be emphasized that horizontal continuous casting plants, no matter whether COMPUTOCAST or HOM-TEC system, featuring a pressure controlled holding furnace, a stationary mould arrangement, an automatic cooling water distribution as well as a modern process control, assure the required competitiveness for a global market. The electromagnetic stirring is a technology, which is can be applied in those cases where the additional investment is justified by the benefits of an improved product quality.
The Outokumpu UPCAST® System L. Eklin Outokumpu Castform Oy, Pori, Finnland
1
Introduction
Copper has a widespread use in various industries such as electrical, telecommunication and information technologies as well as automotive and construction industries. The several physical and chemical properties which make copper the favoured choice include the electrical and thermal conductivities, formability, corrosion resistance and applicability to soldering, brazing and welding. Even though no single property of copper is the most important one, due to application areas in the wire and cable industry the emphasis is often on the high electrical conductivity and formability or drawability of the wire. Coppers can be divided into three different groups, which are pure (unalloyed) copper, alloyed copper and copper alloys. The unalloyed coppers in use are mainly electrolytic tough pitch copper Cu-ETP and oxygen-free copper Cu-OF. Alloyed coppers include copper grades where the copper content is min. 97.5%, these include for example silver-bearing copper. Copper alloys are defined by those coppers which have 2.5% or more alloying elements and examples of such alloys are brass and bronze.
2
Metallurgy of Oxygen-free Copper
According to ASTM B 170-89 standard, oxygen-free copper is defined as a copper with min. 99.95% Cu for standard Cu-OF and 99.99% for Cu-OFE (electronic grade) and containing no more than 10 ppm oxygen for Cu-OF and 5 ppm for Cu-OFE.
Figure 1: Solubility of oxygen into copper
Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
334 By using a very common tool among metallurgists, the phase diagram (Figure 1), the role of oxygen in determining the structure of copper can be easily understood. The solubility of oxygen into copper is approximately 70 ppm at the solidification temperature (1084.9°C) and the maximum solubility of oxygen in copper is around 2 ppm at room temperature. Tough pitch copper Cu-ETP is manufactured in such a manner that the copper melt is purposedly oxidised, typical Cu-ETP contains oxygen in the range of 200-400 ppm. The CuETP cast structure consists of solid solution alfa dendrites and eutectic along the grain boundaries of the dendrites. In the eutectic there is cuprous oxide Cu2O as round precipitates in solid solution alfa matrices. Also, most of the impurities have been oxidised into the grain boundaries. The main principal of manufacturing of oxygen-free copper Cu-OF is that all of the oxygen is removed from the copper by chemical reactions. The usual means are either reduction of oxygen by carbon, or removing oxygen by some deoxidising agent such as P, Al, Ca, etc. The latter method of using deoxidising agents has proven to be rather complicated and to result often in product quality fluctuations. The reason for this is that in order to remove the oxygen efficiently the operator must know the exact amount of oxygen in the melt so that enough deoxidant will be added to the melt. However, if too much is added it will have a negative effect on, for example, conductivity or annealability. Also unnecessary oxides will be introduced into the copper. Removal of oxygen by carbon reduction is the most simple and convenient way of manufacturing Cu-OF. Cu-OF is a single-phase material where both the oxygen (or rather what is left of it) and the impurities originating from the cathodes are in the solid solution. The difference in the grain structure of these two copper grades can be seen in the figures 2a and 2b below.
Figure 2: a) Cu-OF as-cast (160x)
b) Cu-ETP as-cast (160x)
Immunity to hydrogen embrittlement is one key characteristic of Cu-OF. If oxygen is present in the copper and the metal is exposed to a reducing atmosphere at a high temperature, such as for example in welding, the atmospheric hydrogen will diffuse through the copper and react with oxygen and form vapor in the grain boundaries. Such a case can be seen in Figure 3b where ETP copper has been tested for hydrogen embrittlement. It is known that hydrogen embrittlement begins to occur in copper if the oxygen level is around 3- 4 ppm or above. The effect of some of the impurities originating from cathodes on the properties on Cu-OF is harmful. As seen in the figure 4 [1], the electrical conductivity may suffer if certain elements, such as P, Fe, Si, As, etc., are present. In the figure the content of the impurity element is shown on a logarithmic scale on the x-axis, the conductivity in %IACS is on the yaxis on the right, the loss of conductivity in percentage on the y-axis on the left. Also the softening temperature is increased by some elements such as Sb, Pb, Bi, Sn, and P.
335
Figure 3: a) Cu-OF after hydrogen embrittlement test
b) Cu-ETP after hydrogen embrittlement test
Cu-ETP is less affected by these impurities as they are mostly oxidised into the grain boundaries. However, certain critical elements such as Se, Te, Bi and As are not easily oxidised. Also some of the impurities in oxide form in the grain boundaries may cause problems in the casting and hot rolling stages of production of ETP copper. The control of oxygen content is thus very critical and it dependens on the particular copper cathode brand being used. Whenever good quality copper is produced, be it Cu-OF or Cu-ETP, it requires the use of high quality cathodes conforming to the LME Grade A specifications. Practical experience gained at rod mills states that cathode brands differ very much in performance even if required specifications are met.
Figure 4: Effect of certain impurities on the conductivity of Cu-OF
The metallurgical reason behind the excellent formability of copper is the lattice structure which is face centered cubic (FCC) lattice structure. The single-phase oxygen-free copper is
336 more uniform and formable than Cu-ETP [2]. Area reductions of up to 99,99% can be achieved in wire drawing without intermediate annealings as witnessed by some wire manufacturers drawing an 8mm as-cast oxygen-free rod directly down to 0.05mm. However, intermediate annealings guarantee the most consistent quality and homogeneous structure. It should be mentioned also that the creep resistance of Cu-OF is higher compared to that of CuETP [3].
3
Manufacturing
Outokumpu has been manufacturing Cu-OF since the late 1940’s. In the late 1960’s a new concept in casting oxygen-free copper was established with the introduction of upwards vertical casting – the UPCAST® process. Currently over 80 units of UPCAST® for Cu-OF and 40 for alloys are operating worlwide. The highly flexibel UPCAST® process has a proven up-time of over 97%, this being well over traditional CCR processes or a number of less succesful copies of upward continuous casting machines. Variable day-by-day production schedules can be met with no sacrifice to quality. The process is environmentally very safe - Outokumpu has been awarded the EMAS certificate for environmental safety - and clean and can be run economically in the production range of 2000 tons per year to 30000 tons per year and beyond. 3.1
The UPCAST® Process
Figure 5 shows an overview of the UPCAST® line. On the far right behind the furnace installations is the automatic charging machine, which feeds copper cathodes into a channel type induction furnace. The melt is then transferred in batches through a launder into a channel type induction holding furnace. The casting machine is situated above the holding furnace. When casting, the die-cooler assembly is submerged into the melt and the metallostatic pressure forces the molten metal into the die where it rapidly solidifies. The solidified rod is continuously pulled upwards by the withdrawal equipment to end up being coiled in the coilers. The melting and casting are performed in different furnaces due to both capasity and quality requirements. The cathodes charged into the melting furnace as raw material contain oxygen, usually in the range of 10-50 ppm. This oxygen is reduced in the melting furnace by a charcoal cover on the melt. The melt is transferred to the holding furnace through the launder in the presence of a protection gas which prevents atmospheric oxygen from entering the melt. The melt is covered with flake graphite in the holding furnace, performing a “fine-tuning” of the oxygen level. As the oxygen is practically all reduced in the melting furnace and the copper is let to homogenize before pouring to the casting furnace at intervals, the melt quality is consistently high and oxygen level low. UPCAST® rod contains usually about 1-2 ppm oxygen. The process is highly automated and requires only a little amount of manual labour during production. The only critical parameter of the process, oxygen, is largely self-controlled by the process itself, so no process adjustments need to be made during production.
337
Figure 5: The UPCAST® line
3.2
UPCAST® Rod
The copper wirerod at ∅8 mm is only a semi-product to be subsequently processed further and drawn into wire for numerous possible applications. The net shape casting is an advantage because of considerable savings in both capital investments and energy consumption.. In figures 6a and 6b can be seen typical UPCAST® rod structures, in both as-cast ∅8 mm rod 6a as well as in drawn and annealed state 6b.
Figure 6: a). As-cast UPCAST® rod
b) cold deformed and annealed wire from UPCAST® rod
Oxygen free copper is practically free from formation of surface oxidation. When using a galvanostatic measurement system it can be seen that the thickness of the oxide layer is 15-50
338 Ångström in thickness, which is approximately 1/5 of that in ETP. Due to this, the waxing of the rod is not required as it is in traditional CCR processes. The work hardening of the as-cast oxygen-free UPCAST® rod is slightly lower than that of a hot rolled and recrystallized copper, as shown in figure 7 and one annealing after breaking the cast structure is needed for recrystallized small grain strucure. In the diagram the comparison is made between ∅8 mm direct-to-size cast UPCAST® rod and ETP hot rolled into ∅8 mm, but in principal this work hardening diagram is valid for any cross-section size after any further processing stage. For example, after exiting the break-down machine at 2 mm (area reduction ~93%) and annealing, the measured properties of UPCAST ® rod and ETP are the same, but the costly process of hot rolling has not been needed for the UPCAST® product. Wire originating from UPCAST® rod is known to be drawn down to sizes of 15-20 µm in the production of e.g. ultrafine enamelled wire.
500
50
450
45
400
40
350
35
300
30
250
25
200
20
150
15
100
10
50
5
0 0
29.86
46.53
59.36
72.44
85.94
Elongation A100 (%)
Tensile strength (N/mm2)
Work hardening of copper wirerod
UPCAST as cast Rm UPCAST cold deformed/annealed Rm ETP hot rolled Rm UPCAST as cast A100 ETP hot rolled A100 UPCAST cold deformed A100
0 93.11
Area reduction %
Figure 7: The work hardening of Cu-ETP and Cu-OF
4
Process Control
The process control of the UPCAST® system enabels reliable daily operation and is userfriendly to operational personnel. The cathode charging process is fully automized by a PLC (Programmable Logic Control). The furnace inductor power is automized by a very accurate temperature controller system. The maintenance and operational efficieny of the inductors can be monitored from the important R/X values on a constant basis. The control system offers a wide variety of measurable quantities, e.g. all standard measurements (melt temperature,
339 inductor power, water flow, etc.). The PLC is used to collect this data from the field and feed it into the memory and screen of a computer at the plant and all necessary data is available for the further analysis. This provides the management with informative real time production data. Even during total power failure all the control and emergency features work uninterrupted, thus maximizing safety precautions. Also a remote diagnostics system connecting the customer and Outokumpu via modem and a telephone line has been developed.
Figure 8: A view from the UPCAST® control room
5
New Features in UPCAST®
Tighter requirements in plant production economies have led to the development of several sophisticated tools. The latest deliveries of UPCAST® include the possibility to optimize inductor power exactly according to the production capacity. The control system is based on the Siemens S7 logic and the Windows-based interface is easily accessible to operators. The required daily production capacity may be entered as input data and the PLC will then calculate the amount of power needed. This is a handy feature especially at regions suffering from inadueqate power supply and uneven voltage level. As the process control recognizes variations in the voltage fed by the main supply it is able to adjust power accordingly. This power requirement optimisation will be even further improved with a new furnace weight measurement scale which, as its name suggests, recognizes the amount of melt in the furnace. In the actual plant equipment there are some new additions to the existing line. A 24 metric ton melting furnace equipped with four 400 kW inductors will bring the UPCAST® process more accessible to larger capacity producers, as this will bring new possibilites for producers planning to reach capasities up to 30 000 tpa and beyond. New double coilers coiling the φ8mm wirerod into an orbital -shaped pattern has improved the pay off performance of the rod at the breakdown machine. Also new, alternative coiling
340 programs for the old double coiler system have been developed for improved layering and pay-off.
6
References
[1] Lindroos, Sulonen,Veistinen: Uudistettu Miekkojan Metallioppi, Teknisten Tieteiden Akatemia, Helsinki 1986 [2] Opie, Taubenblat and Hsu: A Fundamental Comparison of the Mechanical Behaviour of Oxygen-Free and Tough-Pitch Coppers, J. Inst. Met. 98 (1970) [3] Benson, McKeown and Mends: The Creep and Softening Properties of Copper for Alternator Rotor Windings, J. Inst. Metals 80, 1951-1952
Author Index*
Benum, S. 54 Bergmann, H. W. 109 Birol, Y. 40 Boehmer, J. R. 163 Boiciuc, R. 115 Borge, G. 33 Braathen, S. E. 184 Brandner, D. 61 Brey, M. 325 Buchholz, A. 123, 131 Bulian, G. W. 302
Hayes, J. H. 61 Henriksen, B. R. 184 Holzkamp, U. 94 Hübschen, B. 20
Caesar, C. 169 Cizek, P. 251 Combeau, H. 233 Commet, B. 123, 131 Cook, R. 263 Cooper, P. S. 33
Kara, G. 40 Karkhin, V. 109 Katgerman, L. 77, 138, 199, 239, 276 Keegan, N. J. 20, 26 Kool, W. H. 77, 239 Krone, K. 15 Krug, H.-P. 26 Krüger, J. G. 20
Davidson, C. J. 205, 245 Dopp, V. 316 Dörnenburg, F. 47 Drezet, J.-M. 71, 131, 175 Droste, W. 175 Eklin, L. 333 Engler, S. 47, 269 Eskine, D. 276 Fjær, H. G. 54, 131 Friedrich, B. 15 Grandfield, J. F. 205, 245 Grealy, G. 61 Green, R. 310 Greer, A. L. 149, 154, 218, 257 Gremaud, M. 191 Grün, G.-U. 123, 175 Haferkamp, H. 94 Hardman, A. 26 Härkki, K. 143 Hartmann, D. 269
Immonen, M. 143 Jalanti, T. 191 Jarry, Ph. 233 Jensen, E. K. 61, 184 Jestrabek, J. 15
Langen, M. 302 Lech-Grega, M. 224, 282 Lesoult, G. 233 Lück, M. 300 McKay, B. J. 251 Mortensen, D. 54, 123, 131 Motegi, T. 82 Munteanu, V. 115 Neumann, K. 15 Niedermair, F. 293 Niedick, I. 269 Niemeyer, M. 94 Nosch, E. 15 Opstelten, I. J. 71, 138 Petrache, G. 115 Plochikhine, V. 109 Puschmann, F. 101
*The page numbers refer to the first page of the respecting article Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
342 Rabenberg, J. M. 71 Raihle, C.-M. 89 Rappaz, M. 191 Romanowski, C. 40 Schmidt, J. 101 Schneider, W. 1, 20, 26, 175 Schumacher, P. 213, 251 Simmons, S. 316 Soner Akkurt, A. 40 Specht, E. 101 Storm, J. C. 71 Straatsma, E. N. 77 Stuczy¢ski, T. 224, 282 Swierkosz, M. 191
Tahitu, G. 61 Tanabe, F. 82 Taylor, J. A. 205, 245 Thistlethwaite, S. R. 33 Thorvaldsen, I. J. 61 Tøndel, P. A. 61 Towsey, N. 26 Tremblay, S. P. 310 Tronche, A. 149, 218 Uoti, M. 143 van Haaften, W. M. 239 Vandyoussefi, M. 154, 257 Venneker, B. C. H. 199 Zuidema jr., J. 138
Subject Index* 3D-Modeling 175 3XXX (AlMnMgSi) group, alloys 282 6xxx billets 184 A AA1050 26 AA3004 123 AA5182 123, 239 AA6063 47 AA7075 71 AIRSOL VEIL technology package 302 Al-4.15 wt.%Mg alloys 154 Al 4.5%Cu 233 Alloy additives 33 Alloy development 269 Alloying methods 33 Alloys – aluminum 149, 213, 218, 224, 233, 239, 251, 257, 276, 282 – Al-Ti-B 218 – Al-Ti-C 218 – AlMnMg 77 – (AlMnMgSi)Group 282 – Al-Si-Mg 82 – AZ91 205, 245 – copper 233, 325, 333 – magnesium 154, 245 – nickel 115 AlMgSi0.5 47 Alpha-Al 251 Al-Ti-C inoculants 257 Aluminium 15, 191 – casting technology 293 Aluminium alloys 149, 213, 218, 224, 233, 239, 251, 257, 282 – binary 276 – single-roll strip casting 77 – sheet ingots 61 Aluminum billet casting 302 Aluminum DC casting 123 Aluminum ingots 175 ALZnMgZr 224 Analysis, hot tearing 205 Automatic bleed out detection 300
AZ91 205, 245 B Back-diffusion 191 Billet casting 300 – aluminium 302 Billet quality 47 Billet shell solidification 300 Billets, copper 325 – 6xxx 184 Binary aluminum alloys 276 Bleed out detection 300 Boiling curve approach 138 Boundary conditions, start-up phase 54 Brass billets 325 C Cast magnesium alloy AZ91 205 Cast strip 40 Casting 131 – parameters 40 – practice 184 – trial, alloy AA7075 71 Ceramic foam filter 20, 26 Coherency temperature 276 Columnar transition 245 Combo bag distributor 61, 310 Commercial aluminum alloys 276 Composition, AlMnMg-alloy 77 Continous casting committee 1 ContiSim™ 163 Convection 233 Cooling conditions 131 Cooling rate 282 Copper, properties 333 – vertical continous casting 143 – oxygen-free 333 Copper alloy billets 325 Copper alloy melts 15 Copper alloys 89 Copper ingots 109 Crystal growth morphology 169 Crystallization behaviour 213
*The page numbers refer to the first page of the respecting article Continuous Casting. Edited by K. Ehrke and W. Schneider. © WILEY-VCH Verlag GmbH, Weinheim. ISBN: 3-527-30283-2
344 D Deformations, thermally induced 54 Dendrites 154 Differencing scheme 199 Direct chill, horizontal 205, 245, 293 Directionally solidified Mg alloys 154 Dissolution testing 33 Ductile components 89 E Efficiency improvements 316 Electrochemical hardening 115 Equiaxed transition 245 Extrusion ingot casting 54 F Fiberglass fabric bags 310 Filter 26 Filtration 20 Foam filter 20, 26 Free-growth model 149 G Gas pores 15 Glass 213 Grain refinement 154, 224, 233, 257 – aluminum alloys 149, 218 Grain refiner addition 26 Grain refiner particles 213, 251 Grain refiners 263 Grain structure, copper ingots 109 Grain structures 218 Growth vector 169 H Hardening, electrochemical 115 Heat flow predictions 123 Heat transfer 101 Heterogeneous nucleation 251 High quality rolling ingots 61 High-tensile components 89 Horizontal continuous casting, copper alloy 325 Horizontal direct chill 205, 245, 293 Hot spots 54 Hot tearing analysis 205 Hydrogen concentration 15
I Improved metal distribution 61 Ingot casting 54, 310 Ingot geometry development 175 Ingots, aluminum 175 – copper 109 – sheet casting 131 In-line refining technology 316 Inoculants, Al-Ti-C 257 Instability, TiC particles 257 Intermetallic phases 282 L Linear shrinkage 276 Liquid metal filtration 20 Lubricants 47 M Macro scale, modelling 163 Macrosegregation 191, 233 – modelling 71 – simulation 199 Magnesium 94 Magnesium alloys 154 – AZ91 205, 245 Material defects 15 Material modelling 163 Material properties, alloy AA7075 71 Melt composition 257 Melt pool measurements 123 Melts, copper alloy 15 Metal distribution, improved 61 Metal filtration 20 Metallic powders 33 Metallurgical quality 40 Micro scale, modelling 163 Microsegregation, modelling 71 Microstructure, copper ingots 109 Modelling, macrosegregation 191 Modelling study 184 Molten alloy 82 Morphology, crystal growth 169 Mould 47 Mould systems 302 N Nickel-alloy plating 115
345 Non-uniform spray characteristics 101 Nucleation centers 218 Nucleation process 149 Nucleation studies 213 Numerical diffusion 199 Numerical mass predictions 123 O Outokumpu UPCAST® system 333 Oxygen-free copper 333 P Phases, intermetallic 282 Poisson effect 224 Primary cooling 131 Product quality improvement 163 Q Quality, metallurgical 40 Quality improvements 316 – copper alloy 89 Quenching, water spray 101 R Rapid solidification, aluminum alloys 251 Refining technology 316 Remelt products 293 Rolling ingots 123 S Secondary cooling zone 131 Self-hardening alloys 269 Semi-solid Al-Si-Mg alloy 82 Semi-solid state, alloy behaviour 239 Sheet ingots 61, 131, 310 Shell, solidifying 47 Shrinkage, linear 276 Shrinkage porosity 54 Simulation of macrosegregation 199 Single-roll strip casting 77 Solid state, alloy behaviour 239 Solidification 138, 276, 300 – structures, copper alloys 109
колхоз 5/1/06
Solidifying shell 47 Solute elements 218 Spray characteristics 101 Start-up period 123 Start-up phase 54 – phase, modeling 175 Strains, 6xxx billets 184 Stresses, thermally induced 184 Strip casting, aluminium alloys 77 T Temperature distribution, metal 138 Tensile behaviour 239 Texture, copper ingots 109 Thermal boundary condition, alloy AA7075 71 – DC casting 138 Thermal soft reduction 89 Thermally induced deformations 54 Thermally induced stress 184 Thermomechanical simulations 123 – crystal growth 175 Thin strip casting 263 Thixocasting process 82 Thixoforming feedstock material 269 TiC particles 257 Twin roll casting 40, 263 U UPCAST® system 333 Upwardly solidified Al 4.5%Cu 233 V VDC billet casting 300 Vertical continous casting, copper 143 W Water spray quenching 101 Wear surfaces 115 Y Yield improvements, copper alloy 89
