I

Continuous Casting

Edited by H. R. Müller

Proceedings of the International Conference on Continuous Casting of Non-Ferrous Metals

Deutsche Gesellschaftfür Materialkunde e.V.

II

Further titles of interest:

A. Hazotte (Ed.)Solid State Transformation and Heat TreatmentISBN 3-527-31007-X

D. M. Herlach (Ed.)Solidification and CrystallizationISBN 3-527-31011-8

III

Continuous Casting

Proceedings of the International Conference onContinuous Casting of Non-Ferrous Metals

Edited byH. R. Müller

Deutsche Gesellschaftfür Materialkunde e.V.

IV

Editor:Dr. H. R. MüllerWieland-Werke AGGraf-Arco-Str. 3689079 UlmGermany

All books published by Wiley-VCH are carefully produced. Nevertheless, editor, authors, and publisher do notwarrant the information contained in these books, including this book, to be free of errors. Readers are advised tokeep in mind that statements, data, illustrations, procedural details or other items may inadvertently be inaccurate.

Library of Congress Card No.: applied for

British Library Cataloguing-in-Publication Data:A catalogue record for this book is aailable from the British Library

Bibliografic information published by Die Deutsche BibliothekDie Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliografic data is available in the Internet at <http://dnb.ddb.de>.

© 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Printed on acid-free paper

Printed in the Federal Republic of Germany

All rights reserved (including those of translation in other languages). No part of this book may be reproduced inany form – by photoprinting, microfilm, or any other means – nor transmitted or translated into machine languagewithout written permission from the publishers. Registered names, trademarks, etc. used in this book, even whennot specifically marked as such, are not to be considered unprotected by law.

Composition: W.G.V. Verlagsdienstleistungen GmbH, WeinheimPrinting: Strauss GmbH, MörlenbachBookbinding: J. Schäffer GmbH, Grünstadt

ISBN-13: 978-3-527 31341-9ISBN-10: 3-527-31341-9

V

Preface

In 1976 the Committee for Continuous Casting of the Deutsche Gesellschaft für Materialkunde(DGM) organised for the first time a continuous casting conference for non-ferrous metals. Aperiodical update every 5 years has in the meantime been proven to be useful. Following thevery successful previous meeting in 2000 in Frankfurt, Germany, this year's conference reviewsthe complete range of the processing chain covering both melt treatment and casting technologyas well as specific measures for micro-structural control. A focal point of the programme dealswith modelling and simulation that has become an integral part of modern manufacturing. Cur-rent progress also includes spray forming as an increasingly important processing option. Forthe first time poster presentations are included.

Experts from the manufacturing industry, researchers and scientists from university and indu-stry as well as suppliers of equipment and ancillary products present information on most recenttechnical and economical developments.

The organising committee thanks all authors for their contribution and last but not least the au-dience for their discussion and comments.

Dr. Hilmar R. Müller

Chairman of the Organising Committee

VI

VII

Contents

Melt Treatment 1

Melt Treatment of Copper and Aluminium – The Complex Step Before Casting 3Friedrich, B., Kräutlein, C., Krone, K., IME Process Metallurgy and Metal Recycling RWTH Aachen, Germany

A Study on Surface Defects Caused by Grain Refiners 23Keles, O., Dundar, M., Assan Aluminum, Istanbul, Turkey

Effect of Grain Refiner on Surface Crack of 3004 Alloy during DC Casting 29Morishita, M., Tokuda, K., Kobe Steel, Ltd, Moka-city, Tochigi, Japan

Investigation of Factors Affecting the Extent of Microporosity in an Aluminum Casting 36Savas, O., Kayikci, R., Sakarya University, Sakarya, Turkey

Model Studies of Gas Bubbles Physical Characteristics at Inert Gas Purging into Molten Metals and Alloys 42Stefanoiu, R., Geanta, V., Voiculescu, I., Politehnica University of Bucharest, Bucharest, Romania

Casting Technology 49

Remarks about Process and Technology of Continuous Casting 51Schliefer, H., Khoury, A., Porten, M., Norddeutsche Affinerie AG, Hamburg, Germany; Wolber, P., Boller, K.H., SGL, Bonn, Germany; Dürrschnabel, W., Müller, H.R., Wieland-Werke AG Ulm, Germany; Schneider, St., Deutsche Gießdraht GmbH, Emmerich, Germany; Müller, W.H., Schwarze , M., SMS-Meer, Mönchengladbach, Germany; Oelmann, H., Rode, D., Frankenberg , R., KME, Osnabrück, Germany

Continuous Strip Casting of Magnesium Alloy by a Horizontal Twin Roll Caster 70Watari, H., Oyama National College of Technology, Oyama, Japan; Haga, T., Osaka Institute of Technology, Osaka, Japan; Koga, N., Nippon Institute of Technology, Saitama, Japan; Davey, K., The University of Manchester, Manchester, UK

Strip Casting of Mg-Al based alloy with Ca by Twin Roller Caster 77Matsuzkai, K., Hatsushikano, K., Torisaka, Y., Hanada, K., Shimizu, T., National Institute of Advanced Industrial Science and Technology(AIST), Tsukuba, Japan

VIII

New Strip Casting Process for Magnesium Alloys 81Bach, Fr.-W., Hepke, M., Rossberg, A., Institute of Materials Science (IW), University of Hanover, Germany

Production of Twin Roll Cast AA6016 for Automotive Applications 87Dündar, M., Keles, Ö., Assan Aluminum, Ýstanbul, Turkey; Anger, G., AMAG Automotive GmbH, Ranshofen, Austria

Magnesium Upward Direct Chill Casting 95Bach, Fr.-W., Schacht, S., Rossberg, A., Institute of Materials Science (IW), University of Hanover, Germany

Spray Forming of Advanced High Strength Aluminum Alloys 101Krug, P., Commandeur, B., PEAK Werkstoff GmbH, Velbert, Germany

A Method of VDC Hot Top Mould Design and Setting of Process Conditions 106Bainbridge, I.F., Cooperative Research Centre for Cast Metals Manufacturing (CAST), Division of Materials, The University of Queensland, Brisbane, Australia; Grandfield, J.F., Cooperative Research Centre for Cast Metals Manufacturing (CAST), CSIRO, Division of Manufacturing and Infrastructure Technology, Preston, Australia.

Continuous Casting of Non Ferrous Metal Micro Wrought Shapes 112Bast, J., TU Bergakademie Freiberg, Germany; Bombach, E., Deutsche Solar AG Freiberg, Germany

Influence of Quality of Water and Surface Roughness on Quenching Rate 118Król, J., Specht, E., Otto-von-Guericke-University, Magdeburg, Germany

Electromagnetic Casting of Aluminum and Steel Billet Using Slit Mold 124Park, J., Kim, M., Research Institute of Industrial Science and Technology; Jeong, H., Kim, G., POSCO

Aluminium Alloy Strip Casting Using an Unequal Diameter Twin Roll Caster 131Haga, T., Osaka Institute of Technology, Osaka, Japan; Watari, H., Oyama National College of Technology, Oyama city, Japan; Kumai, S., Tokyo Institute of Technology, Kanagawa, Japan

Fabrication of High Purity Copper Rod with Unidirectional Solidification Structure by Continuous Casting Using Cooled Mold 137Hoon Cho, Duck-young Hwang, Han-shin Choi, Shae K. Kim, Hyung-ho Jo, Korea Institute of Industrial Technology,Yeonsu-gu Incheon, Korea

IX

High Speed Roll Casting of Al Alloy and Mg Alloy Strips 143Haga, T., Osaka Institute of technology, Osaka city, Japan; Watari, H., Oyama national collage of Technology, Oyama city, Japan; Kumai, S., Tokyo Institute of Technology, Kanagawa, Japan

Simulation / Modeling 149

State-of-the-Art in the Modelling of Aluminium and Copper Continuous Casting Processes 151Drezet, J.-M., Computational Materials Laboratory, Ecole Polytechnique Fédérale de Lausanne and Calcom-ESI SA, Lausanne, Switzerland; Gremaud, M., Calcom-ESI SA, Parc Scientifique, Lausanne, Switzerland; Rappaz, M., Computational Materials Laboratory, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland

Numerical Simulation of the Growth of Interacting, Equiaxed Dendrites with a Two-Scale Model 162Jurgk, M., Max-Planck-Institut fur Physik komplexer Systeme, Dresden; Emmerich, H., Siquieri, R., RWTH Aachen, Germany

Monte Carlo Simulation of Grain Growth in Three Dimensions 168Zöllner, D., Streitenberger, P., Otto-von-Guericke-Universität Magdeburg, Fakultät für Naturwissenschaften, Magdeburg, Germany

Thermal Conductivity of Ternary and Multi-Component Aluminum Alloys up to and above the Melting Temperature 174Brandt, R., Neuer, G., Institute for Nuclear Technology and Energy Systems (IKE), University of Stuttgart, Stuttgart, Germany; Bender, W., Grün, G.-U., Hydro Aluminium GmbH, Bonn, Germany

FEM Simulation of Near Net Shape DC Billet of High Strength Al-Mg-Si Alloy 182Nagaumi, H., Takeda, Y., Nippon Light Metal Company Ltd., Kambara, Japan; Suvanchai , P., Umeda, T., Chulalongkorn University, Bangkok, Thailand

Numerical Simulation of DC casting 189Boender, W., Burghardt, A., van Klaveren, E.P., Corus RD&T, IJmuiden, The Netherlands

Continuous Casting of Hypermonotectic AlBiZn Alloys: Experimental Investigations and Numerical Simulation 194Gruber-Pretzler, M., Mayer, F., Wu, M., Ludwig, A., University of Leoben, Leoben, Austria; Moiseev, J., Tonn, B., Clausthal University of Technology, Clausthal-Zellerfeld, Germany

X

Numerical Simulation of the Upward Continuous Casting of Magnesium Alloys 202Landaberea, A., Pedrós, P., Anglada, E., Garmendia, I., INASMET Foundation, San Sebastian, Spain

New Possibilities in the Simulation of Continuous Casting Processes with WinCast-Conti 209Ricken, H., Technische Universität München, Institute of Metal Forming and Casting (utg), Garching, Germany; Honsel, C., RWP GmbH, Roetgen, Germany

Modeling Continuous Casting of Metal Wire Rods 213Chang Hung-Ju, Hwang Weng-seng, Department of Material Science and Engineering, National Cheng Kung University, Tainan, Taiwan; Chao Long-sun, Pan Wensen, Department of Engineering Science, National Cheng Kung University, Tainan, Taiwan; Lai Yi Lin, Metal Industries Research & Development Center

Modeling of Macrosegregations in Continuous Casting of Sn-Bronze 219Gruber-Pretzler, M., Mayer, F., Wu, M., Ludwig, A., Christian-Doppler Laboratory for Multiphase Modeling of Metallurgical Processes, Department of Metallurgy, University of Leoben, Leoben, Austria

Continuous Casting Simulation: From Solidification and Fluid Flow to the Calculation of Grain Structures 226Eberle, R., Wieland-Werke-AG, Ulm, Germany

Mould Temperature Fields during Continuous Casting of DHP-Copper 234Mäkinen, M., Helsinki University of Technology, Espoo, Finland; Uoti, M., Outokumpu Copper R&D, Pori Finland

Simulation of Heat Transfer and Solidification in Continuous Casting of Copper Alloys and the Effect of Fluid Flow 240Vapalahti, S., Louhenkilpi, S., Mäkinen, M., Väyrynen, P., Helsinki University of Technology, Laboratory of Metallurgy, Espoo, Finland

Micro / Macro Structure 247

Spray Forming and Post Processing of Superalloy Rings 249Uhlenwinkel, V., Ellendt, N., University Bremen, Bremen, Germany; Walter, M., Böhler Edelstahl, Kapfenberg, Austria; Tockner, J., Böhler Schmiedetechnik, Kapfenberg, Austria

Macro- and Microstructure of Spray-Formed Tin-Bronze 256V. Kudashov, D., Müller, H.R., Zauter, R., Wieland-Werke AG, Ulm, Germany

XI

Influence of the Crystallization Conditions on the Microstructure and Mechanical Properties of TiAl- and Ti3Al-Based Alloys 265Greenberg, B.A., Kazantseva, N.V., Volkov, A.E., Akshentsev, Yu.N., Institute of Metal Physics, Ural Division, Russian Academy of Sciences, Ekaterinburg, Russia

Effects of Process Parameters on the Characteristics of the Billet Sump and Related Defect Formation during DC Casting of Aluminum Alloys 271Eskin, D.G., Netherlands Institute for Metals Research, Delft, The Netherlands; Katgerman, L., Delft University of Technology, Dept. Materials Science and Engineering, Delft, The Netherlands

Effect of Casting Speed and Grain Refining on Macrosegregation of a DC Cast 6061 Aluminum Alloy 277Kumar Nadella, R., Eskin, D., Netherlands Institute for Metals Research, Delft, The Netherlands; Katgerman, L., Delft University of Technology, Department of Materials Science and Technology, Delft, The Netherlands

Effect of Melt Flow on Macrostructure and Macrosegregation of an Al–4.5% Cu Alloy 283Turchin, A.N., Eskin, D.G., Netherlands Institute for Metals Research, Delft, The Netherlands; Katgerman, L., Delft University of Technology, Department of Material Science and Engineering, Delft, The Netherlands

Quenching Study on the Solidification of Aluminum Alloys 290Ruvalcaba, D., Eskin, D., Netherlands Institute of Metals Research, Delft, The Netherlands; Katgerman, L., Kiersch, J., Delft University of Technology, Delft, The Netherlands

Numerical Study of the Influence of an Applied Electrical Potential on the Solidification of a Binary Metal Alloy 296Nikrityuk, P.A., Eckert, K., Grundmann , R., Institute for Aerospace Engineering Dresden University of Technology, Dresden, Germany

Microstructure and Strain Distribution Influence on Failure Properties in Eutectic AlNi, AlFe Alloys 309Olaru, P., INAV-S.A., Bucharest, Romania; Gottstein, G., IMM, RWTH-Aachen, Germany; Pineau, A., ENSM, Paris, France

Dendrite Coarsening and Embrittlement in Continuously Cast Tin Bronzes 314Virtanen, T., Tampere University of Technology, Tampere, Finland

Continuous Casting of Tin Containing Alloys and their Transformation 320Lebreton, V., Sadi, F., Bienvenu, Y., Centre des matériaux P.M.Fourt, Ecole Nationale Supérieure des Mines de Paris, Evry, France

XII

Suplier Session 327

Horizontal Casting Technology for Copper Products 329Müller, W., Schneider, P., SMS Meer GmbH, Mönchengladbach, Germany

Horizontal Direct Chill (HDC) Casting of Aluminium – The HE Universal Caster 336Niedermair, F., Zeillinger, H., Hertwich Engineering, Braunau, Austria

Aluminium-Semi-Continuous Casting Technic, State of the Art 344Brockmann , G.J., Maerz-Gautschi Industrieofenanlagen GmbH., Tägerwilen, Switzerland

OCP Crucible Monitoring System in Long-Term Tests 353Schmitz, W., Donsbach, F., Otto Junker GmbH, Simmerath, Germany; Hoff, H., Lios Technology GmbH, Cologne, Germany

A New Continuous Casting Process 368Sommerhofer, H., Sommerhofer, P., Sommerhofer Technologies, Graz, Austria

Author Index 377

Subject Index 379

39

Melt Treatment

40

3

Melt Treatment of Copper and Aluminium – The Complex Step Before Casting

B. Friedrich, C. Kräutlein, K. KroneIME Process Metallurgy and Metal Recycling RWTH Aachen

1 Introduction

Aluminium and Copper are second and third in metal production world wide after iron andsteel. But in melt treatment technologies they are probably the number one. Although Copper isthe metal that is longest used technological by mankind and Aluminium one of the shortest, inboth industries a variety of melt treatments and sometimes similar technologies were investiga-ted over time. But as for most industries a look into the technology of a non competing neigh-bour often is missing, because contact is lost due to a lack of connection points. This plenarypaper tries to bridge this gap, here the most important melt treatment techniques are presentedand compared between these two metals. It is accentuated what the industries can learn fromeach other. To see the differences for Aluminium and Copper the most important properties ofCopper are summarised in Table 1.

Table 1: Properties of Copper and Aluminium [2]

The major difference of Aluminium and Copper is the affinity to oxygen. While Aluminiumis a very un-noble element and its melt forms unsoluble oxides rapidly, Copper is considered ahalf noble metal but with a high solubility for oxygen in the liquid state. The major similarity isthe outstanding heat and electrical conductivity of both metals. Although Copper has a 50 %better conductivity than Aluminium the conductivity to density ratio is in favour of Aluminium.This is especially in mobile applications of interest as for example heat exchangers in automo-biles. In comparison Copper heat exchangers are preferably used in stationary and elevated tem-perature applications. For both metals partly similar melt treatment techniques were developed.Impurities like dissolved gases and solid inclusions are battled with the same principles, where-as dissolved metallic impurities have to treated differently. In this paper an overview on themelt treatment techniques of Aluminium and Copper melts are given and gas purging, slagtreatment and filtration as examples are explained in more detail.

Property Aluminium (high purity) Copper (high purity)

Density 2698 kgm–3 8960 kgm–3

Melting Point 660°C 1084°CElastic modulus 70.3 GPa 128 GPaUltimate yield strength (UYS) 90–100 MPa 210–230 MPaElect.conductivity 40 MS 64.5 MSHeat Conductivity 237 Wm–1K–1 401 Wm–1K–1

Electr. Cond./density 14.8*10–3 MSm3kg–1 7.2*10–3 MSm3kg–1

Heat cond./density 87,8*10–3 Wm–1K–1m3 kg–1 44.7*10–3 Wm–1K–1m3kg–1

4

2 Applications of Aluminium and Copper

For increasing applications the Aluminium quality has to meet high performance specificationslike foils, sheets for can bodies and offset plates as well as parts for the production of CDs. Therequirements on the material in respect to cleanliness are very high, e.g. as the thickness of acan body sheet is nowadays only <300 μm. Therefore the number of defects caused by inclusi-ons and gas pores has to be decreased drastically. More than 60 % of the Aluminium productionis used for packaging from which a large part is Aluminium foils. Aluminium foils are used forthe protection of food, e.g. in combination with plastic or paper for the production of juice con-tainers. Inclusions of a size >10 μm lead to holes in the foil and cause spoiled products. Litho-graphic sheets for offset plates have to have a perfect surface.

The highest demands in respect to cleanliness exist for material used for the production ofcomputer discs. At CDs Aluminium is used as material which has to reflect the laser beam. Thereflective material has to have constant reflective properties so that the information stored at thedisc can be read out correctly. The spot where the laser reads the information is only ~ 0.5 μmwide, so the size of surface defects should be as small and their amount as low as possible.

Copper is mainly used for electrical conductivity applications. Besides the day-to-day house-hold uses there are some fields where extreme product cleanliness is necessary. One example isthe use as a cladding material for superconductors as shown in Figure 2. In case of cooling fail-ure (most super conductors still need very low temperatures) the Copper matrix takes over the

Figure 1: High-tech application for Aluminium (Cans, CD, Lithography sheet, beverage containers)

5

conduction because of good deep temperature conductivity until the system is shut off. Contra-rily to the super conductors Copper does not reveal a sharp resistance step at a certain tempera-ture.

Another example for Copper in the high-tech industry is the application in the micro elec-tronics industry. The electronic industry boomed with the development of the so called printedcircuit boards (PCBs) as shown in Figure 2: a Copper foil is applied onto a non conducting sub-strate; alternatively the Copper is deposited electrolytically on the substrate. Then the circuitlines are printed with a special colour, the excess Copper is etched away and the conductor linesremain. Nowadays layers of only 50 μm are applied. The boards with these thin Copper layerscan also be made flexible. This is a great challenge for the material: it has to combine a highconductivity with good mechanical properties and elasticity.One last example for a Copper high-tech application is the use in connector pins, an every day’sbut often not recognised application. In modern automobiles numberless electronic helpers areconnected to one central site. An example of connector pins is shown in Figure 3 Therefore tensof connectors are combined in one connector system. The pins need a high conductivity but also

Figure 2: Copper in printed circuit boards (left) and super conductors applications (right)

Figure 3: Copper strip and connectors (Bilder DKI, KME)

6

flexibility that allows to assemble and disassemble them often. At the same time they should bestiff enough to guarantee the electrical connection even at temperatures in an engine compart-ment and should have the same lifetime expectance as the automobile.

These three fields of application are, of course, only a very small extract of the Copperworld. Copper and its alloys are also used in vacuum switches, vacuum capacitors, electronbeam tubes, welding electrodes, heat exchangers, as moulds for continuous casting of steel,Aluminium and Copper, in generators of power plants and wire.

Aluminium and Copper finished products are sometimes of comparable shape although theirapplication fields are different as for example Aluminium foil for food containers and Copperfoil for conducting purposes. Nevertheless both products require an extremely high amount ofknowledge for their production and because of the effort that was put in their development theyare called high-tech products

3 Impurities in Aluminium and Copper Melts

Impurities in Aluminium melts can be divided into ”solid inclusions” and ”dissolved im-purities”.

Solid impurities in Aluminium have different sources. The exogenous inclusions may comefrom the melt environment as the refractory linings of furnaces, ladles, reactors or launders etc.Mainly these are simple oxides as Al2O3 and MgO, K-, Ca- and Al- silicates, Na-, Ca- Mg- alu-minates, spinels like Al2O3

.MgO or TiB2 cluster originating from grain refining. The endog-enous inclusions for e.g. Al3C4, AlN or AlB2 are formed in the melt during production, e.g. inthe electrolysis cell, at the melt treatment operations esp. during gas purging, or during storageand cooling down steps of the melts. Depending on the material produced the most importantinclusions are Al2O3, MgO and Al4C3.

Dissolved impurities may be foreign metals and dissolved gas. Foreign metals in potroommetal are Na, Li, and Ca coming from the electrolyte. Remelted metal may contain Fe, Si, andCu as impurities. These metals can not be removed industrially and must be diluted by the addi-tion of pure Aluminium or corresponding alloys in the casting furnace. The only dissolved gasin Aluminium melts is hydrogen, because it does not form compounds with Aluminium as othergases (e.g. nitrogen forms AlN, oxygen forms Al2O3). Compared with iron and Copper Alumin-ium has a rather low solubility for hydrogen (at 660 °C liquid Aluminium dissolves 0.69 ppm Hand solid Aluminium only 0.039 ppm H). Hydrogen has to be removed, because bubbles origi-nating during solidification lead to unacceptable gas pores in the produced material. Due to therather small solubility of hydrogen in Aluminium melts its removing is a demanding task. Ele-ments and compounds typically occurring in Aluminium melts are summarised in Table 2.

The issue of impurities in Copper can be separated in two parts: impurities in primary Cop-per remaining or collected after refining electrolysis and impurities in secondary not electrore-fined Copper scrap.

The refining electrolysis produces cathodes with min. 99.995 wt. % Cu, the major remainingimpurities are silver, sulphur, nickel and iron. But the contents are usually so small that they arenot detrimental to the properties of Copper. The more critical elements in this sense namely hy-drogen and oxygen as well as inclusions enter the primary Copper usually during the remeltingand casting process.

7

In secondary materials the impurity matter is more complicated. Remelting of Copper scrapmakes ecological and economical sense because the material does not have to be lead back intothe energy intensive primary electrolysis. There are two types of scrap, the sorted and mostlyclean production scrap which is easily reusable and end of lifecycle scrap (=”old scrap”) con-sisting often of a mixture of different alloys or even compounds with other metals and materi-als. In the process of producing a clean and specified alloy the undesired elements either have tobe removed or diluted. They can e.g. form intermetallic phases in the Copper matrix and as a re-sult decrease the mechanical properties like the ultimate yield strength and the ductility. Due tothe noble character of Copper, most elements like Silicon, Aluminium and iron can easily be re-moved from a Copper melt by selective oxidation up to a very low concentration (activity).Physically and chemically more similar elements like nickel, cobalt, tin and lead have to betreated with more attention. Elements and compounds typically occurring in Copper melts aresummarised in Table 2. Dissolved metallic impurities in small amounts mostly are not detri-mental to the properties of Copper, but some elements as for example Lead and Arsenic precip-itate at the grain boundaries of the Copper materials and lead to embrittlement of the material.

Generally oxygen and hydrogen pick-up can lead to very negative effects. The two gaseshave a high solubility in liquid Copper that decreases sharply during solidification. This canlead to a bubble formation, i.e. porosity in the solid material. Oxygen can also form cuprous ox-ide (Cu2O) above its solubility level that immediately reacts with the moisture of the air formingwater vapour during annealing or welding, this phenomenon is called hydrogen illness. Dis-solved hydrogen and oxygen (or Cu2O) will react to water under extreme pressure in the latticeand will form cracks and lead to embrittlement.

Solid inclusions like intermetallics or oxides from alloying elements or the refractory materi-al usually do not have a negative impact on Copper and Copper alloys. Because the density dif-ference between the Copper melt and the particles is very high the particles tend to float to thesurface (e.g. the density of Copper at 1100°C is 7.96 g/cm³ while iron oxide has a density of5.25 g/cm³ [2]). However Stokes law predicts that even at high density differences very smallparticles tend to stay suspended (In case of FeO particles the diameter smaller than 10 μm risewith 0.59 m/h). This leads to problems where wires with a very small diameter are drawn, forthin Copper foils or where small connector pins are made from thin Copper strip, being etchedor punched.

Generally impurities in metals can be distinguished in dissolved gases, dissolved metals andnon metals as well as solid inclusions, Table 2 summarizes the typical impurity elements andcompounds for Aluminium and Copper.

Table 2: Comparison of impurities in Aluminium and Copper

Impurity type Aluminium Copper

Gas H2 H2, O2Dissolved metals Less noble Li, Na, Mg, K Pb, Sn, Ni, Zn, Fe, Si, Cr, etc.

More noble Fe, Mn, Si, Cu, Ni Au, Ag, PGM

Inclusions Exogenic Al3C4, Al2O3, MgO, Silica-tes, Aluminides

SiO2, Al2O3, SiC etc.

Endogenic Al2O3, MgO, AlB3 Cu2O, MexOy

8

For both metals Hydrogen is a major problem as a dissolved gas, whereas oxygen is unsolu-ble in Aluminium and forms immediately solid compounds. In Copper melts the oxygen con-centration can exceed 1 wt.% and is the second major problem due to the reaction withhydrogen to water vapour or with carbon to CO/CO2. Dissolved metallic impurities are general-ly not a problem as long as they are less noble than the target metal. But as Aluminium is one ofthe least noble elements the variety and amount of more noble metals is much greater than inCopper and Aluminium is difficult to clean. In Copper mainly noble metals like Silver, Goldand PGMs can not be removed from the melt, but also metals with low activities at low concen-trations like lead or nickel, if very high purities have to be obtained. The high oxygen affinity ofAluminium leads to a vast formation of oxides that can harm the products. In Copper ceramicimpurities are mainly from the refractory and from less noble alloying elements not being trans-ferred to the slag.

4 Purity Requirements for High Performance Materials

Potroom Aluminium contains up to 0.3 ppm H, 150 ppm Na, 20 ppm Li, up to 5 ppm Li andmore than 1000 ppm of inclusions, mainly as Al4C3. The ppm-/ppb-/ppt-concentrations of inclu-sions are defined as the total volume of inclusions taken as Al2O3 related to the volume of 1 kgliquid Aluminium. The impurity content of remelted materials sums up to max. 0.6 ppm H2, to40 ppm Ca, 10 ppm Na, and some 1000 ppms of inclusions mainly Al2O3, MgO,MgO.Al2O3,Al4C3 and TiB2.

Alloyed Aluminium melts supplied to the casthouse shall contain not more than 1 ppm inclu-sions, pure Aluminium melts up to 100 ppb inclusions. After filtration this content can be re-duced to less than 10 ppb. Alloyed material for extrusions can contain between a 10–100 ppbinclusions. The highest requirements exist in respect to melt cleanliness for materials used forcomputer disc production. They may contain only between approx. 100–1000 ppt of inclusions.

The alkali-metal content has to be lowered to a few ppms, while the hydrogen content has tobe decreased to smaller than 50 ppb (0.05 ppm). The requirements of different Aluminium ap-plications is shown in Figure 4.

Examples for the increased quality requirements of Aluminium melts are can body sheetsand foils. In the early 1960´s at the beginning of the production of Aluminium can bodies thethickness was approx. 0.5 mm, today it is thinner than 0.3 mm. This means a decrease of 40 %in thickness with a corresponding reduction in weight. Today Aluminium foil is rolled out to athickness of only 6 μm, which is in the magnitude of order of typical inclusions. Without an ef-fective melt cleaning by gas purging and filtration the requirements for the production of thesematerials can not be met.

The most frequent impurity in refined Copper is oxygen. If the alloy is not molten under ox-ygen free atmosphere, Copper melts will always pick up oxygen. Copper melts hyper saturatedwith oxygen lead to the so called hydrogen illness during casting and welding in case hydrogenis present in the material or the atmosphere. Dissolved oxygen can also oxidise less noble ele-ments which leads to the formation of brittle oxides in the material. In technically alloys, an ox-ygen content of a few hundred ppm is accepted whereas in special grade materials the contenthas to be lowered below one ppm, some grades and their respective tolerated oxygen content areshown in Figure 5.

9

In technical alloys the DIN limit for specified metallic impurities mainly varies between 0.1wt. % and 0.5 wt. %. The sum of all not specified remaining elements should not exceed 0.5 wt.%. However some elements are more detrimental as for example Aluminium, arsenic and lead,their content must be kept below 0.01–0.05 wt % [3]. For high quality products limits of < 90ppm Pb, < 50 ppm Sn and < 150 ppm Ni are required. Solid inclusions are especially a problemin thin wire drawing and thin sheets rolling. Equation 1 describes the critical inclusion size, anyparticle bigger will initiate wire breaks:

(1)critical nD KdT

Figure 4: Requirements on Aluminium cleanliness

Figure 5: Requirements for oxygen contents in Copper

10

with dcritical: inclusion size that leads to a break, D: wire diameter, K is a material constant, T: ap-plied pulling tension, n: reduction ratio.

That means that the possibility for a wire break increases sharply with the size reduction ra-tio and the pulling tension. In foils production inclusions lead in the worst case to a foil break.Smaller inclusions lead to surface defects and elongated holes in the material.

5 Processing of Molten Aluminium and Copper

The cleaning of Aluminium melts starts with a simple ladle treatment esp. of potroom metal forthe removal of alkaline metals, before the melt is transferred into the casting furnace. There thealloying is carried out and a further settling operation may take place. From the casting furnacethe molten metal is fed via a launder to the degassing unit for the removal of hydrogen. Grainrefining is carried out by wire injection between the gas purging unit and the filtration station.Sometimes gas purging is combined with a filter in one unit. After the melt treatment the liquidmetal is cast in a DC casting unit to billets, cakes or slugs [5, 6].

The processing of primary Copper usually starts with the melting of Copper cathodes in ashaft furnace. For melting of Copper scrap in general induction furnaces or drum furnaces areapplied. After the preliminary melting furnace the melt is either casted directly or subjected to acasting furnace where the melt is stored, (alloyed) and heated to casting temperature. For Cop-per the continuous wire casting by casting wheels or Hazelett casters is especially important.Besides this also vertical and horizontal slab casting is applied as well as mould casting.

For treatment of Aluminium melts a variety of methods are in industrial use. A cheap andsimple settling procedure in the casting furnace is an easy but ineffective method to clean anAluminium melt. Solid inclusions settle down depending on size, form and density. Because

Figure 6: Processing of molten Aluminium

11

there is only a small density difference between inclusions (e.g. Al2O3) and liquid Aluminiumtheir settling is very slow. Small inclusions do not settle at all. According to the difference inpartial pressure between hydrogen dissolved in the melt and hydrogen resp. water vapor withinthe atmosphere hydrogen can be removed. The return reaction:

(2)

leads to a new hydrogen pick up and oxide formation. Hydrogen and solid inclusions can be re-moved only partially using this method. So settling is only rarely used as a preliminary step totreat Aluminium melts.

By a ladle treatment alkaline and earth alkaline metals can be removed by mechanically stir-ring in salts into the Aluminium melt. Different technical solutions are in industrial application(e.g. the TAC-process). Today ladle treatment is replaced by the development of the RFI proc-esses (see below). Gas purging removes hydrogen as well as solid inclusions, latter only partial-ly by flotation. Also alkaline and alkaline earth metals are removed if chlorine is added to thepurging gas. Melt filtration is used extensively for the separation of solid particles. Elementslike Fe, Si, Mn and Cu, which may be contained in remelted metal in forbidden concentrations,cannot be removed at all and have to be diluted by the addition of pure Aluminium or corre-sponding alloys in the casting furnace.

The ”classical” melt treatment of Copper is the oxidation by air through oxygen injection ortop blowing. With this technique elements that are less noble than Copper can be removed from

2

2 Al 2 3

Hp 3

H O

H O 2Al 6H Al O

aK

p

Figure 7: Processing of molten Copper

12

the melt. Today this technique is often combined with a specific slag that can take up the impu-rity oxides and supplies a certain oxygen potential to improve the impurity separation.

To remove dissolved gases especially hydrogen and oxygen the oldest method is first of allthe right melt handling. Hydrogen can be removed by an excess of oxygen in the melt, oxygenis usually added by blowing air on the melt surface. After the removal of hydrogen by oxidisingconditions, the melt has to be treated under reducing conditions to remove the oxygen. For thereduction since former times the so called poling by tree trunks especially birch is used [4]. Thistechnique is still used today in some places. A new pick-up of hydrogen has to be avoided byshielding of the melt. An advantage of the tree trunks is that they have a CO2 emission of zero,as plants are considered to be regenerative. Alternatively to this classical procedure, moderntechniques reduce the partial pressure in the surrounding atmosphere; this led to vacuum andgas purging technologies like for Aluminium.

The removal of dissolved metals in Copper is an upcoming problem because of the increas-ing recycling material volume not being treated by refining electrolysis. Metals like Zinc, Ar-senic and Antimony can be evaporated by a vacuum treatment, for others like Nickel, Cobalt,Tin and lead a special slag treatment is more economic. Nevertheless very often the high speci-fications of high tech applications cannot be met using only one refining technique. To removesolid particles from Copper melts a simple settling is usually enough for standard qualities. Butthe increasing requirements, e.g. for wire production, led to the development of filtration andflotation techniques.

All the melt treatment techniques are actually batch processes. They have to be implementedin the existing process routes in a way that their effect is not lost before casting and solidifica-tion of the metal. That means that after deoxidation, a pick up of oxygen has to be avoided byproper casting gutters protected by a coal or coke cover or a shielding gas. After filtration a lam-inar flow of the melt through the launder has to be assured to avoid turbulences that promoteabrasion of the refractory and formation of oxides.

Although the impurities in Aluminium and Copper are different, the applicable melt treat-ment principles are almost the same in general gas purging, vacuum treatment, filtration andsettling can be applied. The reactants of the different melt treatment technologies are shown inTable 3.

Table 3: Comparison of melt treatment principles

From both metals dissolved gases can be removed by inert gas purging, where the principlesare actually the same. Most dissolved impurity metals are removed from Aluminium by a treat-

Melt Treatment Aluminium CopperReactant Removal of Reactant Removal of

Gas purging Cl2, Ar, N2 H2, alkalines and floatation of inclusions

Ar, CH4, CO, O2 H2, O2,less noble metals (Fe, Pb, etc.)

Chlorination Cl2 Alkalines , Mg – –Vacuum treatment – Mg, Zn, Pb, H2 – Zn, As, Sb, H2, O2Filtration Ceramic Foam

(Al2O3)Inclusions Ceramic foam

(SiC)Inclusions

Settling ”Time”Inclusions (Al2O3, MgO etc.)

”Time” Inclusions(Cu2O, MexOy)

Slag treatment NaCl, KCl Inclusions (Al2O3, MgO etc.)SiO2 etc. Pb, Ni, Mn, Fe

13

ment with Chlorine while in Copper melts an oxidising procedure is favoured. With a vacuumtreatment dissolved gases as well as highly volatile elements like Zinc, Arsenic etc. can be re-moved. The filtration removes solid inclusions in both metals only the type of filter material isdifferent reflecting the different properties of the metals. Settling is a simple but standard tech-nique to avoid solid inclusions in the melt.

5.1 Gas Purging

First mentioning of gas purging of metallic melts goes back to 1856 for steel [7]. The gas pur-ging of Aluminium melts was mentioned first by D.R. Tullis in 1928, who used pure chlorine.Mixtures of chlorine with inert gases esp. nitrogen were developed very soon. In 1931 Kochproposed the use of a mixture of chlorine and nitrogen for the removal of Fe and Si from com-mercial Aluminium alloys [8]. Based on the basic work of Röntgen and Haas the chlorine/nitro-gen converter was developed and set in operation in 1948 [9]. A couple of those units have beenused up to the early 1960s in Europe. In 1964 the trigas mixture was developed. Carbon-mon-oxide was added to the chlorine-inert gas mixture to lower the Aluminium-oxide formation atthe inner surface of the bubble. This enhanced the transport of hydrogen through the gas-meltinterface. After the development of the gas mixtures the research was focused on the technologyof gas purging.

In the early times of Aluminium melt treatment simple tube lances were used to introducethe purging gas into the melt. Jet injection technology was developed already in the 1970ieswhere using a high-speed jet of gas is injected into the melt via nozzles. The gas is dispersedinto fine bubbles and distributed in the reactor. Porous plugs were introduced in the Aluminiummetallurgy in 1973. Porous plugs are mounted into the furnace technology, so their applicationis limited. But, they are widely spread in the Aluminium industry. At the beginning of the 1990sporous plugs were placed in launders.

In the middle of the 1970ies the rotary gas injection (RGI) technology was developed by dif-ferent companies nearly at the same time. The principle of this technology is the fact, that a gasstream introduced into a melt via a high speed rotor is disintegrated into very small bubbles byshearing forces. A couple of different in–line systems were developed which differ mainly inthe design of the rotor [10]. These units are built in form of boxes, which can be fitted into themelt treatment line easily. They are in use in casthouses worldwide.

The latest development in the middle of the 1990ies was the rotary flux injection (RFI) tech-nology, in which salts, replacing chlorine in metallurgy, are added to the inert purging gas andare injected via a rotor into the Aluminium melt. The main target of this development is a de-crease of chlorine consumption and emission. In the course of the development of gas treatmentsystems the chlorine consumption decreased from up to 0.7 kg Cl/t Al using lances, over 0.1–0.2 kg/t with the RGI-system down to 0.05 kg/t in the RFI-systems.

Gas purging is based on the difference in the partial pressures of hydrogen dissolved in themelt and within the bubbles of the purging gas. The purging gas, usually nitrogen or argon, isintroduced into the melt by lances, nozzles, porous plugs or high–speed rotors. A bubble formede.g. at a pore of a porous plug has a hydrogen partial pressure of nearly zero. Hydrogen atomsdissolved in molten Aluminium are transported to the bubble by convection and via diffusionthrough the melt-gas boundary layer. There the dissolved hydrogen atoms combine to gaseoushydrogen by chemical reaction. The ascending bubble becomes larger because the metallostatic

14

pressure decreases and hydrogen is taken up, theoretically until the thermochemical equilibriumis reached. Normally the retention time of the gas bubbles in the melt is not long enough toreach equilibrium conditions.

The effectiveness of gas purging operations depend on the kinetics of the reactions duringthe degassing process. The speed of the hydrogen removal can be described by a first order re-action and roughly by the equation:

(3)

Therefore the decrease of the hydrogen concentration in the melt c depends on

• the retention time t of the bubble in the melt,• the mass-transfer coefficient ,• the melt volume V, and • the mass-transfer area A (most important).

The mass-transfer area A is the total surface of the bubbles in the melt during gas purging.Consequently the formation of as many and small bubbles as possible in units is essential. Fur-thermore the depth of the melt is important, because the retention time of the bubbles in the meltis determining too.

While using porous plugs a careful adjustment of the gas-throughput is necessary. Only atslow gas velocities small bubbles are formed; at high velocities rather large bubbles are pro-duced because the whole plug surface acts as a bubble source (so called ”flooding”). The small-est bubbles can be produced by the application of high-speed rotor systems Lances are almostineffective for gas purging operations.

Hd

d

c x A

t V

Figure 8: Principle of gas purging

15

State of the art in Aluminium melt treatment is the application of degassing boxes which areused worldwide. They are installed in-line between casing furnace and grain refining unit. Fur-ther developments of the RGI technology are the launder resp. through degassing units usingalso rotor systems.

Compared to the earlier used degassing boxes the launder (compact) degassing technologyhas following advantages:• Reduced production costs by

– diminished metal losses,– decreased process gas consumption and– decreased depreciation due to less expensive equipment.

Figure 9: Gas bubble surface versus bubble diameter

Figure 10: Efficiency of gas purging methods

16

• Increased cleanliness of the melts by– decreased hydrogen content to below 0.05 ppm and– improved inclusion content of lower than 20 ppb.

• Reduced chlorine emissions.• Reduced space consumption.

All further developments must have the following targets:

• Increasing the effectiveness of hydrogen (and inclusion) removal by– lowering the bubble size,– increasing the bubble residence time and– improving the bubble distribution in the reactor

• Further decreasing of operation costs• Decreasing the chlorine emissions to zero, which is the main challenge for the future.

The purging technologies of Copper alloys can be divided in those with inert gases and thosewith reactive gases. Purging with inert gases is based on a low partial pressure of the gas thatneeds to be removed. This process is diffusion controlled, i.e. the speed depends on the diffu-sion constant and the specific surface area of the melt-bubble interface. A diffusion controlledmass transfer can be influenced by rising temperatures (technically not feasible) and a decreasein the thickness of the Nernst layer. Such the process can significantly be intensified by an in-crease of the surface of the gas bubbles by appropriate gas supplying technique. The principlesfor the gas purging was described above. Argon and nitrogen are appropriate inert gases forpurging Copper, especially for hydrogen removal. Their solubility in liquid Copper is negligi-ble.

Also gas purging methods for Copper deoxidation were developed to replace the poling withtree trunks that goes back at least to the year 1200 AD when it was mentioned in ”De Re Metal-lica”. In the 1960s extensive research was conducted on reactive gases with different gaseousand solid reducing agents. Reducing gases were tested like different carbon-hydrates [11], car-bon monoxide as well as hydrogen. Ammonia [12] for gas purging was investigated in the

Figure 11: In line Aluminium gas purging: SNIF Box and Alcan compact degasser

17

1970s. Also oil or coal dust were used as reducing agents, the problem of these substances wastheir impurity content, especially sulphur that enriches in the Copper melt.

Natural gas as a reducing agent was tested by Klein [13] but the process was found to bevery ineffective and slow. In comparison a reformed, partially oxidised natural gas where thecarbon hydrates are reacted to carbon monoxide and hydrogen leads to a fast deoxidation of theCopper melt. This process was implemented at the Phelps Dodge refining works in the 1960saccording to US-patent 2,989,397. An intensive investigation on the kinetics of the Copper de-oxidation by carbon monoxide was carried out by Andreini et al. in 1977 [14]. The chemicalgross equation of the deoxidation with carbon monoxide is:

(4)

It was found that the oxygen diffusion in liquid Copper to the gas bubbles is the rate control-ling step, valid in a concentration interval of 50 to 1000 ppm. Below 50 ppm the kinetics of theoxygen removal decreases sharply. This can not be explained by equilibrium reasons, it is possi-bly due to interactions with sulphur in the melt. This study agrees fairly good with an olderstudy from Nanda et. al. [15] which also found a sharp decrease of the deoxidation speed belowa concentration of 50 ppm oxygen. The lowest oxygen concentration in liquid Copper that couldbe achieved by this process was 10 ppm. As an alternative to carbon hydrates ammonia was in-vestigated as a possible reducing agent by Henych et al. [12]. The gross equation of this reactionis:

(5)

The reducing element in this case is the hydrogen. The developing nitrogen is an inert gasand does not dissolve in the Copper melt. The lowest oxygen concentration in liquid Copperthat could be achieved by this process was 200 ppm. This technique never reached industrialscale.

Up till now the favoured gas injection technology in the Copper industry are tuyeres/injec-tors and top blowing lances even though in other industries different gas supplying techniquesare used as rotary vaned dispersers and porous plugs

5.2 Slag Treatment

A slag treatment of Aluminium and its alloys is usual only necessary when very fine particleslike chips are molten. The natural Aluminium oxide skin prevents the burning of Aluminiummelts up to 700 °C. In Aluminium recycling where fine fractions are molten usually salts on thebasis sodium chloride, potassium chloride are used. The mixture depends on the local depositsand suppliers. This system is selected, because the melting temperature is close to the meltingtemperature of Aluminium but at the same time the system has a rather high evaporation point.This mixture has a better wetting behaviour for oxides that for Aluminium metal and thereforetakes up oxides during a melt treatment and last but not least it is readily available and cheap.Usually to the salts on chlorine base fluoridic compounds like AlF3 or CaF2 etc. are added. Thefluorides accelerate the cracking of the Aluminium oxide layer and therefore improve the coale-scence [15]. Oxidic slags are not used in the Aluminium industry because of their high melting

22O CO COCu

3 2 223O 4NH 2N 6H O

18

point and also because Aluminium is very un-noble and tends to reduce most of the slag oxidesas for example Silicon oxide.

Slags in Copper refining have the task to take up the impurity oxides, formed during the oxi-dation procedure. Their properties should be:

• High solubility for impurity oxides• Low solubility for Copper and Copper oxide• Melting temperature close to the melting temperature of Copper• High thermal stability• Low interaction with the refractory material

Before or during a slag treatment the Copper melt has to be oxidised for example by topblowing of air. A slag that is industrially used is the so called ”Fayalite” slag, which is also ap-plied in the primary Copper metallurgy. It is based on the system FeO–Fe2O3–SiO2 and effectivefor the gross removal of Cd, Fe, Pb, Sn and Zn especially for elements with the valence of twothat can be trapped as silicates. This slag can be used in furnaces with silica refractories [17].This is the slag used mainly today.

On laboratory scale other slag types were investigated as well, for example Calcium-ferriteslags which are based on the ternary system FeO-Fe2O3-CaO. They remove the elements Al, As,Fe, Sb and Sn, especially elements which exhibit an acidic character in a slag [17] at their high-est oxidation levels. This slag can be used with base refractories. Another system is based onCaF2-CaO-MgO-SiO2 this slag shows the same behaviour as the ”Calcium-ferrite” slag but witha lower solubility for Copper oxide.

Also investigated by researchers were salt slags that are mainly based on Sodium carbonate(Na2CO3) but also on other alkaline carbonates like Lithium carbonate (Li2CO3) andPotassiumcarbonate (K2CO3). They are very effective refining slags, but they all attack ”usual” refracto-ries of the Copper industry. It was found that by fluxing with a Sodium carbonate slag theamount of arsenic and antimony could be lowered below 0.1 ppm. The problem that occurredwas, that the binary solution of Sodium carbonate and Antimony oxide let to an increase of sol-ubility of Copper oxide in the slag and therefore to high Copper losses [18]. For fluxes on othersalt base mainly fluorides can be suitable because their melting and evaporation point is veryhigh. Here especially systems based on Calcium fluoride (CaF2) mixed with Aluminium fluo-ride (AlF3) or Sodium fluoride (NaF) are possible candidates.

All fluxes operate in the best way when the reaction surface between metal and slag is in-creased for example by solid flux injection in the metal melt [19] instead of an addition to thesurface.

5.3 Filtration

In 1935 a procedure was proposed for the filtration of light metal melts by DEGUSSA, whichwas transferred to Aluminium melts very soon. The bed filtration (BF) was developed usingbulk petrol coke and/or ceramic particles by ALCAN in the 1940´s. The development of cera-mic foam filters (CFF) started in the beginning of the 1970´s by SELEE. First rigid media filters(RMF), which are called also bonded particle tube filters (BPF), appeared on the market in the1980ies, but were initially not accepted by the aluminum industry. In the 1990ies two stage fil-

19

ter systems were developed having a much better particle removal efficiency. The latest advan-ces in filtration technology is the development of surface active filter systems starting in the mid1990s. By the formation of active surfaces inside the filter itself the effectiveness for the separa-tion of small inclusions was significantly improved.

For the filtration of molten metals the same laws apply as for aqueous dispersions. Two dif-ferent kinds of filtration have to be distinguished: cake and bed filtration. Usually both filtrationtypes occur combined and happen successively. In the case of cake filtration the filtration proc-ess itself happens at least at the beginning by sieve effects. First inclusions larger in size as thepore diameter of the filter settle on the filter surface forming a thin layer. The thickness of thecake increases as more melt flows through and more inclusions are separated. For Aluminiummelt treatment cake filtration is rather unusual and limited to melts with high inclusions con-tents (> 200 ppm) and larger inclusions.

Bed filtration is the common mechanism used for Aluminium melts. In this case the separa-tion of inclusions from the melt is rather complex. It happens mainly by direct collision or adhe-sion of particles to/at the filter surface, sedimentation by gravity as well as by inertia forces,collision of particles by Brown´s movement or/and fluid dynamic effects. Up to now no closedtheory exist of the filtration of Aluminium melts [20]. So a mathematical modeling, whichwould allow calculating filtration efficiencies, filtration times, filter sizes etc. is not yet possi-ble.

The filter materials are generally refractory material, preferentially Al2O3. They can be dis-tinguished between bed filters (BF), ceramic foam filters (CFF) and rigid media resp. bondedparticle tube filters (RMF resp. BPF).

Bed filters (BF) are bulks built of Al2O3-balls or chips with a size of 2–8 mm. Bulks of car-bon or coke are not more used. BF´s are separate in-line units, which need rather much space.They are built in externally heated boxes and are suited for the throughput of large amounts ofmelts up to 1000 t. Particle form and size, layer thickness, and the sequence of different layersare varied to improve the filtration effectiveness. BP filters are suited for the separation ofsmall inclusions < 20 μm from melts with low inclusion concentrations.

Figure 12: Principles of melt filtration

20

Ceramic foam filters (CFF) consist of a labyrintic structured ceramic material in which avery effective cleaning of the Aluminium melt happens by deep bed filtration effects. They areproduced by the infiltration of a ceramic sludge into porous polyurethan foam. During firing theplastics decomposes and the porous ceramic remains. CFF are also built in separate boxes,which must not be heated externally. Filter plates are commercial available in different sizes,thicknesses (normally 50 mm) and pore sizes between 10 to 80 ppi (pores per inch). They areone-way products and rather cheep so that the operational costs are low. This filter type is usedmost often in Aluminium metallurgy [21].

Rigid media filters (RMF) consist of porous ceramic tubes, which are built in, as BF´s in ex-ternal heated boxes in form of pipe bundles. The melt flows from the outside to the inside of thepipes. The filtration processes are very similar to the CFF`s. Because they have a smaller poresize there is a higher pressure drop and they have a shorter lifetime. RMF´s are rather expen-

Figure 13: Filter devices for Aluminium melt filtration (from left to right: CFF, RMTF, BF)

Figure 14: Technology of Aluminium melt filtration

21

sive in respect to investment and operation cost, therefore their application is limited to specialapplications.

State of the art is the application of deep bed filtration in casthouses, where large amounts ofthe same alloy have to be cleaned. For general purposes CFF´s are used. Only for special appli-cations RMF´s are in operation because they are most efficient for the removal of very smallparticles. In normal filter systems single CMF plates or combination of CMF´s with differentpore sizes are built in in separate boxes which must not be heated externally. They are posi-tioned in-line directly before the DC unit. The melt is allowed to run through the filter plate,mostly downwards. CMF´s are used in combination with BF´s and with degassing units, too.Additional ceramic filter clothes may be used at the DC casting unit to retain coarse impuritieswhich can come into the melt after filtration. Targets for the filter development in future are fil-ters that can remove even finer particles with high efficiency at a reasonable pressure drop andwith minimized metal losses.

Filtration of Copper is not very common, because the density difference between the Coppermelt and the characteristic solid inclusions is big enough that they settle in the casting furnace.But because of the increasing demands on the semi finished products as for example Copperstrip with 50 μm the amount of very small particles < 10 μm that do not settle easily has to be re-duced. In 1981 the filtration of Copper melts was first mentioned by Chia et. al. Today filtrationof Copper melts is mainly applied for material that is drawn into wire [22]. The filtration ofCopper melts will increase with a further demand on wire and strip thinness.

6 Summary

In this paper the today melt treatment techniques for Aluminium and Copper are presented. Em-phasis was laid on gas purging, slag treatment and filtration because these techniques are widelyspread in both industries. Until now the development of melt treatment techniques always wasin step with the demands of the semi finished product producers and also with the demands for

Figure 15: In line Aluminium melt treatment: Hydro (VAW)-Filter

22

environmental safety. But the future holds more challenges. Because of size reduction foils andwire diameters will further decrease. For the Aluminium industry the avoidance of Chlorine is amajor point that needs to be solved. For Copper the removal from dissolved elements fromscrap charges has priority to avoid the energy intensive refining electrolysis.

The principles of melt treatment give both industries guidelines in which direction to proceedand to produce the optimum metal quality.

7 References

[1] G. Armstrong Smith, Journal of the Institute of Metals, 100, 1972, 125–130[2] R. Blachnik, D´Ans Lax: Tachenbuch für Chemiker und Physiker: Elemente, anorgani-

sche Verbindungen und Materialien, Minerale; Springer, Heidelberg 1998[3] CEN/TS 13388:2004[4] E. Brunhuber, Schmelz- und Legierungstechnik von Kupferwerkstoffen”; 1959, Schiele

und Schön[5] P. Waite, in Light metals 2002 (Ed.: W. Schneider), TMS, Warrendale Pa. 2002; 841 ff[6] W. Schneider, H.-P. Krug, N. Keegan, Aluminium, 43, 1998, 40 ff[7] Steward, Christopher, Forward, US Patent 23123 (1859)[8] W. Koch, Zeitschrift für Metallkunde, 1931, 34, 95 ff[9] W.Dautzenberg, Metall, 1950, 4, 125 ff[10] K. Krone, Aluminium Recycling, VDS, Düsseldorf 2000[11] F.E. Brantley, C.H. Schack: Bureau of mines R.I. 6113, 1962[12] R. Henych, F. Kadlec, V. Sedlacek, Journal of Metals, 1965, 386–388[13] L. Klein, Journal of Metals, August, 1961, 545–547[14] R. J. Andreini, J.S. Foster, R. B. Phillips, Metallurgical Transactions B, 8B, 1977,

633–638[15] C. R. Nanda, G. H. Geiger, Metallurgical Transactions, 2, 1971, 1101–1106[16] M. Gerke, Dissertation RWTH Aachen, Shaker, 2002[17] J. Gortais, F. Hodaj, M. Allibert, J.-M. Welter, Metallurgical and Materials Transactions

B, 25B, 1994, 645–651[18] H. Fukuyama, T. Fujisawa, C. Yamauchi, Metallurgical Processes for the Early First Cen-

tury, The Minerals, Metals & Materials Society 1994, 443–452[19] B. Zhao, N. J. Themelis: “Kinetics of As and Sb removal from molten Copper”; EPD

Congress, The Minerals, Metals & Materials Society, 1995, pp. 515–524[20] B. Hübschen Dissertation, Shaker, 2004[21] W. Schneider; Gießen von Aluminium-Werkstoffen; VAW Aluminium AG[22] Chia et.al., Us Patent 4277281, 1981[23] B.P. Kamath, R. Adiga, L.K. Sharma, S. Gupta, Copper2003 – Cobre 2003, Fifth interna-

tional Conference, 719–727, 2003

23

A Study on Surface Defects Caused by Grain Refiners

O. Keles, M. DundarAssan Aluminum, Istanbul, Turkey

1 Abstract

The role of grain refiners has been gaining importance in improving characteristics of as-castand end wrought products. Ti-Al-B master alloys have been used for grain refining in aluminumcastings. Agglomeration of the grain refiners causes streaks, which especially affects the sur-face quality of sheets.

Nowadays, customer requirements on surface quality of aluminum sheets have been gettingmore and more demanding. In order continuous aluminum casting technology to compete withDC cast counterparts, products have to be free of surface defects and cost effective. Although,there are many sources for aluminum sheets surface defects, in this study the defect caused bygrain refiners has been examined.

Cast house practices showed that these agglomerates are likely to settle down in launders,nozzles and plugs the filters used. In this study, samples, gathered from ceramic foam filters,casting nozzle, launders as well as end product have been investigated by utilizing an opticaland electron microscopy in order to show agglomerations and their effect on the surface ofproducts.

2 Introduction

In aluminum alloys TiAl, TiAlB, and TiCAl base products have been used for grain refinementfor years. There have been several theories proposed to understand the grain refinement mecha-nisms in aluminum alloys. Some studies have been focused on the production and applicationmethods of these alloys as well as compositions in order to make the grain refinement processefficient and to avoid defects in cast products [1–14].

The chemistry, sizes, shapes, and distributions of grain refiner constituents, feeding rate aswell as feeding position play important roles in reaction efficiency, agglomeration and settlingbehavior of grain refiners. It is known that, as TiBAl based grain refiners consists of TiB2 andTiAl3 phases. Even though TiAl3 dissolve in time, TiB2 phase is known to be stable and abun-dant in aluminum melts [1,4,9,10,15–17].

The sizes of these phases and their concentration in grain refiners vary depending on compo-sition as well as production methods [5,7,13,17,18]. TiAlB base products have a tendency to ag-glomerate and to settle down after a certain period of time.

In aluminum casting, continuous casting technology has become preferable due to its lowcapital investment costs, reasonable productivity measures and low product costs. Technologi-cal developments in continuous casting have been focused on reducing strip thicknesses. Twinroll casting technology offers to cast thin gauge strips [19–21]. In order to utilize the advantagesof this technology, process and product assurances have to be provided.

24

Continuous casting being younger than direct chill casting technology, grain refinement ap-plications are less understood [22,23]. One of the main problems seen in grain refining mecha-nism is the agglomeration of grain refining particles in the components of molten metal deliverysystem [24,25]. These agglomerates can block filters, settle down in launders and caster tips byleading to feeding problems during casting. In addition to this, even flush into cast strips, resultin surface defects in aluminum sheets and foils.

In this paper, to illustrate the complications arising from grain refiner agglomeration filters,caster tips and launders as well as end products are examined.

3 Experimental Procedure

In the present study, 150 tons of AA1050 and AA3003 aluminum alloy is cast by utilizing FataHunter Speed caster. Before a casting campaign is begun, launders are cleaned and a newceramic foam filter and a caster tip are introduced. Launders, ceramic foam filters and castertips are made of silica, alumina, alumina-silica based products respectively. During casting,TiAl5B1 grain refiner rod is fed into a molten metal at the filter box before ceramic foam filterwith the feeding rate of 40 cm/min for a strip having 2120 mm width.

After 150 tons casting, samples are gathered from the metal left in the launder, the ceramicfoam filter and the caster tip. These samples are mounted, grinded and polished for OM exami-nations and then coated with gold, investigated by utilizing SEM &EDS. Furthermore, a sheetand a foil sample are investigated in order to see the faults produced by grain refiners on an alu-minum sheet and also a foil.

4 Results and Discussion

In Figure 1, the pictures of the launder, the ceramic foam filter and the caster tip utilized for thepresent study is shown. First visual inspection of the components has shown that bottom of thelaunders and caster tip have experienced heavy lump and cluster formation. In all parts, their si-zes differ depending on grain refiner feeding rate, casting speed, the amount of metal cast in onecampaign. The sizes of clusters for this study are found to be in a range of 2–20mm in height.Since TiB2 has a greater specific density than that of aluminum and also has a tendency to ag-glomerate, they settle down at the bottom of the materials used for transfer. Although feeding isconducted before the ceramic foam filter, these clusters are observed in the launder as well as in

Figure 1: a) Launders b) ceramic foam filter c) caster tip

25

the caster tip which are located after the filter. It is noteworthy to mention that during castingthese clusters has no or light attachment to the transfer media. During casting, these clustersmay be dislodged and introduced into the cast strips. Even small amount of particles escape intothe cast strips can cause defects as the strips are rolled.

High concentration of Ti compounds and other insoluble particles tend to form a reticulatedlacey pattern. These pattern can be seen in each materials consumed for this study. In Figure 2the cross section of the filter, the clusters taken from the metal left inside the tip and the launderas well as the metal left at the edge of the tip can be seen.

As noticed in Figure 2, the characteristic feature in all microstructure is having a lacey net-work. These networks are also seen in Mabry et al.’s study [26].

In all SEM studies, these boundaries are found to have oxygen, titanium, iron, manganeseand vanadium. In Figure 3, the sizes of TiB2 particles vary from 0.5–4 m. There are alsoneedlelike structures identified to be Al-Fe-Mn phase (see Figure 4). Similar, phases are seen inseveral studies made for AA1050 alloy. Since in the campaign without changing the transfermedia both AA1050 and AA3003 alloy cast, it is likely to see Al-Mn-Fe intermetallics insteadof AlFe ones [15,27,28].

It is well known that fading occurs on prolonged holding times and grain refinement loss canbe reversed by agitation of the melt [29,30]. With the help of agitation TiB2 particles can be sus-pended into the melt and start nucleation process for a second time. This could be the way to in-crease the efficiency of grain refiners but it is important to note that agitation can also cause towash out the agglomerates in products cast. Decreasing the feeding rate of grain refiners couldbe another solution to prolong the settlement period and the quantity of agglomerates in thetransfer media. The tradeoff, in this case, should be made between the quality of coils and thequantity of grain refiners. Lastly, with changing the sizes of constituents in grain refiners, theshapes of grain refiners and finding a way to suspend TiB2 little longer in the melt could makepossible to increase the reaction time as well as grain refiners’ efficiency [5,30].

Due to vibration or erosion in caster tips, the titanium compounds, believed to be TiB2, flushinto cast strips and hot rolled with aluminum. It is hard to distinguish these insoluble particles in

Figure 2: a) filter, b) tip section, c) tip corner, d) launder, e) and f) tip section polarized

26

as cast strips. The processes applied in lithographic sheet production easily reveal those defectsexisting on the surface. They become extremely noticeable as black streaks on the surface afterbrushing, etching, and anodizing operations. In Figure 5 a lithographic sheet having a blackstreak on its surface and related optical microscope and EDS analyses are given. The width ofthe streak is in a range of 1–3mm and the length is approximately 150–200 mm.

As the strip gauge decreases, the clusters bring about a much negative influence. Contrary tothe ductile aluminum matrix that can be easily elongated during rolling, TiB2 particles have lim-ited ability to be deformed. This mechanical incompatibility leads to premature strain accumu-lation within the TiB2 clusters and results in early failure initiated from those areas. While theycan be observed as very small crack on the surface of sheet products, extreme cases results in

Figure 3: EDS analyses a) a cross section of the filter b) the metal left in the tip section

Figure 4: An EDS analysis of needlelike structures

27

high pinhole counts and at the end rupture in foils. In general, the pinholes display consecutive“v” shape appearances (see Figure 6 ).

5 Conclusions

The influence of AlTiB grain refiners in the filter, the launder and the tip used for the casting ofAA1050 and AA3003 alloy campaign have been examined. It is concluded that:

Grain refiners tend to form clusters on the filter, the launder and the tip used. The sizes ofthese grain refiner accumulations may change from 2 to 20 mm in height. Their size maychange depending on the amount of aluminum cast and grain refiner feeding rate.

These clusters have TiB2 and Al-Fe-Mn phases in the metal left after casting. TiB2 particleshave formed a lacey layer structure and the intermetallic phases seen in the grains formed byTiB2 particles.

Figure 5: Micrographs of a lithographic sheet. a. image scanned, b.OM picture, c. EDS analysis

Figure 6: SEM and EDS analysis of foil having a pinhole

28

Especially in continuous casting, as these cluster are introduced into the products due tochanges in metal flow, they cause a black streak in lithographic sheets while showing pinholesand ruptures in foils.

Further studies are being made in order to eliminate or reduce the amount of clusters by un-derstanding the role of contact time, feeding rate and grain refiner type in twin roll casting.

6 References

[1] P. Schumacher, A.L. Greer, Light Metals. 1995, 869–877[2] J.A. Meggy, D. A. Granger, G. K. Sigworth, C. R. Durst, Light Metals. 2001, 943–949[3] M. A. Kerans, S. R. Thistlethwaite, P.S. Cooper, 125th Annaheim Annual Meeting. 1996[4] J.J. Del Campo, M. Martin, L. Galan, Light Metals. 1999, 797–802[5] K. Venkateswarlu, B.S. Murty, M. Chakraborty, Materials Science and Engineering.

2001, A301, 180–186[6] A.M. Detomi, A.J. Messias, S.Majer, P.S. Cooper, TMS. 2001[7] L. Peije, E.G. Kandlova, A.G. Makrenko, V.I. Nikitin, Y. Zhang, A.R. Luts, Materials

Letter. 2004, 58, 1861–1864[8] A. Hardman, F.H. Hayes, Materials Science Forum. 1996, 217–22, 247–252[9] H. Jin, T. Sagstad, Taft , P.T. Zagierski, Materials Sience Forum. 1996, 217–222,

241–246[10] M. Johnsson, Light Metals. 1993, 769–777[11] L. Backerud, S. Yidong, Aluminum 67, Jahrgang 1991,780–785[12] P. Cizek, B.J. Mckay, P. Schumacher, Continuous Casting. 2000, 251–256[13] G. T. Campbell, S. A. Danilak, Light Metals.1992[14] K.T. Kashyap, T. Chandrashekar, Bull Mater. Sci. 2001, 24, 345–353[15] X. G. Chen, Light Metals. 1999, 803–809[16] T.E. Quested, A.L. Greer, Acta Materiala, 2004, 52, 3859–3868[17] C. Limmaneevichitr, W. Eidhed, Materials Science and Engineering. 2003, A355,

174–179[18] Venkateswarlu, S.K. Das, M. Chakraborty, B.S. Murty, Materials and Engineering. 2003,

A351, 237–243[19] O. Daaland, A.B. Espedal, M.L. Nedberg, I. Alvestad, Light Metals 97. 745–752[20] P. Vangala, R. Duvvuri, D. Smith, C. Romanowski, TMS 1993, 333–349[21] M. Cortes, LightMetals. 1995, 1161–1164[22] P. S. Cooper, P. Fisher, Light Metals 1994[23] M. Yun, S. A. Lockyer, J.D. Hunt, Light Metals 2000. 857–861[24] D. J. Pearson, P. Cooper, 6th Australian Asian Pasific Conference. 1999[25] S. R. Thistlethwaite, P. Fisher, TMS, 1995, 249–259[26] G. Mabry, J. Kaems, D. Granger, W.C. Setzer, Light Metals. 1997, 983–989[27] M.W. Meredith, A.L. Greer, P.V. Evans, R.G. Hamerton, Light Metals. 1995, 811–817[28] T. Stucczynski, M. Lech-Grega, Light Metals. 2003, 961–968[29] P.S. Mohanty, R.I.L. Guthrie, J. E. Gruzleski, Light Metals. 1995, 859–867[30] C. Limmaneevichitr, W. Eidhed, Materials Science and Engineering. 2003, A349,

197–206

29

Effect of Grain Refiner on Surface Crack of 3004 Alloy during DC Casting

M. Morishita1, K. Tokuda2

1Kobe Steel, Ltd, Kobe-city, Hyogo, Japan2 Kobe Steel, Ltd, Moka-city, Tochigi, Japan

1 Introduction

An ingot cracking of aluminum alloy during DC casting is one of the most serious problems asless productivity. Therefore to control casting conditions and micro/macro-structure becomesvery important to relax stress in the ingots, because the cracking has been occurred by a solidifi-cation shrinkage, a thermal contraction and a stress concentration within a mushy state.Remarkable progress has been made over the last few years in the numerical simulation of ca-sting and solidification process, and the methods also have had the possibilities to estimate aprediction of stress with good precision [1,2]. However, without quantitative data of the mecha-nical behavior of alloys at high temperatures particularly in the mushy state, it is not possiblepredict to whether a cracking during casting will occur for a given stress and deformation [3,4].

Generally, even if the alloys are the same, ingots of finer grain materials hardly cause crack-ing problems. But fracture criterion is often not described quantitatively in the literature.

Therefore 3004 alloy used as can body was taken up in the present study and the objectiveswas settled next threefold: the first is to identify the crack initiation and clarify the cracking be-havior, the second is to quantify the effect of grain size to mechanical properties in the mushystate and the third is to clarify the criterion of ingot cracking used by numerical simulation insome casting conditions.

2 Observation and Measurement of Mushy State Mechanical Properties

2.1 Observation of Surface Cracking Within a DC Ingot

A surface crack of the actual DC ingot is shown in Figure 1. A fracture surface of the ingot isshown in figure 2. The minute crack on the ingot surface becomes initiation and the surface mi-nute crack progresses to the centerline of the ingot. Figure 2 also shows the surface crackingwas occurred in the mushy state since the fracture near the ingot surface is covered by dendriticmorphology.

It indicates that thermal contraction at the ingot surface is higher than the elongation of a ma-terial in high temperature. It is very important for cracking prevention to control the surfacestrain level by casting conditions and material properties especially elongation by grain size.

30

2.2 Micro-Structure inside Ingot Surface with/without Grain Refiner

Figure 3 shows that the difference of the cracking behavior between fine and coarse grainsample used by RING test method. Grain refiner which was Al-5%Ti-1%B rod was added50ppm Ti into the molten metal to prepare "fine grain" sample. Even if the same test condition,„fine grain“ sample is obviously less breakable.

In order to cast different grain size samples, the same grain refiner condition was applied tothe actual ingot casting. Figure 4 shows the micro-structure inside ingot surface with/withoutgrain refiner and they are also different grain size.

Chemical compositions of two different grain boundaries were measured to predict meltingpoint by Thermo-calc (Table 1). The coarse grain sample has more micro-segregation than fineone. The calculation results show that coarse and fine grain samples are about 26 °C and 15 °C

Figure 1: Surface crack of an actual ingot 600mm x 2000mm about 7000mm length

Figure 2: Shape of fracture surface and micro structure (600mm thickness)

Figure 3: Difference of cracking behavior between with and without grain refiner used by RING tester

31

lower solidus temperature at grain boundary than average solidus temperature of 3004 respec-tively.

Table 1: Chemical compositions of 3004 alloy and grain boundary, and calculated solidus tem-peratures

2.3 Experimental Procedure to Measure Mechanical Properties of Mushy State

Two different grain specimens are taken from ingot surface, and strength and elongation meas-urements were carried out using a developed semi-solid tensile testing instrument which isshown in Figure 5. In order to minimize microstructure changes that occur during heating, a re-sistance furnace is employed to attain rapid heating. Target temperature of each tensile test isranging from 500 °C to 620 °C. Strain rates are 0.01 s–1. Dimension of a specimen is68 mm 8.0 mm diameter and gauge length is 45 mm. To prevent liquid dropping off frommushy area, specimens are covered by a 8.1 mm inside diameter quartz tubes with thermocou-ples.

(mass%) Fe Si Mg Mn solidus temp. (calc.)3004 alloy (average) 0.45 0.25 1.20 1.10 0.2 617 oCgrain boundary fine grain 1.77 0.57 2.16 1.87 0.1 602 oC

coarse grain 1.90 0.88 3.37 1.77 0.0 Cu

Figure 4: Difference of microstructure inside the DC ingot surface between with and without grain refiner (600mm x 2000mm ingot)

Figure 5: Schematic of semi-solid tensile testing instru-ment

32

3 Results

3.1 Mechanical Properties of Mushy State

The stress and strain values vary significantly with temperature and grain except under solidustemperature as shown in figures 6 and 7. They indicate that fine grain materials have high elon-gation values and coarse grain ones have no elongation between 3004 solidus temperature andsegregated solidus temperature. Accordingly the finer grain material must be hardly crackedeven if the ingot are distorted much. Numerical simulation which is to predict distortion andstress are carried out based on the above semi-solid tensile test data below 3004 solidus tempe-rature.

4 Discussion

4.1 3D Numerical Simulation for Cracking

In advance of three dimensional (3D) simulation, unknown boundary conditions such as heattransfer coefficient between a mold and melt interface were fit into actual experimental datawith two dimensional simulation. Then the boundary conditions are applied to 3D simulation inorder to predict a thermal and stress histories. The code is ABAQUS and boundary condition isshown in figure 8.

Since the distortion and stress are strongly effected by not only solidification behavior in thecross section but also vertical temperature variation, a model was designed as follows. Firstlymolten metal shape into rectangular prism, secondly the rectangular prism moves through thewater cooling mold region and direct chill region, and the thermal distribution are calculated ineach time step and finally stress and distortion are calculated based on previous thermal results.

The dimensions of ingot are 2000 mm width, 600 mm thickness and 4000 mm length. Thecasting rate is 50 and 70 mm/min. And amount of water flow rate is 3.9 and 1.5 l/min.cm.Figure 9 shows the example of temperature and distortion distribution. In the figure, location

Figure 6: Ultimate tensile strength of 3004 alloy as a function of temperatures at two different grain material

Figure 7: Elongation of 3004 alloy as a function of temperatures at two different grain material

33

2/4 W, 1/4 W and 0/4 W are at the center, the quarter and the edge respectively. These locationswere chosen because during the casting cracks are most likely to start at these locations.

4.2 Criteria of Cracking Depend on Grain Size and Cooling Conditions

As the ingot cracking progressed from surface of the ingot, calculated surface distortion is com-pared with measured high temperature strain data. In this comparison, liquid at grain boundaryis assumed to be movable freely, and the stress is also assumed to start to build up at 3004 solid-us temperature 617 °C. Therefore calculated distortion values are reset at the temperature to

Figure 8: Boundary conditions calcurated by 2D numerical simulation

Figure 9: Example of 3D simulation results : Rolling surface, 70 mm/min, 3.9 l/min cm(a) distortion distribution of horizontal direction (b) temperature distribution

34

compare with tensile test data. Rested upon the comparison, it is possible to evaluate the materi-al stability for cracking.

The comparison between ingot distortion and material strain is shown in figure 10. Whereasthe strain of fine grain material is always higher than surface distortion of the ingot at all loca-tions, the strain of coarse grain is lower than surface distortion between 3004 solidus tempera-ture and solidus temperature of grain boundary. Consequently it is clarified that the fine grainmaterials are stable against cracking in high temperature and the adding grain refiner is very ef-fective to prevent the crack.

Figure 11 and 12 show results of low speed casting and low water cooling respectively. Thetwo results indicate that both conditions will reduce the distortion. It is possible to conclude thatless crack will occur when we cast by low speed and low water cooling in the mold, because ofless distortion in high temperature state.

Figure 10: Comparison between calculated distortion value and experimental strain data (70mm/min, 3.9 /min cm)

Figure 11: Comparison between calculated distor-tion value and experimental strain data (50 mm/min, 3.9 /min cm)

Figure 12: Comparison between calculated dis-tortion value and experimental strain data (70 mm/min, 1.5 /min cm)

35

5 Conclusions

1. The minute crack on the ingot surface becomes initiation and the crack progresses to the cen-ter line of the ingot.

2. The fine grain materials which are with grain refiner have less micro-segregation and high elongation in high temperature. Therefore fine grain material must be hardly cracked even if the ingots are distorted much.

3. By 3D simulation, it is clarified that the fine grain materials are stable against cracking in high temperature quantitatively. And low speed and low water cooling in the mold are also very effective.

6 References

[1] B. Hannart, F. Cialti, and R. Van Schalkwijk, Light Metals (1994) , 879[2] J. M. Drezet and M. Plata, Light Metals (1995), 941–950[3] M. G. Chu and A. Giron, Proceedings of the M.C.Flemings Symposium on Solidification

Processing and Material Processing, TMS, (2000), 223[4] M. G. Chu and D. A. Granger, Solidificantion Processing 1997, University of Sheffield,

UK, July 7-10, (1997),198

36

Investigation of Factors Affecting the Extent of Microporosity in an Aluminum Casting

O. Savas, R.KayikciSakarya University, Sakarya, Turkey

1 Introduction

Micro porosity has detrimental effects on the tensile strength, ductility and pressure tightness ofcast aluminum parts. Therefore, it needs to be predicted and controlled to minimize a level thatdoes not hinder the required performance of the cast parts. The prediction and control of micro-porosity, particularly in the design stage of a cast part, is an important practical interest which inturn requires knowledge of factors influencing the microporosity formation during the solidifi-cation of the alloy being cast.

Formation of microporosity in aluminum castings is generally caused by three mecha-nisms. (i) rejection of hydrogen from the growing solid into the adjacent liquids as a conse-quence of the large difference in the solubility of hydrogen between the liquid and the solidphases (gas porosity).(ii) interdendritic shrinkage during solidification without gas beingpresent (shrinkage porosity). (iii) Combination of the first two mechanisms [1]. For an efficient-ly fed aluminum casting with an adequate directional solidification the general prospect of po-rosity formation is that for a given melt hydrogen content the pore volume fraction and the poresize of the casting decrease with increased cooling rate and similarly, for a given cooling ratethe pore volume fraction and the pore size decrease with decreased hydrogen content [2–3]. ANumber of work concluded that modification of A356 alloy by strontium addition has beenfound to increase the pore volume fraction and pore size [2,3,4].

Reports from experiments on A356 and A319 alloys indicated that gas porosity after solid-ification depends not only on hydrogen content in the melt, but also on metal cleanliness [5].Porosity level increased for a similar H content when the melts are oxidized by chip addition orby stirring [5]. Liu and Samuel [6] recommended the use of ceramic filters and they claimedthat oxide films have a much more deleterious effect on the mechanical properties compared tothat expected from other inclusion.

A number of researchers have carried out experiments to correlate various castings and so-lidification parameters to predict the level and distribution of porosity in aluminum castings viacriteria functions [9,10,11,12,13]. In previously published literature the most frequently usedparameters were thermal gradient (G), cooling rate (R), local solidification time (Ts), and solid-us velocity (Vs).

In this study, effects of such parameters as mold filling, liquid filtering, cooling ratio, melthydrogen level and liquid treatments of molten alloy with Sr addition, on the final pore size anddistribution in an sand cast A360 alloy were investigated.

37

2 Experimental Procedure

2.1 Experimental Design

The main aim of the present study was to apply the Taguchi method to establish the effects fivecasting parameters on the extent of microporosity in an aluminum alloy. The five design para-meters (factors) and their levels are given in Table 1.

Table 1: Design parameters and levels.

The basic principle of the Taguchi method is to develop an understanding of the individualand combined effects of various design parameters from a minimum number of experiments.Taguchi method uses a generic signal-to-noise (S/N) ratio to quantify the present variation.There are several S/N ratios available depending on the type of characteristics, including “loweris better” (LB), “nominal is best” (NB), and “higher is better” (HB). Since the lower porosity isdesirable in a casting The S/N ratio for the LB characteristics is related to the present studywhich is given by [14];

(1)

where, n is the number of repetition in a trial under the same design conditions, yi represents themeasured values (percent microporosity), and subscript i indicates the number of design para-meters in the orthogonal array (OA) which is shown in Table 2.

Table 2: Experimental lay out and results with calculated S/N ratios for measured porosity.

Factors Level 1 Level 2

A- ModificationB- Hydrogen levelC- Solidification time D- FilteringE- Filling condition

Unmodified Low (0.07 ml/100 gmAl)High (14.0 minutes)UnfilteredLaminar flow

Modified (with 0.185 % Sr)High (0.22 ml/100 gmAl)Low (4.0 minutes)Filtered (with foam filter)Turbulent flow

Experiments Factors and levels A B C D E

Measured porosity (%)

y1 y2 y3

S/N ratio

1 1 2 1 1 2 1.44 1.72 0.79 1.32 –2.752 1 2 2 2 1 0.27 0.05 0.16 0.16 14.673 1 1 1 1 1 0.98 0.51 0.31 0.60 3.584 1 1 2 2 2 0.07 0.04 0.08 0.06 23.665 2 1 2 1 2 0.09 0.15 0.12 0.12 18.206 2 1 1 2 1 1.01 0.58 0.52 0.70 2.637 2 2 2 1 1 0.26 0.08 0.16 0.17 14.828 2 2 1 2 2 1.47 1.69 0.63 1.26 –2.57

2

1

110logn

i

Syi

N n

y

38

In the Taguchi method, a design parameter (factor) is considered to be significant if its influ-ence is large compared to the experimental error as estimated by the analysis of variance (ANO-VA) statistical method from the equations shown below [14]. If this is the case, the designparameter is a fundamental factor in determining the optimal solution to the design problem.

(2)

(3)

(4)

(5)

(6)

where, SST is the sum of squares due to total variation, N is the total number of experiments, SSA

represents the sum of squares due to factor A, KA is number of levels for factor A. Ai stands forthe sum of the total ith level of the factor A, nAi is the number of samples for ith level of factor A.T is the sum of total (S/N) ratio of the experiments, total is the degrees of freedom, Vfactor is thevariance of the factor, SSfactor represents the sum of squares of the factor and Ffactor is the F ratioof the factor.

2.2 Materials

The casting model is shown in Figure 1 which is consisted of four different section thickness in-creased consecutively to obtain higher solidification times from the bottom to the top. The topend section of the casting was designed to behave an effective feeder for the lower sections viamodeling a 3D computer simulation system. In some castings the running system was config-ured to fill the castings under turbulent flow. When necessary 10 ppi foam filters sized55 55 20 mm, were accommodated across the horizontal runner as seen in Figure 1.

Primer ingots of A360 alloy was melted in a 100 Kg electric crucible furnace. Liquid treat-ments such as degassing, regassing and Sr additions necessitating from the experimental lay outshown in Table 2 were carried out in a sequence from Experiment 1 to Experiment 8. Molten al-loy was taken from the crucible using a 10 kg ladle which was poured into the molds at 700 °C.

After cooling down the castings were sectioned and samples sized 15 15 15 mm weretaken from the locations shown in Figure 1. For the microporosity measurements the sampleswere metallographically prepared and their micrographs were taken using an optical micros-cope under 50 magnification. Micro porosity measurements were carried out from the micro-graphs using an image analysis technique and this procedure (including grinding and polishing)

22

1( )

N

TS T

S S iN N

2 2

1

A

i

K

AA

Ai TS S

n N

1totalv N

factorfactor

factor

S SV

v

factorfactor

error

F VV

39

was repeated at least three times to reduce the error arising from the photography and the mea-surement measurements.

3 Results and Discussion

Using the measured porosity values given in Table 2, the corresponding S/N response table wasderived, as shown in Table 3. According to the principles of the Taguchi method, for a givendesign factor, the present study defines high influence on porosity as a maximum S/N ratio. Themean S/N values of each factor level of design parameters are shown in Table 3. Therefore,Table 3 indicates the optimal design parameters combination and the corresponding value ofeach factor (see Table 1), i.e. A1: unmodified alloy, B1: low hydrogen level, C2: low localsolidification time (fast cooling rate), D2: filtered liquid, and E2: use turbulent filling condition.

Table 3: Mean S/N values of factor (factor response).

ANOVA analysis was performed using Equations (2–6) and the resulting data is summarizedin Table 4. The high contribution and variance of factors B and C indicate that hydrogen levelof the liquid alloy and the local solidification of the casting (cooling velocity) are significant.

Factors Level 1 Level 2ABCDE

9.7912.020.238.468.93

8.276.0417.849.609.13

Figure 1: Configuration of casting model

40

Meanwhile, the Ffactor ratio indicates that factors B and C are of 99 % confidence. On the otherhand, the results reveal that design factors marked with (*) namely, factor A, modificationtreatment (i.e Sr additions to the liquid alloy), factor D, filtering the melt during mold filling,and factor E, laminar or turbulent filling conditions have a less significant influence upon themicroporosity formation in the castings.

Table 4: Analysis of variance.

* pooling, ep# At least 99% confidence

According to Taguchi techniques to determine the optimal conditions, and to compare theresult with the expected conditions, it is necessary to perform a confirmation experiment. If thegenerated design fails to meet the specified requirement, the process must be reiterated using anew system until the required criteria are satisfied. In the present study, the requiredconfirmation experiment should satisfy the optimum design parameter (factor) combination as:A1 B1 C2 D2 E2. Since this combination of design parameter had already been included in themain experimental layout (see Table 2, as Experiment 4) there was no need to carry out an extraconfirmation experiment. For confidence the calculated S/N value for Experiment 4 should bebetween 19.18 and 25.33. The S/N value of Experiment 4 (see Table 2) which is the highestvalue indicating the S/N value (23.66) for the lowest porosity (0.06 % mean) within the wholeexperiments validating the confidence of the present study.

4 Conclusions

The present study has applied the Taguchi method to investigate the factors affecting the micro-porosity formation in an A360 aluminum alloys. The conclusion of this study can be summari-zed as follows.1. An A360 aluminum alloy was sand cast under five different design parameters with two level

variations to evaluate microporosity formation within the cast part sections. Results showed that the microporosity is due to the dissolved hydrogen and ranging between 0.04 % and 1.72 % within the castings.

2. The Taguchi design method revealed that the local solidification time of a casting (i.e the cooling rate) was the most influential factor upon the porosity formation. Dissolved hydro-

Factors Sum of squares (SS) Degrees of free-dom ( )

Variance (V) Ffactor P ( %)

A* (Modification)B (Hydrogen level)C (Solidification time)D* (Filtering)E* (Filling condition)e Totalep (error)

4.6371.39620.222.580.098.18707.0915.48

11111275

4.6371.39620.222.580.094.09101.013.10

23.06#

200.35#

Pooled1089PooledPooled1100

41

gen level of the original liquid alloy has an also significant effect on the extent of microporo-sity.

3. The ANOVA results indicated that the addition of Sr into liquid alloy for modification of eutectic silicon, the pouring velocity and the filtering of the liquid alloy in the running system have a less significant effect upon the porosity level of the castings.

5 Acknowledgements

The authors thank to Finitesolutions Inc.,USA for provision of Flowcast® mold filling simulati-on and Erdöküm ve Makina Sanayi A.., Istanbul, Turkey, for providing materials during the ca-sting experiments in the promises of the company.

6 References

[1] J.P.Anson, J.E Gruzleski, AFS Transactions, Vol.99, pp 135–142 (1991)[2] Q.T.Fang, D.A.Granger, AFS Transactions, Vol.87, pp 989–1000, 1989[3] E.N.Pan, H.S.Chiou, G.J.Liao, AFS Transactions, Vol.99, pp 605–621, 1991[4] R.Fuoco, H.Goldenstein,J.E.Gruzleski, AFS Transactions, Vol.102, pp 297–306, 1994[5] G.Laslaz, P.Laty, AFS Transactions, Vol.99, pp 83–90, 1991 [6] L.Liu and F.H.Samuel, J. Materials Science, Vol.33, pp 2269–2281, 1998[7] A.M.Samuel, F.H.Samuel, AFS Transactions, Vol.100, pp 657–666, 1992[8] Y.W.Lee, E.Chang, C.F.Chieu, Metall. Trans.B, Vol.21B, pp 715–722, 1990[9] S.T.Kao, E.Chang and L.C.Chan, AFS Transactions Vol.103, pp 531–536, 1995[10] E.Niyama, T.Uchida, S.Saito, J.Cast metals,Vol6. pp 16–23, 1981[11] S.T.Kao and E.Chang, AFS Transactions, Vol.104, pp 545–549, 1996[12] F.Chiesa, J.Mammen and L.Smiley, AFS Transactions, Vol.106, pp 149–153, 1998[13] F.Chiesa, J.Mammen, AFS Transactions, Vol.107, pp 103–111, 1999[14] P.J. Ross, Taguchi Techniques for Quality Engineering, Loss Function, Orthogonal Expe-

riments, Parameter and Tolerance Design, McGraw-Hill Inc., 1988

42

Model Studies of Gas Bubbles Physical Characteristics at Inert Gas Purging into Molten Metals and Alloys

R. Stefanoiu, V. Geanta, I. Voiculescu Politehnica University of Bucharest, Bucharest

1 Introduction

The shape and dimensions of the contact area between gaseous and liquid phases represent thebest way to stirring the metallic bath and to perform the chemical reactions by the transfer pro-cesses that take place at the gas-liquid interface. The contact performances in the mentioned sy-stem depend on the interface shape: the film-type systems are characterized a continuousinterface between the two phases; a liquid phase dispersed in a fluid constitutes a drop system,while a gaseous phase dispersed in liquid is usually called bubble system.

The bubble-type systems are mainly used in the metallurgical processes. These generate fa-vorable effects due to the operational convenience, and to the dimensions of the interphase sur-face. However, the most important is their disturbing potential that derives from the bubblesintrinsic floating energy.

The study of the gas bubbles physical characteristics (size and distribution) formed at the in-jection orifice is important for establishing the flow regime that takes place at the inert gas in-jection into metallic bath. Also, knowing the gas bubbles size and distribution is determinant inorder to estimate the physico-chemical processes efficiency specific to molten metals and alloysrefining procedure.

Creating fine bubbles leads to intensifying the nonmetallic inclusions removing process byflotation, while creating great bubbles is efficient for thermal and compositional melt homoge-nization in the treatment ladle. Forming bubbles with controlled dimensions is possible throughthe use of a plug-type injection system having a non-oriented porosity for the first case and hav-ing an oriented porosity for the latter. The physical characteristics of the gas bubbles formed byinert gas injection in molten metals depend to a great extent on the injection orifices diameterand also on the injected gas flow.

This paper presents the results of the model experimental researches regarding the inert gasbubble dimension formed inside of the metal baths at the inert gas injection through three typesof injection systems. The experimental researches regarding the size and distribution of the gasbubbles formed by inert gas injection in molten metals were made in order to establish the de-pendence of these factors on the used injection system, and also on the technological parametersspecific to metallic bath refining process by inert gas injection.

2 Experimental

Due to difficulties regarding the direct research of the metallic melt refining by inert gas injec-tion processes in industrial conditions, physical and mathematical modeling, with low tempera-ture models, makes most of the experimental researches.

43

2.1 Inert Gas Injection Systems used for Experimental Researches

In the experimental researches, three types of injection systems were used, defined as:

• The injection system type 1 (I.S.1), which simulates a refractory porous plug having the wor-king surface diameter (the surface in contact with the experimental fluid) of 20 mm;

• The injection system type 2 (I.S.2), which simulates a refractory porous plate having the wor-king surface diameter of 50 mm;

• The injection system type 3 (I.S.3), which simulates a refractory ceramic plug having 18 dif-ferent capillaries with the gas flowing surface diameter of 0.5 mm / channel.

The metallic melt refining by inert gas injection is more efficient when the injection systemis placed at the bottom of the ladle, than for using an injection lance introduced from above thetreatment ladle. Consequently, in all three types of injection systems experimentally used, thegas injection was made from the bottom of the experimental vessel, with the help of a central in-jection system taking into account the vessel symmetry axis.

The injection system features for the first two types used were qualitatively and quantitative-ly determined by microscopic analysis (using the Buehler Omnimet Enterprise, 5.0 versionprocessing software). Thus, the average porosities and the average gas flowing diameter (orificediameter d0) were determined for the two mentioned types. These values are shown in table 1. Itshould be known that the qualitative and quantitative analysis of the injection systems wasmade by taking several pictures of the working surface (the surface which is in contact with theexperimental fluid), and in the subsequent mathematical modeling, and the obtained averagevalues were used.

Table 1: Features of the injection systems used in the experimental program

2.2 Experimental Set-up

The concept shown in fig. 1 was used for the experimental researches. The device is made outof a transparent cylindrical vessel (6) filled with distilled water up to the working level. At itsbottom is it placed an injection system (5), through which air with different flow rates is injec-ted, the air being provided by the compressor (1).

The determination of the gas bubbles characteristic dimensions was made by quickly photo-graphing the area of gas bubbles forming, using a digital camera (7) placed at the bubbles form-ing surface level. For the researches, the three types of injection systems previous mentionedwere used.

The injected flow rate variation was maintained between 1.94 and 8.33 Nm3/s, these valuesbeing in the interval of working flow rate variation, used in the physical modeling of metallic

Injection system Average porosity [%]

Average diameter of the injection orifice ( 10–3) [m]

Type 1 17 0,055Type 2 18 0,067Type 3 – 0,5

44

bath refining by inert gas injection process. Thus, five working gas flow rates were used, foreach of them the gas feeding being continuous. The images were taken 5 minutes after the gasfeeding had begun, in order to stabilize the working flow. For each gas flow rate and for eachinjection system type were taken ten different images.

3 Results and Discussion

In order to determine the gas bubbles diameter the images were processed with the help of CorelPHOTO-PAINT 12 graphical processing software, and the truthfulness confirmation of the ob-tained results was done using the Buehler Omnimet Enterprise, 5.0 version quantitative analysissoftware. The dimensional results obtained for the same flow rate and the same injection systemvaried in maximum ratio of ±10 %.

For the gas bubbles dimension measurement, as a simplified assumption, it was consideredthat the bubbles formed at the injection orifice level are spherical and they are deforming underthe action of the external forces after detaching from the injection orifice, this also being con-firmed by the captured images.

The diagram in fig. 2 shows the gas bubbles diameter variation, determined at the injectionorifice level, function of the gas flow rate used for the three experimental injection systems.From the diagram analysis, it results a quasi-linear variation of the bubbles diameter with the in-jection gas flow rate, for all systems.

By analyzing the data in figure 2, one can notice that the gas bubbles diameter increases withthe increase of the injected gas flow rate, no matter the used injection system and its geometri-cal features. For the same gas flow, the gas bubbles diameter increases with the increase of theinjection orifice diameter.

The greatest values for the bubbles diameter were obtained using the I.S. 3, for which the in-jection orifice diameter d0 has the greatest value.

Figure 1: Experimental device for (air – water) model study of the gas bubbles dimensions generated in gas stir-red systems: 1 – compressor; 2 – manometer; 3 – flow meter; 4 –gas feeding flexible pipe; 5 – gas injection system; 6 – experimental vessel; 7 – digital camera; 8 – lid

45

With the injection gas flow rate variation between 1.94 and 8.33 (*10–5) Nm3/s, the diameterof the bubbles formed at the injection orifice level is: 1.546–2.988 (*10–3) m for I.S. 1;1.893–3.660 (*10–3) m for I.S. 2 and 4.180–5.016 (*10–3) m for I.S. 3.

From the diagram in fig. 2 can be noticed that for S.I. 1 and S.I. 2 types, the slopes of the tworegression curves are bigger than for using the S.I. 3 type. This fact is explained by the increas-ing of the injection pressure for the same gas flow rate.

In the literature there are different equations that express the correlation between the diame-ters of the gas bubble formed at an immersed nozzle level and the technological parameters ofthe injection process.

The empirical relation, obtained from the experimental data, that expresses the bubbles di-ameter variation function of the gas flow rate and the injection orifice diameter is give by theequation:

(1)

The equation was obtained using mathematical modeling with regression analysis by pro-grammed active experiment, using the STATISTICA 6.0 software.

In order to determine the equation, a nonlinear model was used with the independent varia-bles 1 (Qg) and 2 (d0) and the dependent variable 3 (db). Using the Levenberg-Marquardt andGauss-Newton methods, the nonlinear estimation of the model coefficients was established.

The diagram in figure 3 shows the regression lines obtained by equation (1) as well as the ex-perimental values in the case of the three injection systems used in the researches. From fig. 3 itresults that the values of the diameter of the gas bubbles obtained for the three experimentaltypes are close to the regression curves traced for the equation 1.

Figure 4 shows the 3D correlation between the gas bubbles diameter, gas flow and injectionorifice diameter, variation obtained according to equation (1). From the figure it results that thediameter of the gas bubbles formed in a fluid as a consequence of gas injection varies propor-tionally with the injected gas flow rate and with the injection orifice diameter.

0.9 0.406608.38624 0.075057b gd Q d

Figure 2: Gas bubble diameter variation function of injected gas flow rate

46

4 Conclusions

At the inert gas injection into the metal bath, the gas flow rate, the injection pressure and the gasinput surface area all play an important technological role, through this parameters the particu-larly dynamic and transfer processes being controlled.

From the experimental researches, it results that the dimensions of the gas bubbles formed atthe injection point inside of the metal bath depend both on the injection orifice diameter and onthe gas flow rate.

Figure 3: Gas bubbles diameter variation function of gas flow rate and injection orifice diameter. (Comparison between experimental values and regression curves obtained for eq. 1)

Figure 4: Gas bubbles diameter variation function of gas flow rate and injection orifice diameter

47

With the help of the presented diagrams, the gas bubbles dimension value may be deter-mined for any gas flow rate and orifice diameter smaller than 0.5 ( 10–3) m. Also, using equa-tion (1) the diameter of the gas bubbles formed at gas injection through an immersed nozzlehaving the diameter between 0.05 and 0.5 mm can be obtained. At inert gas injection into themetallic melt in the treatment ladle, the interior diameter of the porous systems is between thesevalues.

5 References

[1] D. Mazumdar, R.L. Guthrie, I.S.I.J. International, Japan, vol. 35, 1, 1995, 1–20[2] M. Sano, K. Mori, Transaction I.S.I.J., vol. 20, 1980, 675–682[3] R. Stefanoiu, V. Geanta, Metalurgia International, 1, 2005, 3–9[4] L. Wang, H.G. Lee, P. Hayes, I.S.I.J. International, Japan, vol. 36, 1, 1996, 17–24[5] M. B. Goldschmit, A. H. Coppola Owen, Ironmaking and Steelmaking, vol. 28, 4, 2001,

337–341[6] K. Mori, M. Sano, T. Sato, Transaction I.S.I.J., Japan, vol. 19, 9, 1979, 553–558[7] R. Stefanoiu, PhD Thesis, Chapter 5, Bucharest, 2004

48

49

Casting Technology

50

51

Remarks about Process and Technology of Continuous Casting

H. Schliefer, A. Khoury, M. PortenNorddeutsche Affinerie AG, HamburgP. Wolber and K.H. BollerSGL, BonnW. Dürrschnabel and H.R. MüllerWieland-Werke AG UlmSt. SchneiderDeutsche Gießdraht GmbH, EmmerichW.H. Müller, M. Schwarze SMS-Meer, MönchengladbachH. Oelmann, D. Rode, R. Frankenberg KME, Osnabrück

1 Abstracts

During the last fifty years continuous casting increase in the fields of technology and processalso in the materials and “tools”. It will now given remarks about the history and the future inthis production area.

2 The First Step

The first step into development from single to continuous casting is dated in the literature in1840 and it needed nearly 100 years to become industrial practice. In the historic of the conti-nuous casting Mr. Siegfried Junghans has build on his own a pilot plant in Schwandorf and aftera lot of trials – and errors too – this new technology was born.

After World War II further developments were done by a syndicate of Wieland-Werke AG,Norddeutsche Affinerie AG and Krupp. First for copper and later the steel continuous castingwas created on Krupp side. So for the continuous casting the copper side was the technology en-trepreneur.

This was the first great step from mould to mould cast into a continuous process with all pos-itive effects in it: from the bigger piece and length (10 times and more) to better – over all –quality, from lesser heavy work to better energy consumption, from higher casting speed to bet-ter economic terms.

The important tool for the continuous casting was the heat exchanger (the mould) and includ-ed in the mould is the knowledge of • the heat transfer and the grain solidification / direction in the mould • the cooling rate and cooling zones in and under the mould • the graphite shell / layer into/on the copper jacket and the taper in the mould • the connection between graphite lining and copper jacket without air gaps • the lubrication between melt and mould • the oscillation rate and amplitude

52

This type of mould with graphite inlet – also only with “one way” graphite layer on the cop-per jacket – is also very common and in use all over the world for copper and his alloys.

This figure gives a good overview of the interaction of all parameter into this kind of mould. Generally for all copper alloys – like brass, bronze or alloying of copper with multiple ele-

ments – the knowledge about casting and/or physical parameter is very small and naturally datais hardly to find in books or databases. About a lot of alloys especially the parameter – for the

Figure 1: NA cast shop from last early 1950

Figure 2: Traditional moulds for cakes and billets with graphite shell

53

mould construction and also for the casting technology – near the solidification temperature likei.e. high temperature strength, thermal conductivity/coefficient and also heat transfer rates, areunknown. So for the same alloy different types of moulds and cooling strategy are in use duringthe developments by the casting shops.

As example: CuFe2P, as a connector alloy, is with nearly over 30.000 t/y produced in con-siderable quantities but the casting technology and the moulds are different all over the world.

Figure 3: Scheme transfer and direction of heat and grain growing

Figure 4: Dispersation of Fe in CuFe2P

54

The important parameter of this alloy is the distribution of Fe and the other physical parametersare all predetermined at the time it is cast. So the casting condition must fulfil a slab free ofcracks, shrinkage cavities (pipe, blowholes) and this uniform dispersion of Fe which is abso-lutely necessary for the further treatment.

The big improvement in this matter was done by continuously recording and analysing of allparameter during the casts, mathematical modelling of temperature, stresses – especially for thefresh solidified shell and based on this to optimise the main casting parameter like castingspeed, pouring temperature, volume flow rate of metal and cooling water, different soft second-ary cooling and taper in the mould. And after that, based on the quality of the products, optimis-ing the water quality and the cooling water circuit system, installing a mould maintenanceschedule and also improving the mould-construction and the mould material (i.e. ELBRODUR)for better durability, stiffness and stability. Looking on the control of casting speed and oscilla-tion for smaller tolerances. With the real effects of this developments are the cakes cross sectionhas increased by approx. 50 % and the reject rate has decreased by approx. 60 %. The old oddscreated a reasonable access by try and error in this market section and which new bets will do inthe future?

Today for this alloy a compromise in technical and economic terms is in use. For both wecan imagine – based on all necessarily theoretical data near the solidification point like i.e.

strength, conductivity, heat transfer coefficient, solidification and distribution to build up arealistic mathematical model – and for other alloys too – and based on this parameter and data –as we know this all – we can construct and build a new generation of mould. This new genera-tion will have a pure graphite mould, indirect inner cooling circuit, inner cooling defined areasand cooling water volumes, indirect soft and direct spray cooling areas defined by localisationand cooling water pressure, a defined cooling water quality, pressure and quantities. This graph-ite mould with inline cooling system would create the optimal result for alloys in technical asalso economic terms in any parameter, even if it is more expensive.

Figure 5: Graphite mould (cakes) and inline cooling detail

55

From the beginning this mould was and still is only used for semi-continuous casting due tothe common sense, that this mould is too “fragile” to operate continuously. Nowadays thismould is established in the full continuously production for cakes. Billets are not so commonbecause of dimension intolerances created be regular maintenance. For cakes this mould isworking with very good results for surface and inner quality like shrinks and holes, for excellentcasting speed, for long lifetime of the graphite blocks and inner and outer cooling devices, andmore than three cooling zones with different cooling water volumes and velocity.

This means different heat transfer rates can be installed in the same mould. Any cross sectioni.e. from 130 400 to 1130 350 mm is available. In other words, in the future the progress ofthis mould construction includes the best flexibility and success for the development of copperand alloy casting and additionally the lowest mould cost/t (i.e. ETP, DHP, OF, CuAg, CuFe2P).

This type of mould can also be used in horizontal casting and especially for alloys (i.e. CuSnand CuNi) based on sophisticated heat transfer conditions, on complicated solidification andfirst shell building, on the flexibility in the inside construction for the different heat transfer are-as and localisations from the sidedams to upper- and bottom-side of the mould.

For all this moulds – without connection with the furnace – in the following small remarksabout „tools“ are given. A lot of inventions were done and plenty of creative people haveworked on the development of tools:

Casting without soot means a „white casting shop and people“. The protection of the opensurface in the mould is necessary for lubrication and also helping against oxidation. Protectionby soot or lamp black is very common in the copper industry with all the risk of benzpyrene inthe soot and also the contamination of the casting shop and workers – „going black“. One tech-nology – invented by HITACHI and now promoted by SMS-MEER – is working with a gasmixture, which combines the surface protection with the lubrication between mould and strand.The change from gas to soot is done by temperature and in the presence of molten copper.

This originally HITACHI technology works very well with the result of clean plants andworkers and additionally higher casting speed, i.e. in the HALCOR plant in Greece.

Figure 6: Mould under protective gas

56

The simplification from the included gas generator to a simple gas mixture station will bringthis technology to better economic scale. The combination with a level sensor by ionisation asshown in the picture brings together gas inlet, level control, surface protection and lubricationwith only one tool. This leads to simplification and zero contamination.

Graphite grades are also tools for vertical and horizontal casting. For each alloy differentgraphite specifications are necessary for porosity, density, metal penetration, thermal shock re-

Figure 7: „Cracking“ of gas to soot on the cold borderline

Figure 8: Mould level sensor by ionisation

57

sistance and elongation, conductivity (heat and/or electric), physical parameter like hardnessand strength.

This data sheet shows a rough classification. Horizontal (HC) casting needs more porosity bydegassing than vertical casting (VC). A balance of fine /rough porosity approx. 10 % for CuSnand approx. 5% for brass is common. Penetration is not a big issue for VC by the borderline lu-brication/protection of soot. All physical parameter have to be better in HC. Conductivity de-pends on mould construction and alloy, i.e. lesser than 100 (W/m K) for copper and up to 150(W/m K) for alloys. Depending on the different mould construction and casting parameter it isnot possible today to define general graphite grades for copper and this alloys.

This work can and has to be done only by the casting people themselves and cannot be re-duced to the answer: which graphite grade do you need ? Black graphite!

This means for the future going into a real open minded teamwork to minimise the numberof grades and also a big part of production cost in consideration of the competition on the glo-bal market.

Booster and stirrer have a long tradition in using ultra sonic devices to create smaller grains,improve the distribution of alloys, minimise the porosity and optimise other parameters too.Coupling these ultrasonic waves to the melt is the most difficult part of this technology. A sim-ple way is to bring US in the solid strand, but the mass of metal needs very high energy to haveany effect in the melt. Another technology is to bring the US by booster directly into the meltfrom the top of the mould. The advantage of this technology is the direct connection with themelt and the grains. The disadvantage is the technology itself: The temperature limitation of thebooster material in the melt and the cooling of the US unit above the melt as well as the connec-tion to the booster. Trials were made now and then with common boosters i.e. delivered byKrautkrämer. Metallurgical effects were found in form of very fine grains as long as the boosterwas working, which means the temperature was over 1,100°C in the melt on one end and nearlyroom temperature on the other end. Any US is disturbed and/or deleted by temperature-differ-

Figure 9: Grade comparison continuous casting

58

ences in the booster material. By solving the above material/temperature problems for the directbooster input in the melt chances for the future will increase.

The electromagnetic stirrer is very common in the steel industry and works very well in thelong liquid centre part of the strand after the mould. In the copper industry – with comparativelyshort liquid sump – this technology does not create any kind of process/quality optimisation.Calculations, constructions and trials were done at the RWTH Aachen by Prof. Dr. Block withthe result, that reasonable effects were found by a special constructed stirrer in the copper. Suc-cess in quality, distribution and grain size by this special designed stirrer was only achieved fora defined cross section and due to this reason this technology was “killed” by the economicterms. The use of the stirrer as a tool for copper alloys is an option for the future.

3 The Next Step

The next step – really one and half – was coming up with the Southwire/Contirod/Properzi -pro-cess to produce wire rod directly. This is a continuous casting of a cross section – like a wirebar– by a casting wheel or double belt caster. After casting in one line the direct rolling process inone heat produced wire – near net shape by rolling. The same way, but by another technologywas done by the DIP FORMING technology, which is for the casting part like a “grandfather”of the UPCAST process.

For the wire production copper and aluminium, and partial steel too, the before mentionedtechnology combines the biggest production share in the world wire market.

And which advantages are to be expected in the future by this technology, to fulfil this grow-ing market with a yearly rate near 5 % without investments in the existing plants and minimis-ing the production cost by volume?

First of all within the next 3–4 years higher capacity/plant will be coming up to approx.70 t/h, this means approx. 500.000 t/y by bigger melting/holding furnaces, casting wheel and/or

Figure 10: Scheme of Southwire plant NA

59

bigger cast cross section. This includes step by step more rolling speed/power and/or one or twoadditional rolling units and also high speed clean up/cooling and coiling equipment. This cre-ates new high speed inline test sensors. The coil weight and diameter (maximum truck width)will also increase to 15–22 t/coil – one truck load –, to minimise the transportation and the off-time costs in the cable plants. For market niches smaller (this is a real speed problem) and big-ger diameter (this is a coiling problem) and a production of alloys (i.e. CuAg) in big lots in onlyo n e plant but for different companies is cost efficient and earns together the volume-effects.

Figure 11: Scheme of Contirod plant

Figure 12: Calibration of flat strip

60

The copper flat strip-market worldwide has, including width nearly 100 mm, a market shareover 60 %, i.e. electric and/or ACR / welded tube indicate the next and second step by the pro-duction of flat material in copper wire rod plants.

This can be done with smaller investments of designed rolling units, clean up and additionalcoiler/winder equipments in each wire rod plant. So the production can split in wire and strip inthe same plant that means bigger volume in one unit and two different markets and that will cre-ate better economic terms.

Figure 13: Additional rolling equipment

Figure 14: Additional cooling and winder equipment

61

4 The Second Step

The second step in the development of continuous casting was the “near net shape” castingtechnology included upcast for metal wire and horizontal/vertical thin strip/sheet casting of cop-per/copper alloys and the “Thin slab” casting in the steel industry. But on the other hand a lot ofalloys need a rolling treatment which means "thin” slab. All this before mentioned technologieshave not the biggest advantages in the future:

The vertical “Thin Slab” steel technology works only for a limited amount of grades andneeds a big investment, space and additional heat treatment before rolling near net shape.

The copper VC, its alloy strips and wire technology has a limited potential for better produc-tivity and greater width by the great solidification range and the difficult shell building in amoving mould and/or strand.

The horizontal copper alloy production is the next target. Based on the better mould andgraphite, using pure graphite moulds with inline cooling system instead of the traditional hori-zontal moulds will increase efficiency.

The next big step was what we know as “near net shape casting”: Which is defined as „direct dimension, grade and final surface by one step cast process = one step production“.The single belt process for brass and bronzes is a special example and technology in this ar-

ea. Any kind of copper alloys for sheet and strip production by the traditional way: Verticalcasting cakes – preheating – hot rolling with milling and cold rolling incl. heat treatments in-cludes a higher production than the horizontal casting of strip, milling and rolling including heattreatments which includes a lower production rate but lesser invest and working steps.

This technology was first developed in the steel industry. A pilot plant was later used for testruns for different copper alloys by Wieland-Werke AG. Tin bronzes have to be during solidifi-cation absolutely stress free. The lower part of a horizontal HAZELETT machine can fulfil thisparameter. It was calculated, expected and realized by a single belt to have a high production

Figure 15: Single belt casting machine, scheme

62

rate, approx. 10 to 20 t/h and strip width from 400 to 800 mm with a thickness of approx. 4 mm.The pilot plant was working with 10 to 20 t/h on basis of 400 mm 6 to 10 mm cross sectionfor different copper alloys. Cooling is done indirectly by the under belt. The upper surface wasinsulated to maintain a directed solidification. The planar level of the strip is not so importantbecause of the following rolling of the strip. This technology is working reproducible and wellfor copper and micro-alloyed copper. For all copper alloys with dendritic solidification thistechnology does not work. The single belt technology has also economic limits by the smallvolumes of copper alloys in each cast/plant, which means changing the alloy to often.

For the strip production in the steel and copper industry the future is vertical TWIN-ROLL,in the aluminium industry the horizontal TWIN-ROLL and/or the TWIN BELT HAZELETT-caster.

The vertical TWIN-ROLL caster includes very high production rate from 40 to 70 t/h, nearnet shape casting with cross section from 1,200 12 mm2 to 2,000 0,7 mm2, with good sur-face quality. The low roughness level does not need additional milling. Now the TWIN ROLLis “state of the art” in the steel industry and also in the aluminium. Today the biggest success inthe flat product market is the vertical/horizontal “TWIN ROLL”.

The TWIN BELT caster for sheets is working in aluminium recycling plants.The TWIN-ROLL is a very old technology invented by Bessemer and is also a good and typ-

ical story for patents: it was patented but never producing!These Figures include the technology and process in the future of the production in the flat

market. The numbers and terms in the “flat” steel industry tell of the big success and increasingvolume. The copper industry has TWIN ROLL today only as pilot plant, which is a small in-dustry in comparison to steel and aluminium. The result of this pilot plant tells the TWINROLL- process is really the coming technology in the flat market. Step by step, from pure cop-per to copper alloys and all in big volumes, depending on the possible investments, can be donefor this technology by one company or by one special R+D company with different shareholders

Figure 16: Single belt casting in operation

63

from the industry. This process will create very good economic terms and excellent quality andsurface – as we have seen in the steel industry – and will also create a big chance/change for thecopper industry.

On the way into the future to invest in copper TWIN-ROLL-technology – R+D included – isa small investment in comparison to the steel but a big step for copper. High alloyed copper like

Figure 17: Bessemer patent

Figure 18: Scheme TWIN ROLL vertical NA copper pilot plant

64

brass or bronzes will have the traditional way by continuous casting vertical as cakes or hori-zontal as thin slab or thicker sheet.

Additional remarks about tools in this technologies: Ceramic materials and valves: The ceramic material for flow control valve – especially in the

copper industry – has to fit a long time corrosion and also thermo-shock resistance combinedwith zero penetration by the molten metal and high mechanical strength. This is the imagination

Figure 19: Scheme Castrip® Process

Figure 20: Scheme TWIN ROLL horizontal Alumina plant

65

of a non existing material. Materials with high contents of Al-Oxide, SiC, SiN or high tempera-ture metal-oxides mixtures will fit the most above parameter from the material side, dependingon metal and/or alloy. The other side is that all valves are working only mechanically in metalindustries. Years ago it was described by INTERATOM in patent DE24970 as a channel to con-trol and pump.

Figure 21: Parameters NEAR NET SHEET

Figure 22: Castrip Plant Production

66

This technology is very well working as a pump in a closed sodium-cooling-circuit and wasalso good working in a pilot plant as valve for vertical casting copper. The flow rate variationwas approx. 1:9, which means nearly 4 to 40 t/h – other volumes are possible by construction –in a flat channel of 5 80 mm and a cuprostatic pressure of approx. 300 mm – higher pressure ispossible and tested too. The energy consumption was calculated and recorded lesser than 3 t/hand an investment of approx. 100.000 €/valve each additional valve for the same casting unit

Figure 23: Average UCS Thickness

Figure 24: Scheme electromagnetic pump/valve

67

approx. 15.000 to 25.000 €. This kind of technology in combination with high temperature ma-terials can vary the flow rate by joystick and/or automatic control very exactly – drop by drop –and continuously into the full variation rate. Like cast (drive) by wire without mechanical parts– especially for vertical cast – the E-valve will block every pressure coming from the ladle/tundish. For the horizontal process this valve will give a defined metal speed = weight regula-tion also.

Another tool is the level control by ionisation, very useful in protected and very small areas.By this limitation and in this areas laser, electromagnetic fields or other common sensors areuseless. A good example for this kind of level control is the soot free technology with a “blank”metal surface and the protective area.

The combination of the gas input and the sensor creates an excellent signal of the impedanceby the ionised halo from the out burning gas. This signal can be used for the electronic and elec-tric devices to control the level surface of the melt in the mould.

All this tools combined in the near net shape casting technology for flat products can giveany ideas about the future: The combination of electromagnetic valve with ionisation level con-trols blank metal surface sensor with high speed valve, drop by drop controlling without metalpressure and can bring the continuous casting technology to progress.

5 The Third Step

And what is the next and third step in and for continuous casting? The mathematic modelling is on a good way to become a common tool. The development of

graphite and mould construction as “tools” is a permanent process. The simplification of mate-rials will go on due to cost pressure. The international and national bench marks will bring the

Figure 25: Electromagnetic valve in operation

68

industry in the situation to work better and closer together and to develop tools like booster, stir-rer, magnetic valve, ionized level control, graphite and ceramic grades.

In my opinion the next step has to be the mathematical modelling of alloys on the basis ofMedema and/or Hückel/Debye and the quantenmechanical theory of metals.

It has to be a vision for the future, that it is possible to define and design a real material by hisparameter and physical data. Beforehand the big job is to collect and search the parameter anddata for different temperatures. Based on this to go in a metal databank, using many differentmathematical models for the creating – first in virtual – of a new metal alloy with exorbitant ad-vantages and second out of this knowledge the real alloy can go in production. The hardware forthis third step is present, the software has to be done by brainstorming in the next near future,for the target “model of metal”. This is the third but not the last step of continuous casting proc-ess and technology.

6 Conclusion

Near net shape continuous casting, vertical TWIN-ROLL, sophisticated graphite moulds andnew combined “tools” are one step in the future also the cooperation between companies as thesecond step to reduce the economic terns

The great step in the next years are the “model of metal” by the interaction between metallur-gist and mathematicians to build up databases and models for better material and technologyand also for lesser consumption of our worldwide sources.

7 References

[1] Handbook continuous casting, 1980 [2] div. Patents

UK, 2397, 1861 ; BessemerUS, 2 935 251 ; US, 2 946 100 ; US, 3 098 269 ; US, 3 157 921 ; ASARCOCH, 362 800 ; CH, 361 888 ; Wieland Werke AGDE, 10 2004 027 194.1 ; Norddeutsche Affinerie AGDE, 24970 ; INTERATOMJP, 5 32 96 00; HITACHI

[3] P. Wolber and K.H. Boller, SGL and E. Kress, Sundwiger Messingwerke: Graphite, Per-sonnel information

[4] Dr. W. Dürrschnabel and Dr. H.R Müller, Wieland-Werke AG: single belt casting, Per-sonnel information

[5] Dr. A. Khoury, M. Porten, Dr. St. Schneider, Deutsche Gießdraht GmbH: wirerod South-wire, Personnel information

[6] Dr. W.H. Müller, SMS-MEER: soot free, Personnel information[7] H. Oelmann, Dr. D. Rode, R. Frankenberg, KME: CuFe2P, Personnel information [8] Dr. M. Schwarze, SMS-MEER: contirod, “Rolling of copper flat sections with a contirod

...”and” 25 years contirod, developments, trends, new product mix” [9] R.L. Wechsler, Castrip: report at 5. June, 2005 in New York: TWIN ROLL steel [10] NOVALIS: Reference list JUMBO 3CM/3C, 2005

69

[11] NOVALIS/PECHINEY: JUMBO 3C & 3CM continuous casting technology [12] NOVALIS: Newsletter Februar 2005[13] STAHL UND EISEN: 1891 N°11 P929 ... [14] Pictures:

Norddeutsche Affinerie AG Wieland-Werke AG SGL KME CastripNOVALIS/Pechiney SMS-Meer

70

Continuous Strip Casting of Magnesium Alloy by a Horizontal Twin Roll Caster

H.Watari 1, T. Haga 2, N. Koga 3, K. Davey 4

1Oyama National College of Technology, Oyama, Japan2Osaka Institute of Technology, Osaka, Japan3Nippon Institute of Technology, Saitama, Japan4The University of Manchester, Manchester, UK

1 Abstract

This paper is concerned with the development of a strip-casting technology for manufacturingmagnesium alloy sheets. The aim of the work is to establish a manufacturing process and tech-nology to facilitate the economical manufacture of high-quality magnesium sheet alloys. Ma-gnesium alloy AZ31B was used to investigate the appropriate manufacturing conditions for usein twin-roll strip casting. Temperatures of the molten materials and roll speeds were varied tofind the appropriate manufacturing conditions. The effects of manufacturing conditions on pos-sible forming were clarified in terms of roll speeds and roll gaps between upper and lower rolls.In addition, microscopic observation of the microstructure of the finished casting was perfor-med. It was determined that a magnesium sheet of 2.5 to 3.5mm thickness could be produced ata speed of 20m/min by a horizontal copper roll caster. Mill stiffness and a method of predictingthe cast sheet’s thickness were investigated to determine the appropriate manufacturing conditi-ons. It was also found that the cast magnesium sheet manufactured by roll-strip casting could beused for plastic forming if the appropriate wrought magnesium sheets were produced after theroll casting process.

2 Introduction

Magnesium alloys are expected to play an important role as next-generation materials, with thepotential to help lighten total product weight when magnesium products are used to replace alu-minum and mild steel products. The specific density of magnesium alloy is 2/3 that of alumi-num and 1/4 that of iron. When alloyed, magnesium has the highest strength-to-weight ratio ofall structural metals. Moreover, magnesium has received global attention from the standpoint ofenvironmental preservation because of the ease of recycling metallic materials. The utilizationof magnesium alloys has depended mainly on casting technology (e.g., thixo-forming) becauseof their less workable characteristics due to the crystal structure of the hexagonal close-packedlattice. Recently, demands have arisen in the automotive and electronics industries to reduce thetotal product weight [1]. Unfortunately, high manufacturing costs continue to be a major barrierto greatly increased magnesium alloy use. A key to solving this problem is the development ofroll strip casting technology to manufacture magnesium sheet alloys economically while main-taining high quality.

71

The authors, therefore, investigated the effectiveness of twin-roll strip casting for magnesi-um alloys [2]. This paper describes the forming characteristics of the cast magnesium alloysheets after being hot-rolled in a warm deep drawing test and establishes the appropriate manu-facturing conditions for producing high-quality strip using a purpose-built strip-casting mill.The influences of such process parameters as materials of roll, casting temperature, and rollspeed are ascertained. A simple method of predicting the sheet thickness of cast strip is intro-duced. A warm deep drawing test of the cast magnesium sheets after being hot rolled was per-formed to demonstrate the formability of the magnesium alloy sheets produced by a roll stripcasting process. The microstructure of the manufactured wrought alloy sheets was microscopi-cally observed to investigate the effects of the hot rolling and heat treatment conditions on crys-tal growth in the cast products.

3 Experimental

3.1 Horizontal Twin-RollCcaster and Experimental Conditions

Figure 1 illustrates the horizontal twin roll strip casting process used in the experiment. A sour-ce of molten metal feeds into the space between a pair of counter-rotating internally cooledrolls. The principle dimensions of the horizontal twin roll caster are presented in Table 1. Theinclination angle of the mill in Figure 1 was set to zero degrees. Lc in Figure 1 indicates the con-tact length between rolls and the molten metal. Table 2 presents the experimental conditions forinvestigating appropriate manufacturing conditions to successfully produce magnesium alloysheets by twin roll strip casting. Casting temperatures were varied from 630 °C to 670 °C to findthe best casting conditions, as indicated in Table 2. Temperatures of the molten magnesium inthe melting pot and tundish were measured by thermo-couples.

Figure 1: Schematic illustration of horizontal twin roll casting process

72

Roll casting speeds were varied from 5 m/min to 30 m/min in order to examine which rollspeed was appropriate for solidifying the molten magnesium. The roll gap between the upperand lower rolls was determined to be from 2.0 mm to 3.6 mm by simple calculation resultsbased on basic solidification theory. No shielding gases were used in the experiment.

Table 1: Dimensions of roll caster and tundish

Table 2: Experimental conditions

3.2 Material and its Refining Process

The material used in the experiment was AZ31B. The physical properties of the material are li-sted in Table 3. Magnesium ingots were heated to 680 °C in a melting pot with an electric fur-nace. In the magnesium melting process, magnesium oxide and other suspended nonmetallicmatter were removed with flux that preferentially wet the impurities and carried them to the bot-tom as sludge. After the refining process, the molten magnesium metal in the melting pot wascarried to the strip caster and poured onto the cooling slope to manufacture magnesium strip.

Table 3: Physical properties of material

3.3 Hot Rolling Process after Twin Roll Strip Casting

The hot-rolling process was performed to obtain wrought magnesium alloy sheets with globularand fine microstructures to be used for plastic forming. The cast strip sheets were milled to ob-tain sheets with 2.0 mm thickness to remove oxide film. The cast strip was heated and rolled inthe hot-rolling process. Rolling temperatures were varied from 200 °C to 300 °C. The milledsheet was rolled by several rolling pass schedules until the sheet became 0.8 mm thick. Next,

Rolls

Materials Copper, Copper alloy, Mild steelUpper roll (mm) 300*150Lower roll (mm) 300*150Roll speed (m/min) 0-150m/min (Max.)Inclination angle (deg ) 0TundishMaterial Insulator, mild steel

Temperatures (°C) 630, 650, 670Roll speeds (m/min) 5, 10, 15, 20, 25, 30Roll clearance (mm) 2.0-3.6Shield gas No shielding

Density (kg/m3 103) 1.78Liquidus temperature (°C) 630Solidus temperature (°C) 575Specific heat (J/kg°C) 1040Thermalconductivity (W/m·°C) 96Latent heat (kJ/kg) 373

73

the 0.8 mm-thick sheet was rolled again until the sheet became 0.5mm thick. Finally, the rolledmagnesium sheet was annealed at 350 °C for two hours, and cooled in an electric furnace. A5m/min roll speed was chosen in the hot-rolling process. At 250 °C, cracks were seen during thehot-rolling process, even though the reduction was less than 10 %. A temperature over 250 °Cwas chosen to keep the cast products from cracking.

4 Results and Discussion

4.1 Stiffness of Roll Caster, Relation Between Roll Load and Roll Gap

Figure 2 illustrates the typical relation between roll gap and rolling load obtained by load cellsattached to the mill. The result presented in Figure 2 is the case of a mill fitted with copper rolls.The horizontal axis reveals the differences between initial roll gaps and the obtained sheetthickness. It expresses pure deflection of the mill during the strip-casting process. The circles inthe figure represent results for an initial roll gap of 2.0 mm, the squares for an initial roll gap of2.8mm, and the triangles for an initial roll gap of 3.6 mm. From Figure 2, the spring constant ofthe mill was approximated as 10009 (N/mm). Roll loads per unit width of the cast sheets duringcasting were from 50 to 150 (N/mm) in the present experiment.

4.2 Thickness of Cast Sheet

Thicknesses of cast sheets were measured to investigate the casting phenomenon of magnesiumalloy in the twin roll casting process. The sheet thicknesses of three forming directions weremeasured. Figure 3 presents an example of results for a copper roll caster. The circles indicate

Figure 2: Relation between roll loads and roll gap

74

results for a roll gap of 2.0 mm, the squares for a roll gap of 2.8 mm, and the triangles for a rollgap of 3.6 mm. It is seen that sheet thickness gradually decreases as roll speed increases. Thedotted line represents a theoretical sheet thickness obtained by simple 1-D solidification mode-ling for a 2.0 mm roll gap. Figure 3 indicates that a roll speed of over 20 m/min is necessary tomanufacture strips less than 2 mm thick using a 2.0 mm roll gap.

4.3 Microstructure of Cast Sheet

Figures 4(a), 4(b), present photographs of the microstructures of hot-rolled sheets after the roll-strip-casting process using a copper-alloy caster. The sheet depicted in Figure 4(a) was hot rol-

Figure 3: Relation between roll speed and sheet thickness tween roll loads and roll gap

Fig. 4(a): Microstructure of cast sheet hot rolled at 200°C (without annealing process)

Fig. 4(b): Microstructure of cast sheet hot rolled at 200°C (with annealing process)

75

led at 200 °C, and no annealing process was used. The photo in Figure 4(b) was annealed at350 °C for two hours after hot rolling at 200 °C. We can see that the crystals were well homoge-nized with an appropriate annealing process, although the grain sizes of the crystals became lar-ger by recrystallization.

4.4 Plastic formability of Obtained Wrought Magnesium Alloy Sheet

After the cast magnesium sheets were hot rolled, a warm deep-drawing test was performed toexamine the forming characteristics of the magnesium alloy sheets produced by twin roll stripcasting. The diameter of the punch was 28.8 mm. A lubricant solution was used. The limitingdrawing ratio was investigated by a deep-drawing test at 250 °C. A drawing speed of 30 mm/swas chosen in the test. A limiting drawing ratio of 2.6 was obtained in the warm deep-drawingtest, as indicated in Figure 5. The result presented in Figure 5 suggests that the wrought magne-sium alloy sheets that were hot rolled after the strip casting process had plastic formabilityequivalent to that of the wrought magnesium alloy sheets manufactured by the conventional DCcasting process.

5 Conclusions

AZ31B magnesium alloy was cast by using a horizontal twin-roll caster. The obtained castsheets were hot rolled, and a warm deep drawing test was performed to demonstrate the effecti-veness of twin-roll strip casting of magnesium alloys. The following conclusions were obtained.

1. In the hot-rolling process, a temperature exceeding 250 °C was chosen to keep cast products from cracking.

2. An appropriate annealing temperature was effective for homogenizing the microstructure of the rolled cast sheets after the strip casting process.

Figure 5: Cup drawn in a warm deep drawing test (DR=2.6, hot rolled at 250°C, with annealing)

76

3. The grain size of the manufactured wrought magnesium alloys sheet was less than 10 micro-meters. The obtained magnesium alloy sheet exhibited an equivalent limiting drawing ratio in a warm-drawing test.

6 References

[1] S. Yoshihara, H. Yamamoto, K. Manabe and H. Nishimura, Journal of Materials Proces-sing Technology, Vol. 143-144, 2003, 612.

[2] F. Moll, M. Mekkaoui, S. Schumann and H. Friedrich, Proc. of the 6th Int. Conf. Magne-sium Alloys and their Appln., DGM 2003, 2003, p936.

[3] H. Watari, K. Davey, M. T. Alonso Rasgado, T. Haga, Journal of Materials Processing Technology, Vol 155-156, 2004, 1662.

[4] H. Watari, R., Paisarn, N. Koga, T. Haga, K. Davey, M. T. Alonso Rasgado, Key Engi-neering Materials, Vols. 274-276, 2004, 379.

77

Strip Casting of Mg-Al based Alloy with Ca by Twin Roller Caster

K.Matsuzkai, K.Hatsushikano, Y.Torisaka, K.Hanada and T.ShimizuNational Institute of Advanced Industrial Science and Technology(AIST), Tsukuba, Japan

1 Abstract

AZ61 with 0.25wt%Ca strips were directly prepared from molten by twin roller caster. The ad-dition of Ca enable to perform the strip casting process of Mg alloys in air without cover gas.This is expected to simplify process. The cast strip is composed of fine equiaxed grains with agrain size of 20 m for the thickness of 1.0 mm. The mechanical properties increase with decre-asing thickness, and the yield stress and ultimate tensile strength reach 160 MPa and 240 MPa,respectively, for the strip with 1 mm thickness. This is due to the grain refinement introduced bystrip casting. A further improvement of mechanical properties was achieved by thermomechani-cal treatment. It is thought that the strip casting process is useful for production of Mg alloysheets and the Mg alloy with a small amount of Ca is suitable for strip casting.

2 Introduction

Mg alloy is the lightest metallic constructional material with a density of 1.78 Mg/m3 and havean attractive possibility for weight saving the vehicle. At present time, most of Mg componentsare produced by using Die Cast process because of its good castability. On the other hand, Mgwrought alloys have superior mechanical properties compared to cast alloys. In conventionalMg sheet production process, a thin sheets is produced through many steps, because of poorformability of Mg due to hcp structure . This might lead to an increase in the cost of Mg wrou-ght alloys. The strip casting process makes it possible to directly obtain the sheets with thick-ness less than one or two mm, leading to the reduction in cost and energy consumption.Furthermore, the strip casting have an effective process for refining the microstructure and re-ducing the segregation, leading to the improvement of mechanical properties. However, the re-port for the strip casting of Mg alloys has been limited [1–4].

This paper intends to produce Mg-Al based sheets with Ca by strip casting, and clarify themicrostructure and mechanical properties.

3 Experimental Procedure

Pure Mg, Al Zn and Ca were used in the present study. Firstly, about 100 g of Mg-Al-Zn alloywith 2.5%Ca were melted in the graphite crucible under the argon atmosphere by inductionmelting. The obtained alloy ingot was inserted into a steel crucible with appropriate Mg, Al andZn, and melted into a Mg93Al5.75Zn1Ca0.25 alloy by electric resistance furnace. The mixture of Arand SF6 was used during the melting to protect the molten from oxidation or burning. After

78

them, the molten alloy was move to the twin roll caster and poured into a tundish between rolls.The casting was performed without a protective gas. The steel rolls with a diameter of 300 mmand a width of 50 mm was used. Roll gap was in the range of 0.5 to 4 mm and roll speed was16 m/min. The casting temperature was in the range of 893 to 923 K. The cast strips with athickness of 1.5 mm were subjected to warm rolling and heat treatment. The microstructure wasexamined by an optical microscope. Hardness was measured by a micro Vickers hardness testerand tensile properties were measured by a Shimazu Autograh material testing machine.

4 Results and Discussion

Figure 1 shows the external appearance of a as-cast Mg alloy strip with a thickness of 1.0mm .No oxidation is observed on the surface of the sheet. The sheets with a thickness of 4mm arealso obtained without oxidation. In the case of AZ61 alloy, the strips with a thickness above3mm show black or brown surface due to the oxidation. Moreover, the cover gas was needed toprotect molten alloy in the tundish from burning. An addition of 0.25%Ca is effective to preventthe oxidation or burning. This protective effect of 0.25%Ca was valid for AZ31 and AZ91 al-loy.

Figure 2 shows the microstructure of center area of the strip cast Mg alloy with different thick-nesses. The strip with a thickness of 4 mm is composed of dendrite and equiaxed grain with aaverage size of 40 μm. With decreasing thickness, the dendrite region and grain size decrease.For the 1 mm, the fine equiaxed grains with an average size of 20 m are obtained. Thismicrostructure is favorable for the mechanical properties. It is reported [1–4] that strip castAZ31or AZ91 alloys show a dendrite structure. In the present study, the alloy has a narrow free-zing range and the roll speed is higher compared to that of the reported study. Therefore, the fa-vorable microstructure is obtained. Figure 3 shows the X-ray diffraction patterns of as-castMg93Al5.75Zn1Ca0.25 alloy strips. The strips consist of hcp-Mg and Mg17Al12 and no peaks corre-sponding to Al2Ca are observed. The intensity of (002) peak increases with decreasing thick-ness. This suggesting that the oriented structure with a basal plane parallel to the surface of thestrip is developedMechanical properties of as-cast strips are summarized in Table 1. The yield stress(YS) and ul-timate tensile strength (UTS) increases with decreasing thickness and reach 160 MPa and240 MPa, respectively, for 1mm thickness. There is also an increase in elongation and hardness.It is notable that there is no significant difference in mechanical properties for the strip of

Figure 1: External appearance of as-cast Mg93Al5.75Zn1Ca0.25 with a thickness of 1.0 mm

79

1.5 mm thickness between Mg93Al5.75Zn1Ca0.25 and Mg93Al6Zn1(AZ61). This suggests that theaddition of a small amount of Ca such as 0.25 % is not harmful to mechanical properties of Mgalloy. The increase in mechanical properties is due to the grain refinement, introduced by cast-ing.

Figure 2: Microstructure of as-cast Mg93Al5.75Zn1Ca0.25 with different thicknesses, (a) 4 mm, (b) 3 mm, (c) 1 mm

Figure 3: X ray diffraction patterns of as-cast Mg93Al5.75Zn1Ca0.25 strips

80

Table 1: Mechanical properties of as cast Mg alloy sheets produced by twin roller casting

However, theses value are slightly low compared to those of AZ61 sheets produced by a con-ventional sheet production process. This is maybe due to the inhomogeneous microstructure andthen the improvement of mechanical properties is expected to be achieved by a subsequent ther-momechanical treatment. The strip with a thickness of 1.5 mm annealed for 10h at 673 K showsan improvement of elongation to 12 %.The subsequent rolling at 473K to 1.0mm thickness andannealing at 673 K for 1 h causes an increase in both strength and ductility and YS, UTS andelongation reach 160 MPa, 280 MPa and 18 %, respectively. A further improvement is expectedby optimizing the condition of the thermomehcanical treatment.

5 Conclusion

Mg93Al5.75Zn1Ca0.25 strips can be produced by twin roll caster without cover gas. The addition of0.25%Ca is effective to prevent the oxidation or burning. The strip with a thickness of 1mmconsists of fine eqiaxed grain with a size of 20 m. The mechanical properties increase with de-creasing thickness and YS, UTS and elongation reach 160 MPa, 240 MPa, 10 %, respectively,for the 1mm. A further improvement is achieved by the thermomechanical treatmen. The me-chanical properties the Mg93Al5.75Zn1Ca0.25 strip is comparable to those of the alloy without Ca,indicating that a small amount of Ca have no significant influence on mechanical properties ofMg alloy. It is concluded that the strip casting process is useful for production of Mg alloysheets and the Mg alloy with a small amount of Ca is suitable for strip casting.

6 References

[1] S.S. Park,J.G.Lee,Y.S.Park and N.J.Kim, Materials Science Forum, 419–422,2003,599[2] B.S.You,C.D.Yim,B.S.Kam adnW.W.Park, Materials Science Forum, 488–489,

2005, 337[3] Y.Nakaura and K.Ohori, Materials Science Forum, 488–489,2005,419[4] C.Yang,P.Ding,D.Zhang,F.Pan, Materials Science Forum, 488–489, 427

YS (MPa) UTS (MPa) Elongation (%) Hardness (HV)

Mg93Al5.75Zn1Ca0.25 (1mm) 160 240 10 75

(1.5mm) 120 200 7 75 (3mm) 110 175 5 65 (4mm) 105 160 4 60

Mg93Al6Zn1(AZ61) (1.5mm) 120 210 8 75

81

New Strip Casting Process for Magnesium Alloys

Fr.-W. Bach, M. Hepke, A. RossbergInstitute of Materials Science (IW), University of Hanover, Germany

1 Abstract

One of the main reasons for the hesitant application of magnesium sheets is their high pricewhich is primarily caused by a lack of adequate feedstock for their production. Currently, thick,oval direct chill (DC)-billet cast slabs, which have to be milled before rolling, are used industri-ally. Therefore a main research aim at IW is to cast and deform magnesium near-net-shapestrips in one heat in order to improve the forming capacity and simultaneously cut productioncosts. For this purpose a twin-roll-stand was modified to function as a Hazelett-type caster. Twovariable casting belts, a melt-feeding device, an inert-gas plug furnace and a cooling system hadto be adapted to a twin roll stand. These components were constructed especially for castinghighly reactive magnesium melts contamination-free. Additionally, several security- and con-trol measures were installed.

First casting experiments with the standard alloy, AZ31 (MgAl3Zn1), showed the feasibilityof the new plant-concept. Minor deficiencies were remedied by plant-modifications and/orchanges of the construction materials. Currently, optimized parameter sets for several alterna-tive alloys are under examination. The cast strips produced with the new caster-type were ana-lyzed concerning their suitability to serve as rolling feedstock. The samples show a roll-ablesurface and a significantly finer microstructure than conventional cast products. The propertiesof first sheets (200 200 1 mm) made of this new feedstock have been determined.

In order to evaluate the process holistically, the quality of finished magnesium sheets and thecost saving potential of the new process chain were compared to standard production tech-niques.

2 Introduction

Due to their high mass-specific mechanical properties, magnesium sheets offer a significantweight-saving potential in modern vehicle constructions. To increase the range of applicationse.g. in the automotive industry, magnesium sheets for deep drawing purposes have to complystrict requirements regarding corrosion resistance, geometric tolerances, forming capacity,joining technology and surface quality [1]. The feasibility of automotive applications is subjectto various current research projects. For example in the BMBF-funded project “ULM – Appro-priate process chain for ultralight components made of magnesium sheet metal for transportati-on” an engine bonnet for the VW Lupo was developed as a demonstrator part. This 100%magnesium solution has a Class-A surface, shows acceptable corrosion resistance and is joinedby using various techniques [2]. The range of automotive applications implies among others:closures (e.g. door panels), seat-cups, splash-boards, roof sections, convertible roof covers, in-strument carriers, trunk panels, etc. In contrary to the above said, magnesium sheets have not

82

yet made the step into series application. Besides the natural aloofness of conservative designengineers, the financial aspects of the application of magnesium sheets are the main obstacles.

Firstly, high purity magnesium wrought alloys are at least as expensive as aluminum. Cur-rently the Chinese production is dramatically increasing which relieves this situation [5]. Sec-ondly, the continuous DC-casting process of the rolling feedstock is comparatively complex andinvolves the usage of costly cover gases [7]. Before rolling can begin, the cast surface has to bemilled, because the stationary dies cause cracks and wrinkles in the surface of the billets. Third-ly, the relatively thick slabs have to be hot-rolled due to the coarse grain sizes in the hexagonallattice. This involves several reheating steps between the numerous rolling passes. Finally theproduction of the high quality parts requires heated deep drawing, corrosion protection and join-ing (Figure 1).

The aim of this contribution is to reduce the production costs by improving the casting- androlling process. This can be achieved by applying an appropriate continuous casting principleand thus producing better deformable feedstock.

3 Plant Concept

In the strip casting process, near-net-shape feedstock can be produced very economically. Thisgroup of casting principles is characterized by an increased productivity and a high quality-le-vel. The comparatively rapid cooling causes a fine grain structure. If moving dies are applied,the surface can result smooth enough for rolling without pre-treatment. Two of the main casterprinciples shown in Figure 2 were analyzed regarding their suitability for strip-casting magnesi-um.

The twin-roll casters (Figure 2 left) consist of a nozzle (2) and two rolls (1). The melt solidi-fies directly between the cooled rolls. The orientation can be vertical or horizontal e.g., as

Figure 1: Process chain (from the pre-production to the finished part) [6]

Figure 2: Twin-roll casters, single belt caster [3]

83

shown in [4]. The deficits of this method are the direct contact of the melt and rolls (surfacecracks) and the difficult (inner) cooling of the rolling-cylinders.

The principle of single belt casters, as pursued by [8], is shown on the right image. The meltis poured on top of a cooled belt (1) and has to be covered with protective gases (2). The prob-lem of this caster, besides the gas consumption, is the asymmetric solidification and the surfacequality of the free upper side of the strip. Since magnesium shows strong anisotropy and reac-tivity, problems in these directions are predictable.

The Hazelett-caster (Figure 3) applies two belts (1), at least 4 rolls (2) and a tundish (3). Themelt enters a closed system and the moving belts transport the strip out of the solidificationzone. From the multitude of continuous casting principles, the Hazelett-type was chosen, be-cause it offers several advantages regarding the properties of magnesium melts. Firstly, it oper-ates with a closed die so that there is principally no need for using large amounts costlyprotective gases. Secondly, it functions horizontally which (in case of a failure) minimizes theamount of leakage compared to vertical (gravity) casters. Thirdly, the cooling of the belts ismuch simpler and more effective than an inside cooling of the rolls. Additionally, it is possibleto cast very thin, near-net-shape strips (min. 1 mm). The cast skin is much smoother because thedie moves along with the melt. Finally it is possible to deform the solidified strips between themain rolls which causes homogenization and grain refinement.

In order to benefit from these general advantages, the dimensions and materials of the con-struction had to be adapted precisely to the requirements for casting magnesium.

The general demands for all components in melt-contact are that they do not react with mag-nesium, that the melt does not stick to them and that they endure the occurring thermo-mechan-ical stresses. The functions which have to be fulfilled are melting under protective gas, heatingto casting temperature, dosing, bide a holding period and then continuously solidify, cool, trans-port and deform the strip. Figure 4 gives an overview of the chosen construction. It can be seenthat a plug furnace (6) was adapted above the caster. It is also possible to cast with a pressurizedoven via a dosing tube from below the caster, which may be desired for safety reasons, sincegravity forces the melt back into the system in case of failures. The melt feeder (8) changes theprofile of the melt channel from a round tube to a rectangle. In case of a blocking in the caster(early solidification) it is possible to empty the furnace with a drain plug. This part is made ofcast steel which stays bright after heating and is inert to magnesium. As lateral seals fixed steelpanels (4) were utilized. Fortunately magnesium does not stick to them, thus it was possible toseal the system without using complicated chains (so called dam blocks) which move alongwith the cast strip. The belt adjustment (9) on the one hand suits to tighten the belts. On the oth-er hand the cast- and the roll gap as well as the declination of the belts can be varied widely withthis solution.

Figure 3: Hazelett-caster [3]

84

The cooling zone is shown in detail in Figure 5. Magnesium melt enters the cast gap througha melt inlet (5). The final material choice for this part was titanium because of its low thermal

conductivity which is needed in order to minimize heat losses from the heated into the cooledzone. The casting belts are pressed onto the seals by the pressure bridge (4). It also avoids buck-ling of the belts. Two water-air spray nozzles (1) cool the thin belts and hence the cast strip.

4 Experiments / Casting Parameters

Experiments were so far conducted with the standard magnesium alloy AZ31 (MgAl3Zn1).Firstly geometrical parameters such as the cast- and roll gap and the declination were varied.Secondly the balance between early solidification (blocking of the caster) and melt leakage (Mgflows though the caster) was established by variation of the holding time, melt- and caster tem-peratures, cooling rate and casting speed. Calculations backed by former experiments served asa staring point from which optimization was conducted. Adequate results were obtained usingthe parameter set shown on Table 1.

Figure 4: Magnesium specifically modified Hazelett-caster

Figure 5: Cooling zone

85

Table 1: Casting parameters for AZ31

With these parameters is was possible to cast 1200 mm strips of AZ31 with a thickness of8,3 mm. The length was only limited by the melt volume in the furnace. Anyhow, some deficitswere registered. The strip exits the moving die with a high temperature (approx. 450 °C) , there-fore reactions with the ambient air cannot fully be avoided. A secondary spray cooling at theexit can solve this issue. Secondly, after longer operation times, magnesium tends to stick to thebelts and seals. A continuous lubrication (e.g. with boron nitride) should be adapted.

5 Results and Conclusions

The strips were analyzed regarding their surface quality and their microstructure. Table twoshows an overview of the average grain sizes in common magnesium rolling feedstock. Onlypre-deformed material, e.g. by extrusion has a finer microstructure than the thin cast strips. The-refore strip cast material can be rolled with higher strain from the beginning on. In addition tothe near-net-shape geometry, this is a large cost saving factor in contrast to billet cast feedstock.Figure 6 allows a comparison between the strips produced in the modified Hazelett caster andconventional industrial billet cast slabs. It can be seen that the strip surface (where no reactionwith air occurred) is suited to enter the rolling process without milling, whereas billet casting al-ways demands such a preparation.

In conclusion it can be said that the presented laboratory scale results show the feasibility ofthe new process chain for magnesium sheet production via strip casting. As soon as minor defi-

Range Selected

Melt temperature [°C] 650 – 750 690Feeder temperature [°C] 600 – 670 670Belt speed [m/min] 0 – 7 2,1Holding period [s] 0 – 5 1Cooling [l/min] gas / spray: 0 – 80 argon: 80

Figure 6: Comparison of surface quality and microstructure

86

cits are remedied, high quality, near-net-shape feedstock can be produced very economicallywith the newly modified Hazelett caster. First estimations predict a cost saving potential of ap-proximately 50 %, resulting mainly of less rolling stages, heating cycles and material losses, incomparison to billet cast feedstock.

Table 2: Comparison of average grain sizes for different rolling feedstock

6 References

[1] S. Schumann, H. Friedrich, Materials Science Forum, Vols. 419–422 (2003), pp. 51–56[2] P. Juchmann: Magnesium-Knetwerkstoffe für komplexe Ultraleichtbauaufgaben, Braun-

schweiger Symp. Faszination Karosserie, 2003, Seite 188–195[3] A. Stepanov, A. Neustruev, J. Zilberg, Blechherstellung aus der Schmelze, „Metallurgie“,

Moskau, 1978[4] H. Pircher, R. Kawalla, German Patent No: DE 100 52 423 C1, Verfahren zum Erzeugen

eines Magnesium-Warmbands, Patent owner: Thyssen Krupp Stahl AG, Jan. 2002[5] Fr.- W. Bach, M. Schaper, A. Kuhlmeyer, M. Schäperkötter, J. Weber, A. Rossberg:

Moderne Blechwerkstoffe für die Automobilindustrie; ein Vergleich, Berichte aus dem IWU, Band 22, 2003, Seite 9–26

[6] Fr.- W. Bach, M. Rodman, A. Roßberg: Improvements to the rolling- and deep-drawing process of magnesium wrought alloys, DGM conference Magnesium, Wolfsburg Dec. 2003, p. 285ff

[7] F. Pravdic, D. Leitlmeier, Leichtmetallzentrum Ranshofen, Vertical Direct Chill Casting of Magnesium, DGM conference Magnesium, Wolfsburg Dec. 2003, p. 675ff

[8] H. Palkowski L. Wondraczek: Herstellung von Magnesiumband mittels Single-Belt-Caster: Grundlagen und Ergebnisse, Metall, 58. Jg. Heft 12, 2004

Alloy Steel mould casting Extrusion Billet casting Strip casting

AZ31 500 μm 20 μm 300 μm 80 μm

87

Production of Twin Roll Cast AA6016 for Automotive Applications

M. Dündar1, Ö. Keles1, G. Anger2

1Assan Aluminum, stanbul, Turkey2AMAG Automotive GmbH, Ranshofen, Austria

1 Introduction

mulate development of age-hardenable aluminum alloys to meet the requirements of automoti-ve applications. A continuous increase has been observed in share of aluminum sheet productsamong all aluminum parts employed in auto bodies. Inner and outer panels are typical applicati-ons for sheet products. In this regard, 6000 series aluminum alloys are the primary candidate toachieve comparable performance of conventional material, i. e. steel, that has been employedsince the beginning of the century. The characteristic properties of aluminum; high strength andstiffness to weight ratio, good formability and corrosion resistance exist in AA6016. However,the cost of aluminum and its conversion cost remain the biggest impediment for its large scaleuse in automotive applications [1–3].

Cost effective solutions in production, recycling, or in short, effective life cycle analysis ofaluminum alloys, will bring it to a position at which it can find variety of application fields, notonly limited to automotive industry. Production of aluminum sheet by Twin Roll Casting (TRC)route rather than by conventional DC casting and hot mill method offers an opportunity to sub-stantially reduce the cost barrier, which could lead to an increase in its use. Combination of lowoperational and investment costs to the TRC with the shorter production times and flexibility inswitching from one alloy to another in casting operation stands for strong economical aspects ofthis production method. Superior micro structural features inherited to the material due to thesolidification mechanism and its good response to down stream processes are material relatedissues believed to accelerate the acceptance of twin roll cast sheet in the market [4–5].

Present study investigates the performance of twin roll cast AA6016 aiming automotive ap-plications. As-cast and processed strips were subjected to the microstructural and mechanicalcharacterization techniques to test their compliance to specific application of automotive indus-try.

2 Experimental

Industrial scale coils, having average weight of 9 tons, were produced in ASSAN Aluminum,Turkey by employing 2200 mm wide SpeedCaster® at the width of 1800 mm. Chemical compo-sition of cast strips are given in Table 1. Since this material is aimed to be used in automotiveapplications, one of the basic requirement is that to produce strip having defect free surface. So-lidification mechanism operating at the caster roll-gap, in general, is believed to be inadequateand promote segregation behavior at the roll-strip interface while producing alloys with wide

The growing demand for fuel-efficient vehicles to reduce energy consumption and emission sti-

88

solidification range. Thus, 6000 series alloys are among those that have not been produced witha consistent and reproducible quality, up to date. Ability to increase control over the casting pa-rameters is the key factor for tailoring overall micro structural features of a strip produced withTRC. As-cast strip surface quality, in terms of surface segregations and ripples has exhibitedvery good performance, as far as final product quality expectations were concerned. Down st-ream processes of the coils were carried out in the facilities of project partner, AMAG RollingGmbH, Austria. The strips were processed through cold rolling to 1 mm as well as up to 2,5mm, then solution heat treatment in a continuous heat treatment furnace, water quenching andinvestigated after aging 7 days at room temperature.

Table 1: Chemical composition of TRC AA6016

Through-thickness micro structural investigation of as-cast samples was conducted in twodirections, namely parallel and perpendicular to the casting direction. Samples were polishedwith standard metallographic techniques: ground with SiC paper, polished with 3 m diamondand finished with colloidal silica. Macro and micro structural investigations of the as-cast sam-ples were done by using a Zeiss Axiotech Vario model optical microscope. Micro structural fea-tures were revealed by etching 0,5% HF solution. Grain structure was observed with crosspolarized light after electro polishing with Barker’s solution.

Semi-quantitative analyses of micro structural constituents were conducted by using JEOL5600 SEM equipped with Oxford EDS unit. Line and dot mapping analysis were conducted fordetermining chemical content of individual constituents.

3 Results and Discussion

3.1 Micro and Macro Structure of As-Cast Strip

Important characteristic of twin roll cast alloys, is the heterogeneity of the grain structure throu-gh the thickness of the cast strip. This structural heterogeneity is generally attributed to the coo-ling rate gradient encountered in the twin roll casting process. Very high solidification rate ofinitially contacting liquid metal with caster rolls generates a supersaturated aluminum matrix atthe outermost layers of the strip. During the movement course of solidified metal in the roll gap,metal is forced to pass through the opening between two caster rolls, called indicated castinggauge. This process induces limited amount of plastic deformation at the shallow depth of thestrip. Thus, the macroscopic structure near the surface revealed the characteristic appearance ofa featureless zone at the outer fibers and gradually increasing and elongated ones at the quarterplane of the thickness (Figure 1). Obviously, alignment of the grains, especially those located atthe outer fibers, reveal different grain structures in parallel and transverse to the casting direc-tion (Figure 2 a and b)

Si Fe Cu Mn Mg Al

1,06 0,17 0,07 0,07 0,34 98,25

89

Decrease in casting gauge has pronounced effect on the proportion of plastically deformedvolume through the thickness. The angled position of the grains becomes more parallel to thecasting direction while intermetallic particles, decorating the grain boundaries, are more alignedin the same direction. Center plane of the strip is occupied by more or less equiaxed grains. Likethe other alloys produced with the same technique, centerline segregation at the centerplane ofthe strip is an unavoidable consequence of solidification mechanism of this production method[5–6]. Tubular form of individual segregates of those does not alter equiaxed grain structure ofthe center plane.

Two distinctive segregation patterns were observed under specific casting condition. One ofthem was in the form of solute reach channels called centerline segregation, as was already ex-plained and the others are dispersoids. Limited depth from the free surface of the strip is deco-rated with very fine intermetallic particles in the size of 1–3 m (Figure 3).

Generally four intermetallic phases are possible: Mg2Si as the age-hardening phase, -AlFe-Si(Mn) (Al5Fe3Si(Mn) Chinese script in a as-cast condition, Mn as substitution of Fe possible),

-AlFeSi (Al5FeSi little plates which are reducing ductility and -phase (Al8FeMg3Si6 - is bind-ing Mg) (Figure 4).

Figure 1: Grain structure of as-cast strip at the surface. Note the border of featureless zone marked with the arrows

Figure 2: Grain structure of as-cast material (a) parallel, (b) transverse to the casting direction

90

Detailed investigation of segregate channels at the center plane, called center-line-segrega-tion, has shown that they generally contain coarse eutectic cells and occasional coarse interme-tallic phases.

Despite of two different intermetallic particle geometry in the microstructure, their size anddistribution are very uniform and consistent compared to those observed in its DC cast counter-parts. While intermetallic particle size of DC cast material is larger than 6 m, their morphologyexhibit large aspect ratios, as well. This particle geometry is more prone to align itself in defor-mation direction, i.e. rolling direction.

It has already been known that manipulation of casting parameters might have strong influ-ence on tailoring micro structural features to certain extend. However, in the present study, allefforts were concentrated on optimization of casting parameters to achieve minimum or no sur-face segregations. This approach has allowed not only for obtaining segregation and ripple freesurfaces but also satisfactory productivity values in casting operation.

3.2 Micro and Macro Structure of Heat Treated Material

Twin roll cast strips of alloy AA6016 were cold rolled to 1,15 mm and then subjected to soluti-on heat treatment. The characterization of the material was done after aging for 5 days at roomtemperature to T4 temper. Similar processing steps were also applied to the DC cast and hot rol-led strip for fair comparison.

Table 2: Chemical composition of DC AA6016 material

The DC cast hot rolled AA6016 material which was used for comparison purpose shows avery similar chemical composition as the produced TRC 6016 material (Table 2).

Si Fe Cu Mn Mg Al

1,11 0,17 0,08 0,07 0,35 98,17

Figure 3: Distribution of intermetallic phase at the vicinity of the surface

91

Substantially small grain size was encountered through the thickness of TRC material aftersolution heat treatment which readily leads to re-crystallization. Contrary to the non-uniformgrain size distribution of as-cast strip, re-crystallized grain structure at the final gauge is equi-axed and uniform, regardless of its position in the thickness. Their sizes are almost half of thosein DC cast material (Figure 4). It was previously noted in the literature that the structure of thesheet prior to the final recrystallization, both in terms of the size and distribution of secondphase particles controls the microstructure developed during recrystallization [7].

Microstructural constituents of both materials, after T4 treatment, were also investigated.Figure 5a shows distribution and morphology of intermetallic particles at the mid thickness ofTRC AA6016. Compared to as-cast structure in which interdendritic area is decorated withcoarser particles, along with fine ones, their sizes were found to be relatively smaller after solu-tion heat treatment and T4 treatment.

As was expected, DC cast AA6016 revealed much coarser particles that are aligned in therolling direction. Their aspect ratio is considered to be the major factor for this alignment (Fig-ure 5b). Different solidification characteristics during casting operation of both material deter-mine their particle size and geometry. Coarser particles of DC cast/hot rolled strips are prone tobe breaking up into two or more fragments, especially during cold rolling stage of down stream

Figure 4: Grain structure of TRC (a) and DC cast (b) material after recrystallization

a) b)Figure 5: Intermetallic particle size and distribution at the half thickness of the sheet

92

operation. Corner profile of particle ends, sharp and rectangular, causes intense plastic strainingin the nearby matrix and leads to void nucleation at the corners of the particles. Rolling to thin-ner gauges leads this damage mechanism more likely to occurs.

3.3 Mechanical and Forming Properties of Heat Treated Material

Processing route and their parameters were determined to fulfill the typical requirements of au-tomotive industry for stamping operation of critical parts. The heat treatment was done in an in-dustrial scale on the continuous heat treatment line (floater type furnace). T4 temper wasachieved by solution heat treating at 540 °C/90 s followed by water quenching and aging 7 daysat room temperature. Mechanical tests were conducted in transverse direction to the rolling di-rection. Mechanical properties of paint bake response were investigated on the material exposedto 205 °C for 30 min after T4.

Table 3: Mechanical properties of TRC and conventional DC cast AA6016.

TRC and conventional AA6016 exhibit comparable results in both tempers. T4 and T6 tem-pers of TRC coils has slightly higher yield and tensile strength. Regardless of testing gauge, i.e.50 or 80 mm, almost identical total elongation values of TRC material were obtained with DCcast material. The uniform elongation is controlled by the relative strain hardening rates up tothe maximum load while the extent of post-uniform elongation depends on both strain harden-ing and strain rate sensitivity. Similar to the material produced with conventional methods, sub-stantial fraction of the total elongation belongs to the uniform elongation in the case of TRCmaterial.

All achieved mechanical properties with the current TRC AA6016 are in accord with the pre-viously published data [8–10] and comply the standard European specifications for automotivesheet applications.

Mechanical characterization was extended by constructing Forming Limit Diagrams (FLD)of both materials to determine limiting strains under simultaneously operating complex stressstates, as encountered in industrial stamping operations (Figure 6). FLDs were constructed fortheir T4 tempers at which stamping is carried out. DC cast material has slightly higher limitingstrains for the right side of the diagram than that of TRC. As this behaviour is assessed in thelight of strain hardening exponent, that is relatively higher in DC cast material, might lead todiffuse necking by retarding local thinning and resulting premature failure. However, for the

TRC DCT4 T62 T4 T62

Yield Strength (MPa) 110 230 105 210Tensile Strength (MPa) 220 275 210 260Uniform Elongation (%) 22 9 23 10Elongatin (A50) 29 12 29 13Elongation (A80) 26 – 26 – n value (4–6 %) 0,29 – 0,31 –r value (8–12 %) 0,64 – 0,65 –

93

strain combinations of minor < 0 and major > 0, representing strain combinations in stretching,TRC material exhibits better performance compared to its counterpart.

4 Summary

1. Contrary to the coarse and aligned characteristics of intermetallic particles in the material produced with DC casting/hot rolling route, very fine and uniformly distributed intermetallic particles of TRC sheet have pronounced contribution to the complex straining conditions of stamping operations and failure mechanism.

2. TRC sheet shows satisfactory performance in bending process which is unavoidable joining (such as hemming) and forming method of automotive industry.

3. Fine intermetallic particles improve surface appearance in bending and overall bendability performance.

4. A phenomenon, orange peeling, that deteriorates the surface appearance due to the coarse grain structure, can be avoided with much finer grain structure of TRC sheet.

5. Not only the uniaxial tensile properties, but also complex deformation characterization tech-nique, FLD, shows that TRC sheet has almost identical performance with that of DC cast counterpart.

6. Supersaturated outer fibers of the strip formed upon contacting with the caster rolls and asso-ciated solidification mechanism might result in different age hardening kinetics as exposed to solution heat treatment and subsequent age hardening. This phenomenon needs further detai-led studies.

7. In connection with the same reasoning, influence of alloy chemistry on manipulating mecha-nical performance of TRC strips need further investigation, especially with changing Si con-tent.

Figure 6: FLD of both material at T4 temper

94

5 References

[1] W.S., Miller, L., Zhuang, J., Bottema, P.De, Wittebrood, A., Haszler, Vieregge, 2000, Materials Science and Engineering, (A280), 37

[2] S.A. Arnold, 1993, J. Metals, 45 (6), 12[3] Y., Muraoka and H., Miyaoka, 1993, J. Mater. Proc. Tech. (38), 655[4] K., Sears, Automotive Engineering: Strategic Overview, 1997, (1), 55[5] S, Ertan, M., Dundar, Y., Birol, K. Sarıoglu, C., Romanowski, 2000, Light Metals, TMS[6] M., Dundar, K., Sarıoglu, Y., Birol, A.S., Akkurt, and C., Romanowski, 2002, ed. Das, S.,

Automotive Alloys, TMS[7] G.B., Burger, A.K., Gupta, P.W., Jeffret, and D.J., Lloyd, 1995, Materials Characteriza-

tion, 35, 23–39[8] L., Zhuang, R.De, Haan, J., Bottema, C.T.W. Lahaye and P. De, Smet 2000, Materials

Science Forum; 331–337:1309[9] J., Hirsch, 1997, Materials Science Forum; 242:33[10] S.M., Hirth, G.J., Marshall, S.A., Court, D.J., Lloyd , 2001, Materials Science and Engi-

neering; A319–321: 452

95

Magnesium Upward Direct Chill Casting

Fr.-W. Bach, S. Schacht, A. RossbergInstitute of Materials Science (IW), University of Hanover, Germany

1 Abstract

Since 2001, the development of a new continuous casting technology for magnesium alloys atIW has made mayor steps from laboratory to industrial scale. Compared to the first trials with40 mm diameter billets the process has undergone many changes and enhancements towards thecurrently used technology for casting high quality 90 mm and 203 mm diameter billets. Withproceeding development during a FP5-Project named “EuroMagUpCaster”, various advantagesof the process have become apparent. These achievements do not only affect safety topics, butalso the possibility to significantly increase productivity, to lower installation costs, etc. Com-pared to conventional vertical direct chill casting the upcasting process differs in several issues.This concerns especially the cooling conditions in the mould and the possibilities of exact ther-mal control of the molten metal right outside the mould. An effective secondary cooling zonereaching high cooling rates by usage of only small amounts of water-spray is another highlightof this technology. This contribution is giving a detailed insight into the actual stage of develop-ment of this innovative technology, which has the potential to become the state of the art in con-tinuous casting of magnesium in future.

2 Introduction

In order to increase the industrial application of magnesium wrought products, high qualityfeedstock for rolling and extrusion has to be available at an acceptable price level and in variousalloy compositions. The DC-casting technique, as for steel and aluminum, is suited to producehigh volumes of such material in a continuous casting process. In several research- and industri-al plants magnesium is cast in vertical (gravity) DC-casters (Figure 1, left). Unfortunately onlyfew standard dimensions and alloys can currently be purchased. Alternative feedstock has to beordered in high quantities which are not suited for today’s small series production. Regardingthe quality of such products a standardized level has not yet been established. Especially in thecase of alternative wrought alloys, the resulting microstructures and surface properties often donot fulfill the high requirements of forming processes such as rolling or extrusion. Thereforemilling of the surfaces and homogenization of the material has to be implemented in the processchain.

Another drawback is the security aspect of gravity casters. In the case of magnesium, smalldisturbances in the casting process can easily result in ruptures of the cast billets. In comparisonto aluminum, these failures are more likely due to the lower strength of the just solidified mate-rial and the smaller process window, which is caused by the high thermal conductivity and thelow volume specific heat capacity. Furthermore leakage of magnesium melt is much more criti-cal since the high reactivity with oxygen leads to strong fires at atmosphere which can not easily

96

be extinguished. Often flux (salt) has to be used in order to suffocate these fires. This method isquite effective, but goes along with the emission of corrosive gases, which aggressively damagethe surfaces of surround machines, etc.

Therefore a system which prevents high volume melt leakages offers great advantages forthe secure operation of continuous magnesium casters. At the Institute of Materials Science ofthe University of Hanover, the Upward Direct Chill (UDC) casting principle (Figure 1, right)was chosen for a magnesium specific adaptation. Since this casters requires an active dosingagainst gravity, it also offers the possibility to retain the melt in the furnace in case of billet rup-tures.

In the following article, the optimized plant construction and the so far examined casting ex-periments, which have been conducted within an EU - 5th Framework project shall be presented.

3 Plant Concept

Since the year 2000 a magnesium specific UDC-caster has been developed and optimized at theInstitute of Materials Science [3]. A first laboratory scale version with a diameter of 40 mmshowed the feasibility of the process. In 2001 the EU-funded “EuroMagUpCaster” project wasgranted. The scope of this international collaboration is to introduce this caster principle intonear industrial production. For this purpose two 90 mm moulds and one 203 mm (8 inch) mould were designed and build at IW and at the project partner KME.

Figure 2 illustrates the 3-D construction of the furnace with the attached mould. Within thepressurized container magnesium can be melted in an inert crucible under protective gas. Whenthe gas pressure is increased for dosing, the melt flows upwards through a heated riser tubewhich is via a compensator on top of the oven lid. The mould consists of a heated titanium inlet,a water cooled copper cylinder and a graphite insert. Through this porous ring, a viscous lubri-cant can be pressed into the mould. It avoids sticking of the melt and closes the shrinkage gapbetween billet and mould, which increases the heat transfer into the copper cooler.

Figure 1: Comparison of VDC- and UDC-casting principles [e.g. 1, 2]

97

Above the mould, a secondary cooling ring, which contains several water-gas-spray nozzlesis installed. The spray angle is tilted, so that water cannot drop downwards into the mould.

At several points in the caster, thermocouples and pressure sensors are installed, in order todocument and control the casting speed, the spray pressures, the lubrication flow and the heatertemperatures.

On the photo of plant (Figure 3) the feeding device, which pulls the billet out of the mould,can be seen. This very rigid construction consists of 4 columns, the mould platform, the fixationof the starter block, a spindle and a belt driven motor. This first device allows pulling speeds upto 300 mm/min and a maximum billet length of 900 mm. For the near industrial application inthe EuroMagUpCaster project, the consortium partner, Römer Födertechnik GmbH, has buildthree further plants with a max. pulling length of 2800 mm.

4 Experiments / Casting Parameters

Simulation results of the Spanish project partner, Inasmet [5], and experiences with the 40 mmmould, served as starting point in the determination of suited casting parameters. Fortunatelythe first set of temperatures, speeds and pressures already allowed the successful, stable produc-tion of 400 mm billets made of cpMg (99,5 %). It was soon shown that this process allows largeparameter variations in order to optimize the billet quality. Due to safety reasons first experi-ments were conducted without secondary cooling. Table 1 shows the parameter sets for alloyswith strongly different solidification behavior. In order to increase the casting speed, the secon-dary cooling has to be activated. First results of ongoing experiments lead to the estimation thatthe speed can be increased up to 30% in comparison to Tab. 1.

Figure 2: Pressurized furnace with 203 mm mould Figure 3: Assembly with feeding device [4]

98

Table 1: Casting parameters for d=90 mm mould, lubrication 150–200 ml/m² surface

*1 without secondary cooling

In Figure 4 it can be seen that a stable cooling condition in the mould is reached after approx-imately 200 seconds. The temperature of the titanium inlet (Ti ring) is plotted with higher reso-lution. Its alternating run is caused by early solidification on this part, which also results in theformation of a rough cast skin as can be seen in Figure 5. The thin rings which solidify there tooearly are pulled up after a certain time and the resulting cracks are closed by the following melt.This deficit shall be overcome by: a) higher melt velocity, b) stronger heating, c) better isolationand d) a changed geometry of the Ti-inlet.

5 Results, Conclusions and Outlook

In the first stage, the billets were analyzed regarding the formation of the liquid pool and the ap-pearance of the cast surfaces. Figure 5 allows to evaluate the results for cpMg and AZ80. Bothallows were cast approx. 40°C above their liquidus temperature with strongly different speedsunder otherwise equal conditions. It can be seen, that the intermetallic phases containing alloyAZ80 exhibits quite a different behavior. Still, in both cases the continuous casting was suc-cessful and the loss in the chipping is smaller than 8 mm on the diameter. The cast skin shallfurther be improved by a modified construction of the Ti-ring and the other measures mentionedabove.

AZ31 AZ80 ZEK100

Melt temperature [°C] 650-690 630-670 670-690Casting speed*1 [mm/min] 75-95 60-80 80-105Dosing over-pressure [mbar] 180-200 180-200 180-200Holding period [s] 4-7 4-7 4-6Primary Cooling [l/min] 20-40 30-40 30-40

Figure 4: Data for process analysis (AZ 31; without secondary cooling)

99

Figure 6 gives an example for the microstructure of the cast billets. The results of the alloysAZ80 and ZEK100 were chosen, because here the fine microstructure (approx. 120 μm), thenearly globular grains (strongly reduced appearance of dendrites) and the homogeneous distri-bution of the precipitations (only AZ80) can clearly be observed. As main reason for this positi-ve result, the controlled cooling conditions (high cooling rates) and the contamination free meltprocessing were determined. The alloy AZ31 shows a further reduced average grain diameterwhen the secondary cooling is activated. Results on the quantification of this parameter will bepublished soon to come.

As a next step the cast products will be tested at IW regarding their mechanical propertiesand by the other EuroMagUpCaster project partners with focus on their deformation behavior.Various samples have been delivered for extrusion, forging and even rolling experiments. Judg-ing from the microstructure and small sample tests, significant improvements are being antici-pated.

Furthermore, the cast products of the upscaled mould (d = 203 mm) are awaited with greatinterest, since in case of similarly positive results the step into series application of this moderncaster type comes much closer.

Figure 5: Comparison of liquid pool and billet surface (mould diameter d= 90mm)

Figure 6: Micrograph of AZ80 and ZEK100 (middle of the billet, d=90 mm, as cast condition)

100

6 References

[1] F. Pravdic, P. Egger: Einfluss der Gießparameter auf die Qualität von AZ31-Stran-gussbolzen; Leichtmetallkompetenzzentrum Ranshofen, Gießereiforschung, Band. 57, 2005, S. 26–33

[2] F. Bach, S. Schacht: Vertikaler Strangguß - ein aufstrebendes Verfahren, Werkstoffwis-senschaftliche Schriftenreihe, Band 64, 2004, S. 4–10

[3] U. Holzkamp: Entwicklung einer magnesiumgerechten Stranggusstechnologie, Disserta-tion, Fortschritt-Berichte VDI, Nr. 623, 2002

[4] F. Bach, S. Schacht: New continuous casting process for Magnesium Alloys, IMA confe-rence 2004; New Orleans, Louisiana, 9-12 May 2004

[5] A. Landaberea, P. Pedrós, E. Anglada, I. Garmendia: Numerical Simulation of the upward continuous casting of magnesium alloys, DGM Conference, Continuous Casting 2005, Neu-Ulm

101

Spray Forming of Advanced High Strength Aluminum Alloys

P. Krug, B. CommandeurPEAK Werkstoff GmbH, Velbert, Germany

1 Abstract

Most recently several new types of aluminum alloys were developed at the R&D Centre ofPEAK Werkstoff GmbH, located in Velbert, near Düsseldorf. In any case, these new alloysmake use of the high solidification velocity during spray forming. This leads to interestingmicrostructure with a high volume content of primary phases. These primary phases which canbe pure Si-crystals or intermetallic phases like Mg2Si, Al3Fe are responsible for the goodstrength also at elevated temperatures. In the case of Mg2Si volume fraction of up to 25 % couldbe reached. Due to the low density of the magnesium silicide, the overall density of this alloyhas been reduced down to 2,5 g/ccm, but with comparable strength to normal 2xxx series alloys.

After spray forming a subsequent extrusion is mandatory to close residual porosity resultingfrom the atomising process. In some cases a additional heat treatment is applied to optimiseproperties according to the application.

This presentation will show you the “making of” such alloys, their properties as well aspresent and potential applications.

2 Introduction

Aluminum alloys exhibit several advantageous properties or combinations of such propertieslike e.g. low density together with high thermal and electrical conductivity. Nevertheless,Young’s modulus, thermal expansion and wear resistance can be substantially improved byadding silicon. Unfortunately, casting of such alloys is not appropriate when the Silicon con-tents will exceed 22 wt.-%. Usually Silicon is limited up to 18 wt.-% used in alloys for pistonsor monolitic engine blocks. Exceeding this limit leads to a brittle behaviour of the casting alloyand, therefore, there is a need for a contemporary manufacturing method. Such a method isspray forming of high Silicon-Aluminum alloys. Contents of up to 35 wt.-% can be easily ob-tained with in-situ formed primary silicon particles. It is worth to be noted that 100% of the sil-icon is precipitated primarily out of the melt and no eutectic silicon will be present after sprayforming. The fine and homogeneous distribution of the silicon particles in a spray formed billetwill lead to a good machinability and formability. Such alloys show superior behaviour in termsof stiffness and wear resistance. High surface quality can be installed only by turning and, thus,subsequent grinding and coating is not necessary any more as the example of a spool valve forvariable camshaft timing shows.

102

3 Producing “Impossible” Alloys

Due to the high solidification velocity a variety of beneficial effects will occur. • Extended solubility• Extended stability of phases• Suppressed eutectic phases• Enhanced nucleation of preferred phases

In addition to such metallurgical effects, spray forming offers the opportunity of injectingpowders during spray forming process. Applying different conveying methods it is possible toinject • Spherical powders• Sharp edged powders• Nonmetallic powders (e.g. Carbides, Borides, Oxides)• Metallic powders of same type (Overspray)• Metallic powder of different type (e.g. Si-powder)• Nanoscale powders

The combination of the metallurgical and the injection tools will lead to unusual materialswith interesting properties. In investigating such production methods should lead at the end to aspecial principle of alloy design which should be claimed as “tailoring” of alloys redirected tospecific customer’s requirements [1].

3.1 Metallurgical Tools

The binary system aluminum-silicon is well investigated and the influence of small additions(refining Silicon with Phosphorus or modifying the Eutectic with Sodium, Strontium or Anti-mony) is well established among the foundries. The influence of these small additions will di-minish as solidification velocity increases and the solidification velocity will dominate the finalstructure and occurring phases. An interesting summary is given in [2, 3].

Starting from solidification near equilibrium condition one will expect a microstructure of ahypereutectic Al-35%Si-Alloy primary Si-Particles and Aluminum-Dendrites and Aluminum/Silicon Eutectic. Increasing the solidification velocity a microstructure as shown in Figure 1a.will occur. The picture reveal the microstructure of a spray formed Al-35%Si alloy (DISPALS220). The Primary Silicon becomes the dominating phase and the whole silicon content will bepreciptated as small and homogenous distributed particles. Increasing the solidification velocityfurthermore , i.e. by electron beam welding of the spray formed material will lead at least againto a microstructure with aluminum dendrites and a granular eutectic (Figure 1b). The coarse Sil-icon primary particles shown in that figure nucleated at already existing Silicon particles whichdid not dissolve completely during welding.

It is worth noting that other alloy systems show similar behaviour. For example in Figure 2an Al-Mg-Si alloy is given [4]. Clearly the chinese script morphology of the Mg2Si–Al-eutecticcan be examined (Figure 2a). In hypereutectic alloys a irregular eutectic and cuboidal primary

103

Mg2Si will occur (Figure 2b). Increasing solidification rate leads to an uniform microstructurewith globular and fine primary Mg2Si-precipitates. This alloy system is of specific interest sincethe density of a 25 vol-% Mg2Si –alloy will be below 2,5 g/ccm.

The Al-Mg-Si alloy shown in Figure 3 is spray formed. Although there is a certain size dis-tribution among the Mg2Si particles the show only a weak tendency to coarsen during heat treat-ment which make this alloy applicable for high temperature service.

A couple of alloys have been sprayed to “play around” with high volume contents of prom-ising phases. In Figure 4 two of these experimental alloys are shown. The captions also indicatethe Vicker’s hardness of the as sprayed material.

Hardness as sprayed: 207 HV30.

3.2 Injection Tool

As mentioned before spray forming offeres the opportunity to inject almost every type of pow-der. The activities with Overspray powder reinjection led to a stable, high efficiency process.First attempts to use the same equipment for injection of ceramic particles – especially of abra-

Figure 1a: Al-35%Si (DISPAL S220) , spray formed. Silicon is completely pre-cipitated as primary phase

Figure 1b: Electron beam weld seam in same material as figure 1a. Aluminum dendrites and fine eutectic is reintroduced to the microstruc-ture

Figure 2a: Al-8%Mg-3%Si, sand cast. Chinese script Mg2Si and Al-dendrites

Figure 2b: Al-9%Mg-5%Si, sand cast. Irregluar eutectic and primary Mg2Si pre-cipitates

104

sive powders like silicon carbide – were not successful. During a public funded project an alter-native method of powder transportation was elaborated. This so called “dense pack conveying”principle allows low velocities of the powder during transport into the spraying chamber. In themeantime several powders have been injected successfully up to 30 to 40 vol.-%.

A interesting example is the injection of particles which will already be present in the alloy.For example, during the spray forming of an Al-35%Si-alloy additional Silicon powder was in-jected. At the end one will receive a multimodal distribution of Silicon particles, ex-situ and in-situ formed. This method enables to influence the Si distribution when for wear resistance asmall amount of coarse Silicon is required.

The next step will be the injetion of nano scale powders. This should lead to supreme mate-rials since really nanodispersion hardened alloys can be produced. Due to the insolubility andthermal stability of nano-Al2O3 in Aluminum alloys a superior hot strength and creep resistancecan be implemented. There is a chance to replace the costly and time consuming mechanical al-loying.

Preliminary feasability studies have revealed that the injection of nanoscale powders is pos-sible but needs further modification and optimisation of the existing equipment. A public fund-ed project is set up within the WING programme of the Federal Ministry of Research andTechnology to work out an efficient as well as an effective processing route for such nano-MMC’s.

Figure 3a: Al-16%Mg-8%Si, spray formed

Figure 3b: Same as figure 3b, higher mag-nification

Figure 4a: Al-30%Cu-10%Mg spray formed. Hardness as sprayed: 180 HV30

Figure 4b: Al-25%Cu-6%Mn spray formed

105

4 Summary

Spray forming is more than just another production of existing alloys. It offers unique opportu-nities to create supreme materials with outstanding and tailored properties. We just openend thedoor a little bit. Let us enter a space of nearly unlimited metallurgy.

5 Acknowledgements

The authors would like to thank the Federal Ministry of Research and Technology (BMBF), theDeutsche Forschungsgemeinschaft (DFG), Stiftung Industrieforschung and Arbeitsgemein-schaft industrieller Forschungsvereinigung (AIF) for funding research projects related to sprayforming.

6 References

[1] P. Krug, B. Commandeur, “Sprühkompaktieren von Aluminium-Hochleistungs-Legie-rungen – Pflicht und Kür“, Abschluss-Kolloquium des SFB327, Band 7, 2004, p.123–136.

[2] W.J. Boettinger, J.W. Cahn, S.R. Coriell, J.R. Manning, R.J. Schaefer, „ Application of Solidification Theory to Rapid Solidification Processing, Semi-Annual Technical Report, Springfield, Va. : NTIS, 1982

[3] W.Kurz, D.J. Fisher, “Fundamentals in Solidification”, Transtech Publication, Switzer-land, 1989

[4] H. Hanemann, A. Schrader; Ternäre Legierungen des Aluminiums, Atlas Metallographi-cus Band II, Teil 2, 1952, Verlag Stahleisen m.b.H., Düsseldorf, S.128

Figure 5a: Al-35%Si (DISPAL S220) , spray formed. Addition Silicon powder was injected

Figure 5b: Same as figure 5a. with differ-ent Si-size injected.

106

A Method of VDC Hot Top Mould Design and Setting of Process Conditions

I.F. Bainbridge1 & J.F. Grandfield2

Cooperative Research Centre for Cast Metals Manufacturing (CAST)1 Division of Materials, The University of Queensland, Brisbane, Qld. 4072 Australia2 CSIRO, Division of Manufacturing and Infrastructure Technology, Locked Bag 9, Preston, Victoria, 3072, Australia

1 Abstract

The design evolution of VDC hot top moulds has occurred largely by innovation and practicalrefinement. Similarly, conditions under which a particular mould design is operated are set andremain unchanged throughout the cast. During a cast process conditions within the mould maychange such that control settings become inappropriate, or are at such a variance as to result inthe production of scrap product. The paper presents an approach to mould design and the deter-mination of process conditions based on an understanding of the critical parameters such asmould length, metal head and air pressure balance which control cast product quality. Newknowledge of fundamental properties controlling molten metal meniscus behaviour within themould is applied to the design of a 152mm hot top mould.

2 Introduction

Hot top moulds are used for the casting of extrusion billet. Several mould designs are availablecommercially and a number of in-house designs are in general use. Whilst hot top moulds havegenerally not been used for the casting of rolling ingot, design and operating principles now ap-plied to rolling ingot systems are similar to those used for the hot top billet moulds. Hot topmould designs are generally innovative and practical, but they frequently incorporate featuresthat are thought to be necessary or desirable, with many of these features evolving as attemptsare made to differentiate systems competing commercially, or to improve cast product quality.

The understanding of the complex interacting processes occurring within the mould to pro-duce the cast product has not necessarily kept pace with the practical development of mouldsand associated systems, including process control systems. With the advent of computer mod-elling, attempts are being made to simulate these complex processes and thereby refine existingmould designs and to generally improve the process as demanded by the constant competitivepressures for the production of a superior product at a lower cost.

The work introduced in this paper proposes an approach to mould design and the setting ofprocess conditions based on application of basic principles of mould heat flow and solidifyingproduct physical properties.

107

3 Mould Design

The position of the liquid/solid metal surfaces as solidification commences within the mould isa critical factor to be considered in the mould design. The product surface structure within themould is shown schematically in Figure 1. The important components determining the productsurface quality have been identified as the meniscus, the molten metal head and the solid sur-face formed by the two cooling mechanisms, viz., heat extraction from the mould and heat ex-traction by the sub-mould cooling. The refractory overhang is also important, too short and themeniscus will not be stable and gas will bubble up through the melt. The key dimensions are(Figure 1): • A and B, the height and width of the meniscus;• C the molten metal head height;• D the shell length, i.e., the total length of the solid surface projecting into the mould from the

point of impact of the sub-mould cooling; and• E the UCD.• F the refractory overhang• G the effective mould length.

The height A, and width B, of the meniscus may be determined by application of the modelof Baker and Grandfield [1] (Figure 2) which solves the Laplace-Young equation and is de-pendent upon:

Figure 1: Schematic of hot top mould, showing surfaces and important dimensions

108

• the surface tension of the molten metal ;• the contact angles formed at the mould and hot top refractory surfaces, M and R;• the pressure balance between the molten metal head height H and the gas pressure Pg.

The height A is given by

(1)

In order to calculate the meniscus width B, one must solve equation (2) a first order ordinarydifferential equation for y(x) (the shape of the meniscus which together with the initial conditiony(0) = 0 can be solved using a suitable numerical method such as Runge-Kutta; there is no exactintegral for the function).

(2)

Stable meniscus sizes for three different alloys were calculated by the above model usingmolten metal surface tension data for various aluminium alloys available from the work ofBainbridge [2]. The approximate mould length required for a particular mould diameter maythen be obtained by adding the values for the UCD, calculated from the method of Flood et al.[3] or a full 2D heat and fluid flow model (using casting speeds from standard commercial prac-tice) to the meniscus height for a particular alloy (Table 1). A definitive mould length cannot beobtained unless accurate data is available for the mould heat flow, which in turn is dependentupon the mould gap thermal conductivity. Published data for a range of alloys and casting con-ditions is not presently available, but is part of an on-going research program at CAST.

2 2cos cosM RA H H

2tan acos cos +2M

yH A y y

x

Figure 2: Meniscus model from [1]

109

Table 1: Calculated stable meniscus size and UCD for three different alloys, for a 152 mm diameter mould , a R of 10° and a zero head height; plus critical mould dimensions as calcu-lated from these two parameters (refractory plate overhang is calculated as B + 20%).

The data as derived above may then be applied to the critical dimensions of the mould. The

effective mould length is the sum of the stable meniscus height and the shell length (approxi-mated by the UCD) (A and D Figure 1), whilst the minimum refractory overhang dimension (Fin Figure 1) is based on the stable meniscus width. The final mould dimensions may then be es-timated by adding the effect of metal head height to the stable meniscus calculations. Theabove principles have been applied to the design of a mould for different alloys and operatingconditions.

A mould designed for the casting of one type of alloy is not commercially acceptable, hencea compromise in terms of effective mould length had to be chosen and then the critical operat-ing conditions of casting speed and casting temperature (these affect the UCD) and metal headheight (this affects A) chosen to obtain a meniscus that remained within the calculated stablesize range for the particular alloy being cast. A fully instrumented version of the mould hasbeen built and used to cast product on a pilot VDC casting unit. Results from the mould trialshave shown excellent agreement with the model’s predictions. Billet quality has been shown tobe totally predictable, e.g., Figure 4, showing the effect of a metal head height increase on castsurface. The smooth surface was cast within the mould design parameters, the rough surface re-sulted when the metal head was moved outside these parameters.

Alloy (Nm–1)

Meniscus Size (mm)A B

UCD (mm)

Effective Mould Length (mm)

Refractory Plate Overhang (mm)

1050 6063 7075

0.63 0.48 0.77

6.0 5.3 6.7

4.3 3.7 4.7

12.9 16.1 8.8

18.9 21.4 15.5

5.2 4.4 5.6

Figure 3: 6063 alloy billet cast in experimental mould

110

4 Process Conditions

The models used to derive the basic dimensions of the mould include operating parameters, inparticular metal head height, casting speed and molten metal temperature. A mould based on adesign that incorporates these factors must obviously be operated in a manner that maintainsthese parameters within acceptable ranges for the mould design and alloy being cast. The moulddesign calculations presented above were initially based on known casting speeds, metal tem-perature and molten metal head height conditions used commercially. The effect of a variationin one or more of these process parameters on the critical mould dimensions is shown inFigure 4, and had to be taken into account in the final mould design and then in the manner inwhich the mould was operated. Note the significant effect that metal head height has on effec-tive mould length.

Present process controls generally rely on pre-set values for each of the critical operating pa-rameters, with operator interception to change any particular value during a cast only occurringif a major problem arises. There is no real time measurement of mould conditions and the useof these conditions to control the process based on comparison of the measured values and mod-el requirements.

5 Discussion

Existing mould designs have been derived over time, largely by empirical methods. The availa-bility of data on critical properties of the cast product as it forms within the mould now permitsmould design to be based on an approach that recognises factors critical to the formation of high

Figure 4: Effect of a variation in metal head height or casting speed on the effective mould length

111

quality cast product. This modelling approach is still rudimentary considering the complexity ofthe processes occurring within the mould. The two models proposed take into account five var-iables for the calculation of the meniscus dimensions and eight variables for the calculation ofUCD. If the latter calculation was modified to calculate the shell length, i.e., the UCD plus thatsolid formed due to mould cooling then at least two further variables would be added. Using themodels it is possible to identify the variables that will have the most significant effect on theprocess and the limits to which each must be maintained to guarantee the casting of high qualityproduct.

Notwithstanding these limitations of the present model, the methodology is a significant de-parture from conventional mould design methods and recognises the conflicts and compromisesthat must be made to arrive at a commercial mould design capable of casting a range of alloys.Further, the application of the concept to the process conditions recognises that the mould wasdesigned using specific parameters that normally form part of the operating envelope. It is sug-gested that this link is not commonly made in commercial operations at this time.

6 Conclusion

A method of designing a VDC casting mould based on molten metal meniscus size and UCDhas been derived and the results of the initial testing of the concept suggest that the combinationof the mould design and operating parameters linked to the design may result in cast product ofmore consistent and higher quality. The concept is a significant departure from the present em-pirical design methods with the potential to offer to the industry more reliable and cost savingcasting systems. In addition, the concepts may be applied to the DC casting of other metals, e.g.,magnesium.

7 Acknowledgement

The work was supported by the CRC for Cast Metals Manufacturing (CAST). CAST was estab-lished under, and is supported in part by the Australian Government’s Cooperative ResearchCentres Scheme.

8 References

[1] Baker, P.W. and J.F. Grandfield. The Role of Surface Tension Forces in Gas Pressurized VDC Casting. in Seventh Australian Asian Pacific Conference Aluminium Cast House Technology. 2001. Hobart, Tasmania, Australia: TMS, USA. p. 195–204

[2] Bainbridge, I.:The Influence of Molten Metal Surface Properties on the Formation of Sur-face Defects on Vertical Direct Chill Cast Aluminium Alloy Products:Engineering, Phys-ical Sciences and Architecture:The University of Queensland:2005

[3] Flood, S.C., P.A. Davidson, and S. Rogers. A Scaling Analysis for the Heat Flow, Solidi-fication & Convection in Continuous Casting of Aluminium. in Modelling of Casting, Welding and Advanced Solidification Processes VII. 1995: TMS, USA. p. 801–808

112

Continuous Casting of Non Ferrous Metal Micro Wrought Shapes

J. Bast1, E. Bombach2

1TU Bergakademie Freiberg2Deutsche Solar AG Freiberg

1 Introduction

In the last years micro technology has developed into a key technology. An increasing need ofmicro-structured products is projected for the future. Information and communication technolo-gy, aircraft and aeronautics applications [1], as well as medical technology [2] and in particularthe automotive industry [3] are regarded as growing markets for the use of micro-structured sy-stems. Micro-components are not only used in mass produced parts, but also in products whichare manufactured as prototypes or in small quantities.

At present the mass and large-scale manufacture of micro-structured components is dominat-ed by silicon etching techniques, lithography methods [4] and injection moulding. Since thesemethods are usually two-dimensional, it is complicated to produce a third dimension in suchcomponents. Furthermore, because of the high cost of equipment these methods are not suitablefor small and medium quantity production of micro-structured components.

That is why methods are gaining importance, that are based on conventional mechanical cut-ting technologies, e. g. drilling, milling, turning, grinding, micro eroding and laser machining[5]. Unfortunately, with the exception of the above mentioned injection moulding, castingmethods play only a minor role for the production of micro-structured components, althoughthey permit a wide geometric spectrum and allow the manufacture of different materials.

2 Continuous Casting

Conventional continuous casting has an interesting potential for near net shape production ofwire and rods direct from the melt. Continuous casting methods are suitable for producing rodsmade of materials that are difficult to forge or roll and for producing products, where the castingprocess leads to significant cost reductions because of cross-section and dimensions of the pro-ducts. Presently the smallest economically produced cross-section made by a conventional con-tinuous casting process is above a diameter of 5 mm. With the conventional continuous castingprocess a metallic melt is poured into a mould. By cooling the mould walls the heat is removedfrom the melt, so that the metal solidifies inside the mould. The metal assumes the shape of themould and can be pulled-out as rod material. In the mid-1980s at Chiba Institute of Technologyin Japan a crystal growing process for the production of mono-crystals was developed [6]. Thisprocess makes it possible to produce very small cross-sections, with rod dimensions well belowthe dimensions of conventionally produced rods. In this process the mould is heated. The tem-perature of melt is held a little higher than the solidification temperature of the cast material.Heat is extracted from the cast product by means of a cooling device located nearby the mould

113

exit. The temperature of the mould is higher than the temperature of the rod [7], so that the heatdissipation occurs parallel to the cast direction and the nucleation of crystals on the mould sur-face is prevented [9]. The difference between conventional and micro-continuous casting is de-monstrated in Figure 1.

3 Experimental Micro-Continuous Device with Horizontal Arrangement

Based on the OCC-method [6] an experimental horizontal micro-continuous device was de-veloped in the Institute of Mechanical Engineering of the Technical University BergakademieFreiberg [8]. See Figure 2. The metal is molten and held in the furnace. At one end of the fur-nace wall is an outlet pipe which also holds the mould. The pipe and the mould are heated. Themould opening corresponds to the cross-section of the rod to be cast. The height of the melt inthe furnace is adjusted so, that the outlet is well filled with molten metal by metallostatic pres-sure. At the beginning of the casting process a starting dummy is attached directly to the meltnearby the mould exit. The dummy is cooled. The melt located in the mould loses enough heat,to solidify inside or nearby the end of the outlet while coupled to the dummy. The dummy ismoved backwards by a pull-out device and the attached profile wire is pulled out of the mould.

An experimental apparatus was designed and built for the production of micro-profiles. SeeFigure 3. The equipment consists of four main components: the melting system with mouldheating, the mould, the cooling system and the pull-out system. Furthermore, the equipment is

Figure 1: Schematics of conventional and micro-continuous casting processes

Figure 2: Schematics of the experimental horizontal micro-continuous device

114

connected to a computer system for data collection and process control. A resistance heatedmuffle furnace was used for melting and holding. It contained a crucible made of steel. Thepipe-like outlet holding the mould protrudes from the melting furnace. A resistance heated tubefurnace surrounds the outlet and maintains its temperature. See Figure 4.

The moulds are exchangeable and consist of steel or graphite. In the centre of the moulds isan opening with a cross-section equivalent to the geometry of the profiled rod. See Figure 5.

The design and the surface characterization of the mould strongly influence the parametersof the casting process and on the quality of the rods. For the production of rods with a high qual-ity the cooling system is very important. During the investigation three systems were tested:

• Cooling by water flow• Cooling by water bath• Cooling by air.

Because of the high specific heat of water, the turbulences when impacting the rod and thehigh heat transfer by cross-flow, flowing water is a very effective cooling system. Unfortunate-

Figure 3: Basic design of the trial equipment for micro continuous casting of small profiled rods

Figure 4: Melting system

115

ly, the direct impact of the water stream on the cast rod leads to vibration with resonance ef-fects, which ruptured the rod and influenced the quality of the rod negatively. Because of thelow distance between the cooling stream and the mould during the casting process cooling wa-ter was sprayed on the mould. This leads to sudden chilling of the mould and to a change of thecooling conditions of the system. In the investigation showed that air cooling was the bestchoice. See Figure 6.

4 Casting Results

The developed micro continuous casting device was used for the production of profiled rods oftin and aluminium with a cross-section below 1 mm and 900 mm length. The profiled rods hadcircular and more complicated cross-sections. During the investigation the following tin rodswere produced:

• Circular profile with a diameter of 1.0 mm, 0.9 mm, 0.5 mm and 0.3 mm with 900 mm length,

• Star-like profile with 900 mm length,• Square profile with 400 mm length.

Furthermore, circular profiles of aluminium and lead were produced with 400 mm length.Figure 7 displays the enlarged images of the profiled tin rods. For the examination a metallo-

Figure 5: Mould in the melting crucible Figure 6: Cooling of the rod by air

Figure 7: Rods with circular (top) and star-like (bottom) cross-section

116

graphic specimen of the rods was embedded into resin and a micrograph was made. Figure 8shows the outline of the cross-section, the geometry of the mould and the cross-section of thecast rod. The surface quality corresponds to the results which can be obtained with conventionalmanufacturing methods (micro-drilling, grinding, high-speed milling) or the die casting proc-ess. The surface roughness is 0.5 to 1.5 microns.

5 Summary

The continuous casting process permits near net shape production of rods directly from the meltand processing of a wide variety of materials. With a growing need of micro-structured com-ponents it is possible, that the micro-continuous casting process of small rods can widen thespectrum of existing processes. For the production of micro-structured rods a horizontal micro-casting device was developed. With this equipment micro-structured profiles of tin and alumini-um with dimension from 0.3 to 1.0 mm were produced. Furthermore, experimental and theoreti-cal investigations about the analysis of process parameters and process control were carried out.

6 Acknowledgements

The financial support provided by the Deutsche Forschungsgemeinschaft Germany is gratefullyacknowledged.

Figure 8: Outline of rod, geometry of the mould and micrograph of profiled cast rod

117

7 References

[1] J. Fahrenberg, Werkstattstechnik 2000, 90, 11/12, 484–486[2] I. Beltrami, et. al. In Tagungsband Aachen 10.-11. Juni 1999, 39– 420[3] T. Seubert, in Micromaterials for Automotive, Leipzig, 26. Juni 2003, 39[4] W. Ehrfeld, Feinwerktechnik, Mikrotechnik, Messtechnik, 1992, 100, 282–286[5] J. Hesselbach, et. al., Werkstattstechnik 2003, 93, 3, 119–128[6] A. Ohno, Metals, 1986, 38, 14–16[7] H. Soda, G. Motoyasu, A. McLean, A. Ohno, Advanced Materials & Processes, 1995, 4,

43–45[8] E. Bombach, Diss. TU Bergakademie Freiberg, April 2004[9] H. Soda, G. Motoyasu, A. McLean,S. D. Bagheri, D. Perovic, Cast Metals, 9, 1996, 1,

S. 37–44

118

Influence of Quality of Water and Surface Roughness on Quenching Rate

Jacek Król, Eckehard SpechtOtto-von-Guericke-University, Magdeburg, Germany

1 Introduction

When measuring the heat transfer from a hot surface quenched by a spray, researchers typicallyuse water that has been carefully distilled so as to remove any dissolved impurities. From expe-rience in industrial practice it is known that the quality of the water has an influence on thequenching rate in continuous casting. The effect of additives on pool boiling heat transfer hasreceived much more attention than their role in spray cooling. Pool boiling studies done by Na-jibi [1] showed that dissolved salts precipitate on the heater surface during boiling. Surfactantsalso have an important effect on pool boiling [2]: they promote bubble nucleation and foamingin the liquid and significantly increase heat transfer. Qiao and Chandra [4] found that dissolvinga surfactant in the spray water significantly increased heat transfer. The surfactant reduced theliquid-solid contact angle and produced foaming in boiling droplets; both effects increased thesolid area wetted and enhanced surface cooling. King et al. [5] observed the evaporation of saltsolution droplets placed on a hot stainless steel plate. In opposite, they found that the dissolvedsalts reduced the vapor pressure of water and therefore decreased the droplet evaporation rate.Until now no studies are know which considered the influence of the quality of the water syste-matically.

The roughness of the strand’s surface influences the Leidenfrost temperature and there wisethe quenching rate, too. But there are no papers which quantitatively investigated the influenceof the roughness. Both effects were experimentally researched using a relatively new method ofmeasuring technique.

2 Experimental Set-Up

The measurement set-up, sketched in Fig. 1, was used to investigate atomized spray quenching.In Atomized Spray Quenching [6], the spraying water is atomized into fine droplets by com-pressed air and sprayed onto a hot surface. The drops partially evaporate and then move awayfrom the superposed airflow. Thus, the vapor film is avoided as it is the case for other quen-ching techniques. The main component of the measurement set-up was a thin metallic sheet,electrically heated up far above the Leidenfrost temperature. This metal sheet was cooled fromone side by the water spray. During heating up, the water spray was covered. When the stationa-ry temperature was reached, the water spray has been started. On the black colored oppositeside the transient surface temperature was measured using an infrared camera. Because of thesmall thickness (0.3 mm) and high conductivity the temperature on both sides are similar. Fromthe given heat flow and the temperatures the local heat transfer coefficient especially in the tran-sition range from film to nucleate boiling can be determined with a high accuracy. To correlate

119

the heat transfer with the water spray characteristics, the distribution of the drop velocity andthe drop diameter of the spray were measured with a combination of 2D-Phase-Doppler-Ane-mometer and patternator. Internal mixing air blast atomizers were used for the water spray ge-neration. With these nozzles, the water is mixed with compressed air inside the nozzle andejected afterwards. The air used for atomization was supplied up to pressure of 0,5 MPa. Thequality of water was regulated with different concentrations of salt and soap. With sand paper adefined roughness of the surface was adjusted.

The impingement density was measured with the patternator sketched in Fig. 2. It consistedof several tubes with a diameter of 5 mm, arranged in line array parallel to the spray axis. Thewater amount Mw was collected by the tubes over a period of time t. The collected water wasstored in small bottles. The mean impingement density was calculated from Eq. (1),

Figure 1: Experimental set-up

Figure 2: The patternator for measuring of the impingement density

120

(1)

where dt is the tube diameter. Therefore, a stagnation flow similar to the flow pattern used forquenching was measured.

The heat flow caused by water spray qsp was computed from the energy balance by Eq. (2) atthe investigated location, where dT/dt was measured

. (2)

The heat loss by radiation, free convection and conduction in metal sheet was negligible. Theheat transfer coefficient of the water spray sp was calculated from the temperature differencebetween the hot surface Ts and the water spray Tsp by Eq. (3)

. (3)

3 Experimental Results

The aim of the measurements of the atomized spray quenching was to find out the influence ofthe main parameters on heat transfer such as: diameter of the droplets, velocity of the droplets,impingement density, quality of water, surface roughness.

Figure 3 depicts profiles of the impingement density, measured with the patternator in aplane 200 mm in front from the nozzle. The measurements were carried out with a constant wa-ter flow through the nozzle with different air pressure supplying the nozzle. It can be seen thatthe maximum of the impingement density is at the spray centre. If the water flow is constant, themaximum impingement density decreases with increasing air pressure.

ws 2

t

4 Mm

d t

sp p losddT

q c s qt

sp sp s spq T T

Figure 3: Impingement density distribution in radial direction

121

Figure 4 presents profiles of the mean volumetric diameter d30. The water flow through thenozzle is constant at 7 kg/h, and the air pressure is varied from 0.2 MPa to 0.4 MPa. The largestdrops were measured in the centre of the spray. For a constant water flow through the nozzlewith an increasing air pressure, the drop diameter decreases. To estimate the expansion of thespray, a qualitative profile of the impingement density is added to Fig. 4. The measured maxi-mum velocity and mean volumetric diameter are about v = 30 m/s and d30 = 20 m, respectivelyin the centre of the spray.The heat transfer coefficient of atomized spray quenching is presented in Fig. 5 depending onthe surface temperature for a high and a low impingement density. For both impingement densi-ties, the heat transfer coefficient is independent from the surface temperature for the values abo-ve the Leidenfrost temperature of about 300 °C.

The heat transfer coefficient depending on the impingement density is presented in Fig. 6 forwater spray quenching and for atomized spray quenching. In atomized spray quenching the va-por film does not form like in water spray for surface temperatures above the Leidenfrost tem-perature. The impingement density exerts the highest influence on the heat transfer coefficient.

Figure 4: Distribution of mean drop velocity and diameter

Figure 5: Heat transfer coefficient of atomized spray quenching vs surface temperature

122

For atomized spray quenching the heat transfer coefficients also increase with the air pressure.This occurs because of the changing droplet characteristics, mainly the change in the velocity.The heat transfer coefficient increases with the droplet velocity. The effect of droplet size on theheat transfer coefficient was not found to be significant in this investigation of atomized waterspray. Atomized Spray Cooling leads to much higher heat transfer coefficients than water spraycooling.

Figure 7 presents the mean heat transfer coefficient for atomized spray quenching dependingon the surface temperature for different concentration of salt and soap. The experiments showthat the amount of the dissolved salt and soap in distillated water decreases the heat transfer co-efficient compared to the pure distillated water.

The run of the temperature is presented in Fig. 8 for smooth and rough surface. With sandpaper a defined roughness of the surface was adjusted. The roughness of the surface exerts highinfluence on the quenching rate. This occurs because of the larger droplet contact area for a

Figure 6: Heat transfer coefficient versus impingement density

Figure 7: Heat transfer coefficient vs surface temperature

123

rough surface than for a smooth surface. For increasing impingement density the difference ofthe quenching rates between smooth and rough surface decreases.

4 Conclusions

The experiments demonstrate that the heat transfer coefficient is mainly determined by the im-pingement density. Atomized spray cooling has a much higher heat transfer coefficient thanspray cooling. The amount of the dissolved salt and soap in distillated water decreases the heattransfer coefficient compared to the pure distillated water. Higher concentration of salt and soapincreases a little the Leidenfrost temperature. Surface roughness does not influence the Leiden-frost temperature, but exerts high influence on quenching rate for the smallest impingementdensity.

5 References

[1] Najibi, S.H., Muller-Steinhagen, H., Jamialahmadi, M., “Boiling and Non-Boiling Heat Transfer to Electrolyte Solutions,” Heat Transfer Engineering 1996, 17, 46–63

[2] Hetsroni, G., Zakin, J.L., Lin, Z., “The Effect of Surfactants on Bubble Growth, Wall Thermal Patterns and Heat Transfer in Pool Boiling,” Int. J. of Heat and Mass Transfer 2001, 44, 485–497

[3] Qiao, Y.M., Chandra, S., “Spray Cooling Enhancement by Addition of a Surfactant,” ASME J. of Heat Transfer 1998, 120, 92–98.

[4] King, M.D., Yang. J.C., 1997, “Evaporation of a Small Water Droplet Containing an Additive,” Proceedings of the 32nd National Heat Transfer Conference, Vol. 4, pp. 45–57

[5] Puschmann, F., Specht, E., “Atomized Spray Quenching as an Alternative Quenching Method for Defined Adjustment of Heat Transfer,” Steel Research 2004, 75, 283–288

Figure 8: Run of temperature with time for smooth and rough surface

124

Electromagnetic Casting of Aluminum and Steel Billet Using Slit Mold

J. Park1, M. Kim1, H. Jeong2, G. Kim2

1Research Institute of Industrial Science and Technology, 2POSCO

1 Introduction

Many defects, such as an oscillation mark, a crack, etc., were formed in the surface of the billetinto which the steel was cast through the continuous casting. Such surface defects affect seri-ously the productivity and workability in the continuous casting process 1–3). Surface defectcauses any delay in the process, any labor loss and even any material loss. Steel industries,which have to reduce the used amount of fossil fuel, are striving to realize the hot direct rollingprocess without any process of reheating the billet as one of the energy saving processes. Fromthis viewpoint, EMC technology is being studied. This technology has been developed for themoldless casting of any material with a light specific gravity and a good electric conductivityusing the electromagnetic force in stead of the mold 4–5). However, the specific gravity of thesteel is relatively high, and its thermal conductivity and electric conductivity are low. Besides,its casting speed is high. Therefore, it is thought that the moldless casting of the steel is almostimpossible. This is why steel industries turn their attention to the study on application of EMCin the soft contact casting form using the mold. There are two EMC methods. One is to use thehigh frequency magnetic field of tens of kHz or higher6–11), and the other is to use the low fre-quency magnetic field of 60Hz to 200Hz12). According to the former method, the meniscus hasa good stability but it has a difficult point that a special mold and a special power device are re-quired. On the other hand, the latter method has a major challenge that the meniscus is unstable.

This study has been conducted as a part of the study to improve the surface quality of the bil-let into which the steel was cast through the continuous casting by applying the EMC technolo-gy using the high frequency magnetic field to the steel. In this study, the effect of the moldshape on the billet surface shape was examined through the continuous casting experiment ofthe steel, and the effect of various factors on the surface quality of the billet was researched byobserving the shape of the early solidified shell and measuring the meniscus shape and the moldflux consumption13,14).

In the aluminum continuous casting, what has not yet been solved in aluminum continuouscasting technology is a simultaneous improvement of surface quality of the cast and castingspeed. In this research, an electromagnetic casting technology has been proposed that can in-crease casting speed as well as improve the surface and inner quality of the cast through appli-cation of electromagnetic casting technology and electromagnetic stirring technology by meansof the slit mold being studied in electromagnetic casting of steel. The possibility of the proposedidea in this work has been confirmed through basic study.

125

2 Concept of EMC

Following is a brief explanation about the concept of soft contact EMC using the high frequen-cy magnetic field. As shown in Figure1, in case the electric current is applied to the coil, themagnetic field is induced in the mold and the electric current gets to be induced by the magneticfield. This electric current generates the magnetic field and an electric current in the moltenmetal inside the mold through the mold segmented by slits. This induced current not only heatsthe molten metal but also generates the electromagnetic force in the molten metal by acting onthe magnetic field. This technology using the Joule heating and the Lorentz force is just theEMC principle using the high frequency power. The electromagnetic force enlarges the menis-cus curvature of the molten metal in contact with the mold, thus, improves the inflow of themold flux and at the same time reduces the contact pressure between the shell and the mold. Themeniscus is heated by the Joule heating and a thin early solidified shell gets to be formed in thelower part of the meniscus, which inhibits any hook, the root of the oscillation mark from takingplace.

3 EMC of Steel Billet

3.1 Lab Scale Experiment

The billet surface morphology of face are shown in Figure 2, respectively. It can be known fromFigure 2 that in case the electromagnetic field was not applied, the OSM was clearly formed. Itcan be also known that in case (b) the coil current 670 A was applied, the face surface shapewas much improved, but the OSM still remained in the corner. Further, it can be known that incase (c) the coil current 830 A was applied, both the surface shape of the face and the surfaceshape of the corner were greatly improved so that the OSM was formed about 0.15 mm deep orbelow. It was also identified that in case (d) the coil current 1100 A was applied, the respective

Figure 1: The concept of EMC using slit mold

126

surface shapes of the face, the corner and the off-corner were all greatly improved, but groovestook place along slits in the casting direction of the billet. It is thought that the electromagneticfield acted more intensely than under any proper condition and a great magnetic pressure actedon the molten steel in the slit parts where the magnetic flux density was relatively high, and as aresult, the molten steel was pushed into the mold, and thereby these grooves were formed. Thismark is about 1mm deep, being deeper than the OSM in case the electromagnetic field was notapplied. From this result, it can be known that any optimal electromagnetic condition is requiredfor improvement of the billet surface.

Since the forming position and the shape of the early solidified shell have a great effect onthe surface quality of the billet, it is very important to identify the effect of the electromagneticfield on the shape of the early solidified shell. 100 g of the FeS powder was wrapped in the Althin sheet, and it was added directly to the molten steel in the mold at the terminal stage of thecasting. Figure 3 shows the shape of the early solidified shell respectively in case the electro-magnetic field was not applied and in case the coil current 830 A, which was thought to be theoptimal condition in this study, was applied. The black part in the photograph is the part wheresulfur was intermixed, and it is also the solidified part after FeS was added. It can be known thatin case (a) the electromagnetic field was not applied, the early solidified shell was thick and un-even, having a hook, the root of the OSM formed, and further that after the solidified shell wasformed, the molten steel containing sulfur overflowed so that it existed even in the outer face ofthe solidified shell. On the other hand, in case the coil current 830 A was applied, the early so-lidified shell was thin and even, having no hook formed.

Figure 2: Surface morphology of the billet

Figure 3: Early solidified shell of steel Billet

127

3.2 Conventional Scale Experiment

The mold plate was the copper alloy identical to the material used in POSCO’s billet casting op-eration. Its length was 800 mm, the size of casting section was 163 163 mm and an inside cor-ner radius of 8 mm.

The amount of the mold powder consumption of the EMC operation is shown in Figure 4 forvarious casting conditions compared to the conventional casting operation. In the EMC opera-tion, it was in the range of 110–130 % of the conventional casting. This can be explained as thework of the magnetic force in such a way that the magnetic force made the meniscus into acurved shape, and thus the mold flux would easily flow into the gap between the strand and themold wall. Joule heat also helped in melting the powder and in keeping the molten flux at rela-tively high temperature compared to the conventional casting. Particularly, it is noteworthy thatthe mold powder consumption in case of EMC without mold oscillation was no less than that ofthe conventional casting.

Figure 5 shows typical billet surfaces for various coil currents. In case that the coil currentwas 0 A, i.e. the conventional casting, billets had the OSMs of its nominal depth in the range of0.45 0.15 mm. When the coil current was 500 A, the OSM disappeared a lot to its nominaldepth in the range of 0.20 0.05 mm except the corner region. When the coil current was1000 A, the OSM disappeared everywhere to its nominal depth in the range of 0.10 0.04 mm.When the coil current increased to 1200 A, the OSM appeared again in the form of wave markalthough its nominal depth was not severe in the range of 0.18 0.05 mm.

Figure 4: Mold powder consumption versus casting conditions

Figure 5: Typical appearance of billet strands

128

4 EMC of Aluminum Billet

In the case of general moldless EMC, there is a limit to increasing of the casting speed due tothe unstable molten metal surface. Since the depth of the molten metal pool is also low due tothe low casting speed, it is difficult to stirring the molten metal to make fine grains. But in thecase of EMC technology proposed in this study, it is shown that casting speed can be increasedsince molten metal surface and mold contact through slit mold, and then the molten metal ismaintained stably. Fine grain can be produced through stirring of the molten metal by installa-tion of an electromagnetic stirring (EMS) device at the below of the EMC coil, which is due tothe increased depth of the molten metal pool.

4.1 Experimentals and Method

A schematic view for experimental devices are shown in Figure 6. The mold is a cylindricalshape with an inside diameter of 100 mm and a length of 200 mm. The slit is machined in somepart of mold for longitudinal.

After about 40 kg of 2024 Al alloy and A356 Al alloy were melted, respectively, and castingwas made through variation of casting speed, EMC current, EMS frequency, and current. Theeffect of various parameters on the quality of the billet was reviewed. Moreover, the depth ofthe molten metal pool within the mold was measured with a 2 mm stainless steel rod duringcasting. The shape of the solidified pool was observed through injection of molten zinc into themolten metal within the mold at the final stage of casting. The specimen after casting was takenand microstructure was observed.

4.2 Results and Discussion

In Figure 7, a surface morphology of the 2024 billet produced by general casting is shown.Casting speed was elevated to 0.25 m/min and breakout occurred at 0.3 m/min. It can be seen

Figure 6: EMC mold system

129

that a severe ripple mark was formed on the surface of the billet and the mark became weaker ascasting speed increased. The surface morphology of the billet produced by the EMC at a coilcurrent of 1000 A is shown in Figure 8. The ripple mark is disappeared. Moreover, castingspeed was also increased up to 0.45m/min and breakout occurred at 0.5 m/min.

5 Conclusion

5.1 At the EMC of Steel Billet

The depth of the oscillation mark was decreased from about 0.6mm to 0.1 mm or below and anyappropriate electromagnetic condition was required for obtaining the optimal surface shape.Electromagnetic continuous casting apparatus was suited on a billet caster of POSCO works,and commercial scale casting test was successfully performed. The major observations with theEMC operation include that the oscillation mark was improved a lot to its nominal depth of lessthan 0.1 mm]

5.2 At the EMC of Aluminum Billet

Casting speed as well as surface quality of the billet can be improved with electromagnetic con-tinuous casting technology.

Figure 7: Surface appearance of the 2024 Al billet without EM

Figure 8: Surface appearance of the 2024 Al billet at EMC under coil current 1000 A

130

It was confirmed that surface quality can be improved and internal structure can be globularthrough the simultaneous application of electromagnetic casting technology and electromagnet-ic stirring technology.

6 References

[1] S. Kumar, I.V. Samarasekera and J.K. Brimacombe: ISS Trans., 1997, June, 53[2] K. Kawakami, T. Kitagawa and Y. Hiratani: Tetsu-to-Hagane, 67(1981), 1190[3] K. Kawakami : Tetsu-to-Hagane, 74(1988), 1204[4] D. C. Prasso, J. W. Evans and I. J. Wilson: Metall. Mater. Trans. B, 26B(1995), 1243[5] Ch. Vives : Met. Tran. B, 16B(1985), 377[6] I. Sumi, K. Sassa and S. Asai: Tetsu-to-Hagane, 78(1992), 447[7] T. Li, S. Nagaya, K. Sassa and S. Asai: Metall. Mater. Trans. B, 26B(1995), 353[8] J. P. Park, H. T. Jeong, D. J. Sim and H. Y. Kim: Korean Ins. of Met. & mat., 36(1998),

1598[9] S. Itoyama, H. Tozawa, T. Mochida and K. Kurokawa : ISIJ Int., 38(1998), 461[10] H. Nakata, M. Kokita and K. Ebina : Tetsu-to-Hagane, 80(1994), 711[11] S. Furuhashi, M. Yoshida and T. Tanaka : Tetsu-to-Hagane, 84 (1998), 625[12] T.T oh, E. Takeuchi, M. Hojo, H. Kawai and S. Matsumura : ISIJ Int., 37(1997), 1112[13] H. Kim, J. Park, H. Jeong and J. Kim, ISIJ Int., Vol.42 (2002), No.2, p171–177[14] J. Park, H. Jeong, H. Kim and J. Kim, ISIJ Int., Vol.42 (2002), No.4, p385–391

131

Aluminium Alloy Strip Casting Using an Unequal Diameter Twin Roll Caster

T. Haga1, H. Watari2, S. Kumai3

1Osaka Institute of technology, 5-16-1, Omiya, Asahiku, Osaka city, 535-8585, Japan2Oyama national collage of Technology, 771, Nakakuki, Oyama city, Tochigi, 323-0806, Japan3Tokyo Institute of Technology, 4259, Nagatuda, Yokohama city, Kanagawa, 226-8502, Japan

1 Abstract

An unequal diameter twin roll caster with a long solidification length was devised to cast alumi-nium alloy strip with a thickness of about 5 mm at speeds higher than 20 m/min. The characteri-stics of the unequal diameter twin roll caster with a long solidification length are as follows.The diameter of the lower roll is four times larger than that of the upper roll. The solidificationlength is long. The casting speed is high. Using the unequal diameter twin roll caster, a 4.5-mmthickness of 6111 strip could be cast at a speed of 30 m/min. Low superheat casting and semiso-lid casting was adopted to the unequal diameter twin roll caster with a long solidification length.The microstructure of the as-cast strip was equiaxed and spherical, not columnar. The mechani-cal properties of the strip rolled from roll-cast strip were almost as same as that of the strip madefrom cast ingot.

2 Introduction

In automobiles, decreasing weight is one of most important problems to be solved. Aluminiumalloy parts are adopted instead of steel parts to reduce the weight. However, aluminium alloyparts are more expensive than steel parts. Many parts in automobiles are made from thin plate. Itis desirable to economically produce aluminium alloy thin strip, and roll casting has this capabi-lity. Aluminium alloy strip can be cast directly from molten metal by a twin roll caster. The twinroll caster can eliminate the need for many processes: for example, grinding of the ingot sur-face, hot rolling and some degree of cold rolling. This shows that much energy and operationalcosts can be saved. Moreover, the need for many machines and space for machines can be redu-ced. The cost of installation can be saved, too. Thus, the twin roll caster has the advantage ofcost saving. The twin roll caster also has the advantage of rapid solidification. The microstruc-ture becomes fine by rapid solidification, and the mechanical properties can be improved. Insummary, the twin roll caster has many advantages. However, the twin roll caster also has dis-advantages: slow casting speed, limitation of usable alloys and center segregation. These disad-vantages must be addressed. In the present study, an unequal diameter twin roll caster with longsolidification length was devised in order to reduce the disadvantages of the roll casting of alu-minium alloys. The unequal diameter twin roll caster with a long solidification length is diffe-rent from the conventional twin roll caster for aluminium alloys and the conventional unequaltwin roll caster. In the present study, 6111 aluminium alloy was cast into strip using an unequal

132

diameter twin roll caster with a long solidification length, and the mechanical properties wereexamined.

3 Experimental Conditions

The unequal diameter twin roll caster with long solidification length (UDTRCLS) used for theexperiment is shown in Figure 1. Experimental conditions of roll casting by the UDTRCLS areshown in Table 1. 6 kg of 6111 aluminium was melted in the crucible by an electric furnace inair. The melt in the crucible was transferred to the UDTRCLS. The melt was poured on the lo-wer roll through the cooling slope. The roll was rotated at desired speed (30 m/min) when themelt was poured. The special operation was not carried out at the start of the casting. The lubri-cant was not used on the roll surface. The melt head was maintained at 30 mm at the top of theroll. The cooling slope was used in order to decrease the temperature of the melt. The coolingslope was water cooled, and the surface was coated by BN. Casting load was 0.18 kN per unitwidth. This load was very small, because the hot rolling was not desired. However, this loadwas enough to make heat transfer between the strip and the roll better. The cast strip by theUDTRCLS was homogenized and cold rolled to 2 mm. Intermediate annealing was applied tothe cold rolled strip, and then the strip was cold rolled to a thickness of 0.5 mm. T6 heat treat-ment was applied to the 0.5-mm thickness of cold rolled strip. Tension and180 degrees bendingtests were used to examine the mechanical properties.

Figure 1: Photograph of unequal diameter twin roll caster with long solidification length in casting

133

Table 1: Experimental conditions

4 Result and Discussion

The 6111 aluminium alloy strip could be cast continuously at a speed of 30 m/min using theUDTRCLS. The thickness of the 6111 strip was 4.5 mm. In the CTRCA, a 7-mm thickness ofstrip was cast at speeds slower than 5 m/min. The productivity of the UDTRCLS is three timessuperior to that of the CTRCA, and so the problem of low productivity in the CTRCA is solvedby the UDTRCLS. The UDTRCLS strip was cast from the semisolid slurry, which contained asolid fraction of about 5 %. The semisolid slurry did not solidify around the nozzle and the damplates, and clogging did not occur. Figure 2 shows the surfaces of the as-cast UDTRCLS strip.The upper surface of the as-cast strip is different from the lower surface. The lower surface isrich in metallic luster, and the upper surface is poor in metallic luster. When solidification oc-curred, the solid fraction of the semisolid slurry in the upper side of the strip was larger than thesolid fraction in the lower side. Therefore, the upper side was not the same as the lower side.However, the upper side became the same after the cold rolling.

Figure 3 shows the microstructure of a cross section of the 6111 as-cast UDTRCLS strip.Figure 4 is an enlarged view of Figure 3. The microstructure of the cross section of the strip castby the CTRCA is usually a columnar structure and is symmetric in the thickness direction. The

Upper roll Material: mild steel, Diameter: 250mm,width: 100 mmwater cooling, non-lubricant

Lower roll Material: mild steel, Diameter: 1000mm, width: 100 mmwater cooling, non-lubricant

Roll speed 30 m/minSolidification length Upper roll: 60 mm, Lower roll: 300 mmLoad of roll 0.18 kN/mm (per unit width)Specimen 6111 (6 kg)Melt temperature 655 °CHeat treatment : T4,T6 Solution: 530 °C – 4 h, Water quenching

Aging: 160 °C – 6 hHomogenization 540 °C – 4 hIntermediate annealing 530 °C – 2 hCooling slope Material: mild steel, Length: 300 mm, Width: 100 mm

Water cooling

Figure 2: Surface of 6111 as cast strip

134

thicknesses of the solidification layers cast by the two rolls are the same. As confirmed by Fig-ures 3 and 4, the microstructure of the strip cast by the UDTRCLS is different from the micro-structure of the strip cast by the CTRCA. The microstructure is equiaxed, not columnar. Themicrostructure around the interface of the solidification layers is a spherical structure. This isthe effect of the semisolid casting. The thicknesses of the solidification layer cast by the upperand lower roll are not the same. The solidification layer cast by the lower roll is thicker than thesolidification layer cast by the upper roll, as shown in Figure 3. This effect is due to the differ-ence of the solidification length. The solidification length of the lower roll is five times longer

Figure 3: Cross section of 6111 as cast strip

Figure 4: Enlarged view of cross section of 6111 as cast strip

135

than that of the upper roll. The interface between the solidification layers is near the upper sur-face. This reduces the segregation. The material around the interface toward the upper roll so-lidified rapidly because the thickness of the upper solidification layer is thin. This centersegregation remains an important problem to be solved. This problem could be solved or the de-gree of segregation could be decreased.

The microstructure of the cross section of the 0.5-mm-thickness strip cast by the UDTRCLSafter homogenization, cold rolling and T6 heat treatment is shown in Figure 5. The nonuniform-ity of the microstructure is somewhat improved. The roll-cast strip is usually used after sometype of treatment, such as homogenization and cold rolling. Therefore, the nonuniformity of themicrostructure in the thickness direction of the as-cast strip is not a defect, if the nonuniformitycan be improved by heat treatment and plastic deformation.

Figure 5: Cross section of 6111 strip of T6 condition after some treatments

Figure 6: Outer surface and cross section of 180 degrees bent strip. Thickness was 1mm and heat treatment was T4.

136

The mechanical properties of the strip cast by UDTRCLS after T6 heat treatment are shownin Table 1. The mechanical properties of the roll-cast strip are almost the same as that of thestrip made from ingot. It is said that the mechanical properties of the roll-cast strip are inferiorto that of the strip made from ingot. This disadvantage is lessened by the UDTRCLS. Figure 6shows the result of the 180-degree bending test on the UDTRCLS specimen. There was nocrack at the outer surface. This shows that the roll-cast 6111 strip has enough ductility for hemforming. Hem forming is an important forming application used in the panels of automobiles.

5 Conclusions

An unequal diameter twin roll caster with a long solidification length was designed and assem-bled. Experimental casting was performed to investigate the characteristics of this twin roll ca-ster and the strip cast from semisolid slurry. This caster can cast a 6111 aluminium strip of 4.5-mm thickness at a speed of 30 m/min. This shows that the casting speed can be increased wit-hout decreasing the thickness of the strip. A tensile test and 180-degree bending test were per-formed. The strip showed good mechanical properties.

6 References

[1] A.I.E.Nussbaum, FATA Hunter SpeedCaster Inauguration, LIGHT METAL AGE, Oct. (1997)34–38

[2] N.Toyama, H.Aho, H.Arai, H.Yoshimura, Direct casting of stainless sheet by unequal-diametered twi roll method, Tetsu-to-Hagane, Vol.71(1986)A245-A248

[3] D.B.Love, J.D.Nauman, Controlling the physical and mechanical properties of cast stain-less steel band, TMS Proc. of an international Symposium on casting of Near Net Shape Products, Hawaii, 1988, 597–611

137

Fabrication of High Purity Copper Rod with Unidirectional Solidification Structure by Continuous Casting Using Cooled Mold

Hoon Cho, Duck-young Hwang, Han-shin Choi, Shae K. Kim, Hyung-ho Jo. Korea Institute of Industrial Technology; 994-32 Dongchun-dong, Yeonsu-gu; Incheon, 406-130, Korea

1 Abstract

It is generally known that OCC (Ohno Continuous casting process)using heated mold has to beintroduced to produce cast rod representing single crystal and uni-directional solidification mor-phology. However, in the present study, it is expected that unidirectionally solidified high-puri-ty copper rod could be fabricated by optimization of withdrawal speed in typical continuouscasting process using cooled mold. This paper discusses the production of small cross sectionhigh-purity copper rod by the process of continuous casting, and the effect of varying the pa-rameters of withdrawal speed and casting temperature on the cast grain morphology of the rodand its drawing characteristics to fine wire. The heat transfer within the graphite mold is alsodiscussed with respect to the withdrawal speed and casting temperature.

2 Introduction

Recently, the application of copper as a substitute of gold bonding wire has been investigatedbecause formation of brittle intermetallic compound such as aluminide between aluminium sub-strate and bonding wire can be suppressed. Especially, copper being cheaper than gold and withhigher conductivity and better stiffness, is a viable, cost-effective alternative[1].

In order to manufacture copper ultra fine wire for bonding wire in integrated circuit package,continuous casting process, which can produce high purity copper rod with small cross section,has to be optimized to produce cast rod without internal defects and to control microstructureorientation which can prevent wirebreaks in wiredrawing process[2].

The paper presented here is mainly aimed at investigation of influence of varying parametersof pulling speed, superheat and rod diameter on grain morphology of the casting rod and on itsdrawing characteristics to ultra fine wire.

3 Experimental Procedures

In the vacuum vertical continuous casting system, the graphite mold (nozzle) was attached tobottom of melting furnace and was designed to cast rod in the range of 7 to 11 mm in diameter.Especially, the top of the mold was heated by copper melts and the bottom of the mold is cooledby cooling jacket. Therefore, it is very important to understand heat transfer within the graphitemold during continuous casting process. Several thermo-couples were inserted to the mold with

138

different depth to measure temperature profile of the mold and to forecast solidification frontposition in the mold. In order to prevent the contamination of copper cast rod and to control heattransfer within the mold, high-purity copper rod with high conductivity was used for starting barduring continuous casting process.

Table 1: Chemical composition of high purity copper used in this study (unit : ppm)

Microstructure evolution and mechanical properties of high-purity copper casting rod wereperformed with respect to casting temperature, withdrawal speed and cast rod diameter. In orderto investigate the variation of micro-orientation with different process parameters in continuouscasting process, EBSD(Electron Back Scattering Diffraction, Oxford INCA Crystal) was used.

4 Result and Discussion

4.1 The Influence of Withdrawal Speed on Unidirection Solidification

The present work on casting 9mm diameter high-purity copper rod has shown there is an opti-mum withdrawal speed around 30 to 50 mm/min, which leads to small columnar grains oriented

Element Ag Al Cr Fe Mn Ni Pb S Si

5N-Cu 0.1 0.1 0.04 0.5 0.002 0.2 0.2 2 0.5Cast-rod 0.04 0.08 0.04 0.5 0.002 0.018 0.11 0.9 0.5

Figure 1: Longitudinal cross sections of the Cu rod with different withdrwal speed(diameter : 9 mm, casting temp. 1200 °C)

139

at an angle to the axis of the cast bar. Figure 1 shows longitudinal cross sections of the Cu rodwith different withdrawal speed when rod diameter is 9mm and casting temp. is 1200. With awithdrawal speed of up to 30 mm/min, macrostructure of the cast rod represents columnargrains, which is solidified and grown in parallel direction with continuous casting direction. Incontrast, as the withdrawal speed increases, the angle between the direction of casting and thepredominant crystallization direction increases[3].

Figure 2 shows pole figure of the copper rod with respect to different withdrawal speed. Asshown Figure 2 (a), unidirectionally solidified copper rod with (100) direction produced by acontrol of solidification front position within mold through continuous casting process using acooled mold. As the withdrawal speed increases to 100mm/min, the misorientation angle ofgrains increases and solidification occurs in several directions such as (100) and (110). The sol-id-liquid interface(Solidification front) position was measured by making a temperature profilewithin the mold.

Figure 3 shows schematic illustration of solidification front and crystal growth direction withdifferent withdrawal speed(V, mm/min) when rod diameter is 9mm and casting temp. is1200. As withdrawal speed increases, distance between top of the mold and liquid/solid inter-

Figure 2: Pole figure of the Cu rod with different pulling speed(diameter : 9mm, casting temp. 1200 °C)

Figure 3: Schematic illustration of solidification front and crystal growth direction with different withdrawal speed(V, mm/min) when rod diameter is 9 mm and casting temp. is 1200 °C

140

face(solidification front) becomes far. As shown in Figure 3 (a), the solidification front couldenter the location where the mold temperature is similar with copper melts when withdrawalspeed is 30mm/min. Thus, heat transfer mainly occurs according to copper rod used for startingbar. It can be mentioned that unidirectionally solidified grain result from heat transfer directionparallel with casting direction. In contrast, the solidification front could enter the location wherethe mold temperature is lower than copper solidus temperature when withdrawal speed is100 mm/min. Thus heat transfer also may occurs according to perpendicular to casting direc-tion. The heat transfer induces solidification of crystal with several growth direction such as(100) and (110) as shown in Figure 2(b).

4.2 The Influence of Casting Temperature on Unidirection Solidification

In order to investigate the effect of continuous casting temperature on unidirection solidificationof Cu cast rod, the continuous casting with different casting temperature varied 1150 °C and1200 °C was carried out. The macrostructure showing unidirection solidification is illustrated inFigure 4 and the pole figure of Cu cast rod uis shown in Figure 5.

As shown in Figure 4 and Figure 5, Cu cast rod was solidified unidirectionally when with-drawal speed is set to be up to 30 mm/min even though the casting temperature was varied to1150 and 1250 °C.

Consequently, it is very important to control and predict the position of solidification frontwithin continuous casting mold in order to produce Cu cast rod with unidirection solidification

Figure 4: Longitudinal cross sections of the Cu rod with different withdrwal speed(V) and casting temperature (T)

141

structure. Among the continuous casting process parameters including withdrawal speed, cast-ing temperature and cast rod diameter, it can be mentioned that withdrawal speed during contin-uous casting process is dominant process parameter to control the position of solidification frontwithin the mold.

5 Conclusions

It is expected that unidirectionally solidified high-purity copper rod could be fabricated by opti-mization of withdrawal speed in typical continuous casting process using cooled mold.

The unidirectionally solidified grain represents growth direction in (100). As the withdrawalspeed increases to 100mm/min, the misorientation angle of grains increases and solidificationoccurs in several directions such as (110) and (100).

It is very important to control and predict the position of solidification front within continuoucasting mold in order to produce Cu cast rod with unidirection solidification structure. Amongthe continuous casting process parameters including withdrawal speed, casting temperature andcast rod diameter, it can be mentioned that withdrawal speed during continuous casting processis dominat process parameter to control the position of solidification front within the mold.

Figure 5: EBSD measurement result of the Cu rod with different withdrawl speed ( ) and casting temperature (T)

142

6 References

[1] Harman G., Wire bonding in microelectronics: materials, processes, reliability, and yield.(New York : McGraw-Hill, 2nd ed., 1997), 1–11

[2] Wilson R. et al., “Continuous casting of high purity small diameter copper rod”(Paper presented at copper ’90, refining. Fabrication, markets, Vaesteras, Sweden, 1-3 Oct. 1990) 238–244

[3] Ohno A. et al., “ Studies pertaining to the position of the solidification front in horizontal Ohno continuous casting system”( Paper presented at AFC-8, Bangkok, Thailand, 17-20 Oct. 2003) 689–699

143

High Speed Roll Casting of Al Alloy and Mg Alloy Strips

T. Haga1, H. Watari2, S. Kumai3

1Osaka Institute of technology, 5-16-1, Omiya, Asahiku, Osaka city, 535-8585, Japan2Oyama national collage of Technology, 771, Nakakuki, Oyama city, Tochigi, 323-0806, Japan3Tokyo Institute of Technology, 4259, Nagatuda, Yokohama city, Kanagawa, 226-8502, Japan

1 Abstract

A high speed twin roll caster focused on aluminum alloy and magnesium alloy was designedand assembled. Vertical type was adopted from the point of easiness of pouring of molten metalin the roll bite. Improvement of increase of cooling rate and casting speed was attained in thiscaster. A356 for casting was cast to investigate the characteristics of this roll caster. AZ31 ma-gnesium alloy for forming, AM60 and AZ91 magnesium alloy for casting were cast to investi-gate the roll-castability of magnesium alloy. Aluminum and magnesium alloy strip thinner than4 mm was cast at speeds from 60 m/min to 150 m/min. Use of copper roll, non-use of partingmaterials and reducing of the thickness of the strip were contributed to improve the cooling rateand casting speed of the strip. The microstructure of the strip cast by the high speed twin rollcaster was usually duplex structure. Center of the thickness was equiaxed or spherical structure,and surface sides were short columnar or equiaxed structure. Ununiformity of the microstruc-ture at thickness direction could be improved after cold rolling and heat treatment. This casterwas useful for aluminum alloy which freezing zone was wide. Strip of A356 showed good duc-tility. Annealed A356 strip was not broken at 180 degrees bending test. Cold deep drawing ope-rated to T4-strip. Warm deep drawing of AM60 and AZ91 strip could be operated, and LDRwas greater than 2.0.

2 Introduction

Aluminum alloy for casting has good castability. Therefore, the roll casting of the aluminum al-loy for casting is easy. However, it is said ductility of the aluminum alloy for casting is poor.Therefore, aluminum alloy for casting is not suitable for the forming. It is possible that im-provement of the poor ductility of the aluminum alloy for casting by the rapid solidification us-ing the twin roll caster. If the aluminum alloy for casting can be used for forming, the variety ofthe aluminum alloy for recycle can be reduced. Improvement of poor ductility of aluminum al-loy for casting was investigated. Properties of the cast strip were investigated by metalography,tension test, bending test and deep drawing.

Magnesium sheet is normally made from slab by hot rolling. It is difficult to do cold rollingof magnesium alloy. The hot rolling is essential for magnesium sheet. Therefore, the slab or thestrip must be heated up to hot condition at every rolling pass. This makes the sheet of magnesi-um very expensive. If the thin strip was cast directly from molten metal, number of heating androlling could be reduced. As the result, the cost of the magnesium alloy sheet could be reduced.The twin roll caster is suitable for the magnesium alloy, as the strip can be cast directly frommolten metal. The high speed twin roll caster is better to make the price lower. AZ31 is normal-

144

ly used for sheet metal forming. AZ61 and AM60 are better than AZ31 from the point of thestrength. However, it is difficult to make sheet of AZ61 and AM60 as the hot rolling of thesemagnesium alloys is very difficult. Therefore, roll casting of AZ61 and AM60 is very useful. Inthe magnesium alloy, castability at high speed roll casting was investigated. Properties of castmagnesium alloy strips were showed by the hot rolling and warm deep drawing. The hot rollingwas operated on as-cast strips. Warm deep drawing was tested on the hot rolled strip.

3 Experimental Conditions

A high speed twin roll caster (HSTRC) is shown in Figure 1 and experimental condition isshown at table 1. Magnesium alloy was cast in the air without inert gas.

Table 1: Experimental conditions

4 Roll Casting of Aluminum Alloy

Figure 2 show roll casting in operation and as cast A356. A356 could be cast to the strip conti-nuously at speeds from 60 m/min to 150 m/min. The strip was thinner than 4.0 mm. TheHSTRC was able to cast at speed 10 times higher than that of the conventional twin roll caster

Roll material Material: copper, Size: diameter 300 mm, width 100 mmLubricant: non-use, Cooling: water

speed 60, 90, 150,180 m/minaluminum alloy A356Magunesium alloy AZ31, AM60, AZ91superheat 15 °CCooling slope

Material: mild steel, Size: length 300 mm, width 100 mm,Inclination angle: 60 degrees, Coating: BN, Cooling: water

separating force 0.14 kN/mmSolidification length 100 mmmelt head 100 mm

Figure 1: Schematic illustration of a vertical type high speed twin roll caster (HSTRC)

145

for aluminum alloy (CTRCA). The strip did not stick to the roll without the lubricant. The stripcast by HSTRC was half thickness of the strip cast by the CTRCA. The strip cast by the HSTRCwas thinner than that cast by the CTRCA without the hot rolling. The microstructure of the stripcast by the HSTRC is different from that of the strip cast by the CTRCA. The surface and thick-ness distribution were improved by the rolling. 20 % reduction was enough to improve the sur-face and the thickness distribution. The cause of this defect is oscillation of the meniscus at thetip of the nozzle. This defect could be improved by cold rolling. The surface of the strip cast bythe CTRCA is same at both sides.

The microstructure of most of the cross section was equiaxed structure or spherical structure.This is the effect of semisolid casting and rapid solidification. Microstructures of cross sectionof A356 strip at as cast and T6 condition are shown in Figure 3. The microstructure of as cast

Figure 2: Experiment of roll casting and as cast strip

Figure 3: Microstructure of cross section of A356 strip

Figure 4: Surface of A356 strip after 180 degrees bending. treatment : homogenization, cold rolling, annealing and 180 degrees bending

146

strip was not uniform at thickness direction. This ununiformity of the microstructure was im-proved by cold rolling and annealing. The microstructure of T6 strip was almost uniform. Eu-tectic Si of T6 strip was spherical and very fine. Annealed 0.5mm thick strip could be bent at180 degrees without crack at outer surface as shown in Figure 4. Figure 5 shows the result ofcold deep drawing of T4 heat treated A356 sheet. The cold deep drawing could be operated toT4-A356 of 1.0 mm thick. These results show roll-cast A356 had good ductility.

5 Roll Casting of Magnesium Alloy

High speed roll casting of magnesium alloys was tried. The casting was operated in the air wit-hout sealed gas.The magnesium strip could be cast at the speeds up to 150 m/min. The castabili-ty became better as the Al content became greater. The castability of AZ91 was better thanAZ31. The relationship between the roll speed and the strip thickness was as same as that of thealuminium alloy. The strip thickness became thinner as the roll speed became higher.The strip,which was thinner than 2.0 mm, could be cast by the high speed twin roll caster. The thicknesscould be reduced down to 0.5 mm only by 3 times of hot rolling. This shows that the magnesi-um sheet could be made at very low cost.

Figure 6 shows surface of the as-cast strip and the rolled strip. The as-cast strip had metallicluster at proper casting condition. They say that it is difficult to make the sheet of AM60 andAZ91 as hot rolling of slab is difficult. However, hot rolling of AM60 and AZ91 could be oper-ated, and the sheet of 0.5 mm thick was obtained.

Figure 5: Cold deep drawing of T4-A356

Figure 6: Surface of roll cast AZ31 strip at speed of 120 m/min

147

Figure 7 shows the microstructure of roll cast AZ31 strip. The microstructure of as-cast stripwas not uniform at thickness direction. The microstructure of center area was spherical struc-ture, and the near surface was the equiaxed structure. This was not the character of the magnesi-um alloy but the high speed twin roll caster. This non-uniformity of the microstructure becamealmost uniform after the hot rolling.

The warm deep drawing of the roll cast magnesium strip could be operated. Result wasshown in Figure 8. L.D.R. (limiting drawing ratio) was greater than 2.0. The warm deep draw-ing of the magnesium alloy for casting like AM60 and AZ91 was able. The warm deep drawingof AZ91 is more useful than that of AZ31, as the AZ91 is stronger than AZ31.

6 Conclusions

A high speed twin roll caster of vertical type was designed and assembled to cast aluminum al-loy and magnesium alloy thin strips. Some devices were adopted to realize rapid solidificationof the strip. Casting was operated, and ability of the high speed twin roll caster was estimated.The A356 aluminum alloy could be cast at speeds from 60 m/min to 150 m/min. The magnesi-um alloy strip could be cast at high speed, too. Thickness of the strip was from 1.5 mm to3.5 mm. The microstructure of the strip was not columnar structure but equiaxed structure. It

Figure 7: Microstructure of roll cast AZ31 strip at speed of 120 m/min

(a) AM60 (Pd = 40 mm, t = 0.5 mm) (b) AZ91 (Pd = 32 mm, t = 0.5 mm)Figure 8: Warm deep drawing of roll cast magnesium alloy (T = 250 °C, LDR = 2.0)

148

became clear that the high speed twin roll caster has ability to improve the deterioration of thealuminum alloy by impurity. The 0.5 mm thick of sheets of Am60 and AZ91 magnesium alloyswas obtained by hot rolling from the roll cast strip. Warm deep drawing could be operated toAZ60 and AZ91 sheet, and their L.D.R was greater than 2.0.

7 References

[1] M.Yun, X. Yang, D.V.Edmonds, J.D.Hunt, P.M.Thomas: Cast Met.Vol.4-2(1991),p.108[2] D.V.Edmonds, J.D.Hunt: Extraction Refining and Fabrication of Light Metals, CIM,

Ottawa, (1991), p.257[3] T. Motegi: Proc. of the ICAA-6, (1998), p.297

149

Simulation / Modeling

150

151

State-of-the-Art in the Modelling of Aluminium and Copper Continuous Casting Processes

J.-M. Drezet1,2, M. Gremaud2 and M. Rappaz1

1 Computational Materials Laboratory, Ecole Polytechnique Fédérale de Lausanne, IMX-STI, Station 12, CH-1015 Lausanne, Switzerland2 Calcom-ESI SA, Parc Scientifique, CH-1015 Lausanne, Switzerland

1 Abstract

With the advent of powerful and cheap computers, modelling of solidification processes at themacroscopic scale has become a standard practice in industry, in particular in continuous ca-sting processes. Indeed, commercial software packages are available for the modelling of heatand fluid flow, as well as for stress-strain calculations. Such software can even be used in an“inverse way” in order to deduce casting parameters (e.g., heat transfer coefficients) from mea-surements (e.g., temperature). Electro-magnetic stirring, which is increasingly used to refinegrain structures, can also be modelled by coupling hydromagnetic and thermal aspects. In moreadvanced approaches, convection in the liquid, heat exchange and stress developments are cou-pled all together in a mixed Lagrangian-Eulerian formulation. The start-up phase, which is cru-cial for many continuous casting processes (e.g., DC casting of Al alloys), is another field ofdevelopment for which mixed formulation is promise full. Modelling of macrosegregation isstill a critical issue as it can have different origins: convection, solidification shrinkage, grainmovement/sedimentation, deformation of the mushy zone. Substantial progresses have beenmade in this area as well. Besides macroscopic aspects, modelling of microstructure and defectformation is an active field of research, in particular for microporosity and hot tearing. The pre-sent contribution will review the state-of-the-art modelling of aluminium and copper semi-con-tinuous or continuous casting processes, at both the industrial, more macroscopic, approach andthe still more academic, microscopic, level. This review is based on the communication givenby M. Rappaz at the occasion of the 40th anniversary of the R&D Centre of Hydro-Aluminium(formely VAW) in Bonn in May 2004 [1].

2 Introduction

From an academic point of view, the VDC (vertical direct chill) casting process is schematicallyshown in Figure 1. Although fairly simple in its principle, it involves several interplaying phe-nomena which finally render modelling approaches complex. At the macroscopic scale (topics 1to 4), heat and fluid flow modelling is of course the first step in order to predict the delivery ofmetal, the melt pool depth, the thermal gradient, the local solidification rate, etc ... Fluid flow isalso essential in determining macrosegregation, i.e., transport of solute species at the macrosco-pic scale, and grain structure, i.e., influence of convection on dendrite fragmentation and graintransport. Almost as important is the calculation of stress build-up and strains, since this condi-tions air gap formation between the ingot and the mould or bottom block, and thus heat transfer.

152

Knowing the deformation of the slabs also helps to calculate the mould shape that will minimisescalping operations, while stress assessment is a key element to predict hot cracking formation.

Although fluid flow and solid deformation are governed by the same basic conservationequations (conservation of mass and momentum), their associated rheologies are so differentthat combined Eulerian/Lagrangian approaches are required to handle the large displacements/small stresses of the fluid and the small displacement/large stresses of the solid. Although sever-al softwares can now calculate in a coupled way these two aspects, the approaches usually re-main “one-phase”. If such approaches are very useful in predicting the combined interactionbetween heat flow, fluid flow and solid deformation, they cannot address in details phenomenaoccurring in the mushy zone. Recently, one has seen the emergence of “two-phase” approachesin which the solid and liquid equations are averaged over a typical volume element, with appro-priate exchange terms, in order to predict hot cracking tendency.

For a metallurgist, macroscopic entities such as cooling rate, residual stresses, etc., are fine…but not sufficient! He wants to have access to macro- and microstructures as well as to defectssuch as microporosity and hot cracking (topics 5 to 8 in Figure 1). While Cellular Automata orgranular-type approaches in the mid-nineties were a step forward for the prediction of grainstructure formation, the advent at the same time of the phase field method in the materials sci-ence community gave great hopes to solve the problem of microstructure prediction. Amazingprogresses have been made, but the technique still remain delicate to use and very CPU-intensi-ve. For the prediction of microporosity, one has seen recently the emergence of 3D computati-ons combining solidification shrinkage, gas segregation, nucleation and growth of pores.Finally, the prediction of hot tearing has really become a …. “hot” topic since the first approachcombining strains of the solid skeleton and liquid feeding was published in 1999. The presentpaper reviews very briefly the state of the art of solidification modelling, focusing mainly onDC casting, and will outline some of the challenges that remain. It is largely inspired by the

Figure 1: Schematics of the VDC casting process. Topics 1–4 correspond typically to a macroscopic approach, whereas topics 5–8 correspond to a microscopic one. Typical variations of the thermal gradient, G, of the solidifi-cation rate, v, and of G/v are shown as a function of x [1].

153

work carried out in two European research projects, EMPACT (1996-2000) and VIRCAST(2000-2004) and in the on-going project POST (Porosity Stress).

3 Macroscopic Modelling

3.1 Heat Flow

Heat extraction is of course essential in solidification modelling, not only at the process scale,but also at the level of the microstructure-defects. The thermal gradient, G, and solidification ra-te, v, can be used in microstructure maps [2]: the fineness of the microstructure is essentially afunction of (Gv), i.e., of the cooling rate, while the type of microstructure (e.g., columnar,equiaxed) depends on G/v. Niyama’s criterion for microporosity prediction is also a function ofG/v [3]. It should be pointed out that, even in the steady state regime, the solidification rate v atthe liquidus position depends on position and is not equal to the casting speed, vc, as shown inFigure 1. Modelling of heat flow in DC casting is no longer a challenge from a numerical pointof view, in the start-up or steady state phases. The enthalpy method with evolving/activatedmeshes is robust in handling the strong non-linearity associated with latent heat release. Themain questions in this area are related to materials properties and boundary conditions. Forboth, inverse methods have become a standard practice to calibrate the calculations [4]. As anexample, Figure 2 shows the heat flux deduced at the surface of a DC cast ingot as a function ofthe distance from the top liquid surface. One can clearly distinguish the heat flux associatedwith primary cooling, the air gap formation (nil flux) and the strong heat extraction associatedwith water cooling.

3.2 Fluid Flow

Despite the advanced CFD (computational fluid dynamics) software on the market, fluid flowcalculations remain a task of specialists, especially when solidification occurs simultaneously.Since such calculations are performed in a single domain, containing the liquid and solid pha-

Figure 2: Computed heat flux as a function of the distance to the top liquid level for an AA5182 alloy [4]

154

ses, a penalty method is used to make the velocity in the solid resume to the casting speed. Theviscosity can be made a function of the volume fraction of solid or, better, a drag term similar toDarcy’s equation for a porous medium can be introduced in the momentum conservation equa-tion. On the other hand, as the laminar viscosity of metals is very low, a turbulence model isalso required to get realistic fluid flow pattern and intensity. One of the challenges that remainin this field is the validation of the calculated velocity field. Although some probes were desi-gned by Vives [5], one has very often to rely on water models to calibrate the flow [6–7].Another challenge is the interaction between the flow and the mushy zone. Besides fragmentati-on of dendrites [8] and formation of feathery grains [9,10], which are both favoured by convec-tion, it is not easy to model accurately the narrow boundary layer near the liquidus in a transientregime, since it moves over time and normally requires adaptive-evolving meshes. As an ex-ample, Figure 3 shows the velocity field calculated with Fidap in a round billet fed from the la-teral side [10]. The maximum shear rate of the liquid occurs at the opposite side of the gatingand favours the growth of twinned dendrites in this region.

3.3 Stresses and Strains

Deformations of the solid during cooling and stress build-up are important issues in DC castingof aluminium alloy [11–13]. Deformation both limits the heat exchange with the mould and bot-tom block, and thus the production rate, and modifies the shape of the ingot. As shown schema-tically in Figure 1, butt curl, butt swell and lateral faces pull-in make the final shape of the slabto deviate from a parallelepipedon. In a transverse section as well, a convex mould has to beused if a rectangular cross section at the end of solidification is desired, since pull-in at the cor-ners is smaller than at the mid-rolling faces. Stresses are also important as they induce hotcracking or even “cold” cracking.

In this area, modelling is very mature, partially thanks to the EMPACT program, and manycommercial software exist. However, deformation especially of the lateral faces is strongly in-fluenced by the inclination of the thermal gradient, and thus by the sump shape. For example, an

Figure 3: Velocities and temperatures at the surface and in the vertical symmetry-plane for an AA1050 DC cast billet, injected from the lateral side, as calculated with the Fidap software. Temperatures are in °C, (Tliq = 657 °C, Tsol = 645 °C). In-flow velocity is 0.02 m/s, from [10].

155

increase of the casting speed makes the liquid sump deeper and thus the thermal gradient morehorizontal, in which case pull-in is increased. By the same argument, the shape of the liquidpool being influenced by convection, and therefore by the distribution bag, accurate solid defor-mation calculations require to couple them, not only with heat-, but also with fluid-flow simula-tion. This challenge is not met by all commercial software. By the same argument, compressivestresses of the mushy zone can expulse interdendritic liquid out (i.e., sponge effect [14]) thusleading to deformation-induced segregation. Tensile stresses on the contrary lead to an openingof the mushy zone, inducing either segregation (healed hot tears) or hot cracks.

3.4 Macrosegregation

Despite the effort done during the EMPACT program, macrosegregation, i.e., solute inhomo-geneity at the scale of the ingot, still remains a challenge in DC casting for several reasons [15].First, macrosegregation can be induced by solidification shrinkage (inverse segregation) andexudations, by forced or natural convection, by grain movement and sedimentation and by de-formation of the solid skeleton in the mushy zone. These effects have been addressed separate-ly, sometimes in a combined approach (e.g., shrinkage and deformation [14] or shrinkage andnatural convection [16]), but never in a comprehensive model. Second, each of this topic iscomplex. For example, deformation-induced segregation requires to couple stress-strain calcu-lation, including the mushy zone, with heat-, mass- and solute-transport in the liquid phase [14].Third, some of these phenomena are very localized. For segregation induced by natural convec-tion, the origin is the region where both the liquid velocity and the solute gradients are non-zero.It is limited to a region of maybe 1 mm thickness at most near the liquidus surface [16,17].

Figure 4: Macrosegregation results for a 2D section of an AA5182 DC cast slab, as calculated with the Cal-coSOFT software. The various pictures represent, from left to right: the steady state temperature profile, the iso-fractions of solid showing clearly the mushy zone position, the velocity field, the streamlines and the average Mg concentration [16].

156

Nonetheless, some interesting attempts to model macrosegregation have been made. For ex-ample, a result obtained in a 2D section of a DC cast slab of an AA5182 alloy is shown inFigure 4 [16]. It has been computed with the software CalcoSOFT, considering shrinkage andnatural convection. The main contribution to macrosegregation in this case was found to be dueto shrinkage: it induces a negative segregation zone near the centre of the ingot, the magnitudeof which corresponds to concentration profile measurements.

3.5 Coupled Heat Flow, Fluid Flow and Solid Deformation

As stated before, the thermal field induces stresses and strains in the coherent solid and convec-tion in the liquid region. Fluid flow transports also heat, while deformation of the solid modifiesthe thermal contact with the bottom block or with the mould. It is therefore necessary to couplethese aspects. One of the goals of the POST project [18] is to implement a mixed Lagrangian-Eulerian representation in the 3D code ProCAST [19]. Indeed, a significant difficulty which ari-ses when one wants to model the start-up phase of continuous casting, is the large change in sizeof the calculation domain in the direction of the casting velocity. This expansion cannot be ac-counted for by simply deforming the finite elements: such a method would lead to a severe dis-tortion of the elements with a damaging effect on the accuracy of the calculation. A solution tothis problem is provided by the use of a dynamic mesh, i.e. a mesh in which new elements areautomatically added during the course of a calculation. The strategy is therefore based on thetechnique of element activation. All the elements exist in the mesh from the beginning of thecalculation so that the total number of elements is constant all over the simulation, but the ele-ments can be active (regular) or inactive. Although present in the mesh, inactive elements areskipped during the assembly process and therefore have no influence on the calculation. A zonemade of a sufficient number of layers of inactive elements is defined in the initial mesh betweenthe liquid distribution system and the moving part of the casting. When the material begins tomove following the bottom block, the nodes in the expanding domain will move at the samespeed because of the Lagrangian nature of the description in this region. The layer of inactiveelements in contact with the active region is activated and the corresponding elements are stret-

Figure 5: Automatic generation of the accordion domain. The thickness of the elements is set here to a small value to make the elements visible. One layer of elements of the accordion domain is already stretched (POST project, [18]).

157

ched to account for the expansion of the domain. As soon as the thickness of this transition layerreaches a given value, the elements get fixed in size and the next layer of inactive elements isactivated. The inactive elements act as a reservoir which compensates for the expanding size ofthe domain by progressive stretching and activation. The behavior of this set of elements mi-mics the unfolding of an accordion: this is why it is called below the "accordion domain" (seeFigure 5). Periodic boundary conditions are used to ensure that the layers of inactive elementshave no effect on the calculation, in other words, to ensure the continuity of the solution fromone side of the accordion domain to the other side.

4 Microscopic Modelling

4.1 Grain Structure Formation

In continuous casting, grain structure formation has to consider the nucleation of grains, thefragmentation of dendrite arms and grain growth at a microscopic scale, the motion of the grainsand their thermal history at a macroscopic scale. All these phenomena are of course influencedby the temperature field and by fluid flow. Conversely, the presence of equiaxed grains modi-fies the apparent viscosity of the fluid and their movement also transports heat (and solute). Asthese phenomena are fairly complex, a semi-coupled approach can be made: the macroscopicaspects of the process are first calculated using appropriate average conservation equations (andthermo-physical data [20]). Then, the motion and growth of equiaxed grains are calculated as a“post-processing” module, using as input the local heat extraction and the velocity field compu-ted at the macroscopic scale.

For Cu based alloys, the influence of electromagnetic stirring (EMS) on grain refinement hasbeen studied in a Bridgman furnace by Campanella and co-workers [21]. Metallographic in-spection of the specimens, temperature measurements and numerical simulations performedwith CalcoSOFT [22] revealed that the efficiency of EMS is strongly dependent upon the pene-tration of the liquid in the mushy zone and therefore upon the position of the convection vorti-ces with respect to the liquid front. The results were analysed on the basis of a dendritefragmentation criterion similar to Fleming‘s criterion for local remelting of the mushy zone.

Figure 6: Schematic representation (left) of local fluid flow in the liquid, ul, and in between the dendrites, udl. The

projection of udl on the thermal gradient (parallel to the z-axis), ud

l,z, has to be compared with the velocity of the isotherms VT . On the right, schematics of remelting (1) and fragmentation (2) of secondary dendrite arms [21].

158

The new criterion tells that local remelting occurs when the component of the fluid flow veloci-ty along the thermal gradient becomes larger that the speed of the isotherms, as depicted in Fig-ure 6.

4.2 Microstructure Formation

The prediction of the macro- and microstructures obtained at the end of solidification is of greatinterest for aluminium producers. Indeed, the mechanical properties of aluminium alloys, butalso defects such as hot tearing, are strongly dependent upon these parameters. This is probablywhy the phase-field method has attracted the attention of so many scientists over the past deca-de [23–26]. Indeed, with a fixed mesh, such a model is capable in principle to predict the forma-tion of dendrites, eutectics, peritectics, etc ... It can directly use information coming from phasediagram calculations. Nevertheless, it has some inherent difficulties. In particular, the meshmust be fine enough (typically 0.1 micron) to correctly predict the curvature and the diffusionfield around a dendrite tip, thus making the time step of the explicit scheme small as well. Thismakes the calculations CPU intensive and limits the applications to domains of typically 1 mmin size or less, unless adaptive meshes are used.

4.3 Microporosity Formation

Although probably less important than hot tearing for DC casting, microporosity is the key de-fect in shape castings. While resulting from the same combined effects occurring within themushy zone, namely solidification shrinkage and gas segregation, this defect is usually classi-fied into shrinkage and gas porosity. Taking into account these two phenomena, a quite generalmodule has been developed within the software CalcoSOFT [27]. Darcy’s and mass conservati-on equations are solved only for the mushy zone, using a dynamic and regular finite volumegrid superimposed to the fixed finite element mesh used to solve the thermal exchanges withinthe casting. In this way, an accurate description of the shrinkage-induced pressure drop in themushy zone can be obtained. Appropriate boundary conditions have to be set at the limit of themushy zone, taking into account free liquid pockets (regions directly connected to ambient air),closed pockets (regions of liquid totally surrounded by the solid and/or mould) and semi-openregions of liquid (regions of liquid connected to an open region through the mushy zone). At thesame time, segregation of gases such as hydrogen is calculated in each volume element. Usingthe pressure and temperature fields, conditions for the nucleation and growth of pores are thendetermined, taking into account the important overpressure associated with the curvature contri-bution. Knowing the amount of microporosity formed in each volume element, macroporosityin hot spots and pipe shrinkage at free surfaces can be deduced simultaneously.

4.4 Hot Crack Formation

Modelling of hot cracking is probably one of the most challenging issues in solidification. Inde-ed, it involves thermal stresses/strains of the coherent solid, transmission of these deformationsto the partially coherent mush and interdendritic liquid feeding. For many years, very simple

159

criteria, such as the solidification interval, were used to simply address the composition-depen-dence of the susceptibility of an alloy to hot cracking [28]. In 1999, however, a new approach,based on Niyama’s criterion for microporosity formation [3] was derived at EPFL (so-calledRDG model after the names of the authors [29]). Assuming that the strain rate component per-pendicular to the thermal gradient at the roots of the dendrites is uniformly transmitted to thenon-coherent mushy zone, a simple mass balance combined with Darcy’s equation allows us tocalculate the pressure drop in the interdendritic liquid induced by these deformations and solidi-fication shrinkage. If this pressure falls below a cavitation pressure, a crack is assumed to form.Due to its simplicity, such a physically-sound approach can be easily implemented as a post-processing calculation of a deformation simulation [30]. In order to refine the RDG approachand be more rigorous in the handling of the solid and liquid interactions, the viscoplastic volu-me change (dilatation/densification) associated with the thermally-induced deformations in themush, and thus to the thermally-induced “opening-up” of the dendrite network have been re-cently included in a compressible model of the mushy zone [31–32]. A challenge in the area ofhot cracking, and to some extent of porosity formation, is the description of the gradual trans-formation of a continuous interdendritic liquid film to a continuous and coherent solid. In lowconcentration alloys, which are the most sensitive to hot cracking, this is achieved by coale-scence or bridging of the primary phase [33–35]. Eventually, the numerous parameters enteringthe compressible model describing the mechanical behaviour of the mushy zone have been de-termined not only for binary Al-Cu alloys [36] but also for industrial alloys, such as theAA5182 alloy [37].

5 Conclusion

Like DC casting, the development of models is a continuous process. It evolves with the powerof computer, with the increased knowledge gained from experimental observations, with the ad-vent of new numerical techniques (phase field, cellular automata, granular methods, etc.) andwith the development of new approaches (hot tearing, two phase method, etc …). Nowadays,macroscopic simulation tools are being routinely used by industry, despite their apparent highcost (hardware, software, training of people, maintenance, creation of data bases and expertisefor thermo-physical and other properties, validation, …). Yet, there is not a single companywhich could imagine not using such tools, since this is the only way to get an increased know-ledge of the process and therefore to reduce scrap, improve quality and diminish productioncosts. New sophisticated methods developed recently by academic partners for microstructureand defect formation will also make their way to industry in the future. This is the next logicalstep to characterize cast ingots from a metallurgical aspect, thus leading the way to propertiesprediction.The authors would like to acknowledge the scientific and financial support of the EU researchprojects EMPACT and VIRCAST and of the project POST.

6 References

[1] M. Rappaz: State of the art and new challenges in modelling of aluminium sheet ingot castings, in Hydro-Aluminium Bonn publications, 2004, p.14

160

[2] W. Kurz, D. Fisher, Fundamentals of Solidification (Trans. Tech., Aedermansdorf, Switzerland, 1998)

[3] E. Niyama, T. Uchida, M. Morikawa, S. Saito, A method of shrinkage prediction and its application to steel casting practice, AFS Int. Cast Metals J. (Septembre 1982) 52

[4] J.-M. Drezet, M. Rappaz, G.-U. Grün, M.Gremaud, Determination of thermophysical properties and boundary conditions of DC cast aluminum alloys using inverse method, Met. Trans. 31A (2000) 1627

[5] S.A. Argyropoulos, Measuring velocity in high-temperature liquid metals: a review, Scandin. J. Metallurgy 30 (2001) 273

[6] Dong Xu et al.: The Use of Particle Image Velocimetry in the Physical Modeling of Flow in EM or DC Casting of Aluminum: Part I. Development of the Physical Model, Mater. Trans 29B (1998)1281

[7] J. Dong Xu, W. Kinzy Jones, Jr., J.W. Evans, The Use of Particle Image Velocimetry in the Physical Modeling of Flow in Electromagnetic or Direct-Chill Casting of Aluminum: Results of the Physical Model Including Bag Geometry, Blockage and Nozzle Placement, Met. Trans 29B (1998) 1289

[8] A. Hellawell, J.A. Sarazin, R.S. Steube, Channel convection in partly solidified systems, Phil. Trans. R. Soc. Lond. 345A (1993) 507

[9] S. Henry, P. Jarry, M. Rappaz, <110> Dendrite Growth in Aluminum Feathery Grains, Met. Mater. Trans. 29A (1998) 2807

[10] S. Henry, G.-U. Gruen, M. Rappaz, Influence of Convection on Feathery Grain Forma-tion in Aluminum Alloys, Met. Mater. Trans. 35A, pp 2495–501, (2004)

[11] J.-M. Drezet, M. Rappaz, Modeling of Ingot Distorsion during Direct Chill Casting of Aluminum Alloys, Met. Mater. Trans. 27A (1996) 321

[12] W. Droste, J.-M. Drezet, G.-U. Grün, W. Schneider, 3D-Modeling of Ingot Geometry Development of DC-Cast Aluminum Ingots During the Start-Up Phase, in Continuous Casting, Eds: K. Ehrke and W. Schneider, (DGM, Wiley-VCH, Frankfurt, 2000) p. 175

[13] W. Droste, G.-U. Grün, W. Schneider, J.-M. Drezet, Thermo-mechanical modeling to predict shrinkage, shape and mold openings for dc-cast rolling ingots, Light Metals 2002, Ed. W. Schneider (TMS) p.703

[14] T. Kajitani, J.-M. Drezet, M. Rappaz, Numerical simulation of deformation-induced segregation in continuous casting of steel, Met. Mater. Trans. 32A (2001) 1479

[15] C. Beckermann, Modelling of Macrosegregation: Applications and Future Needs, Int. Materials Rev. 47 (2002) 243

[16] T. Jalanti, Etude et Modélisation de la Macroségrégation dans la Coulée Semi-Continue des Alliages d’Aluminium, PhD Thesis 2145, EPFL (2000)

[17] Th. Kaempfer, M. Rappaz, Modeling of Macrosegregation During Solidification Proces-ses Using an Adaptive Domain Decomposition Method, Model. Simul. Mater. Sc. Engng 11 (2003) 575

[18] POST Research Project, Calcom ESI, CH-1015 Lausanne, Switzerland[19] ProCAST User Manual, Version 2005, Revision 2.1, ESI Group, 75761 Paris, France[20] H. Combeau , B. Appolaire , B. Dussoubs and N. Houti: Prediction of the macro and

microstructures of solidification in DC casting of aluminium, Aluminium (2004, vol. 80), p. 614

161

[21] Th. Campanella, Ch. Charbon and M. Rappaz: Grain refinement induced by electro-magnetic stirring: a dendrite fragmentation criterion in Met. Trans. A. vol. 35 (2004), p. 3201

[22] CALCOSOFT User Manual, Version 2004, Revision 3.2, Calcom ESI, CH-1015, Lau-sanne, Switzerland

[23] J. A. Warren and W. J. Boettinger, Prediction of Dendritic Growth and Microsegregation Patterns in a Binary Alloy Using the Phase-Field Method, Acta mater. 43 (1995) 689

[24] J. Tiaden, B. Nestler, H. J. Diepers, I. Steinbach, The multiphase-field model with an inte-grated concept for modelling solute diffusion, Physica D 115 (1998) 736

[25] B. Nestler, A. A. Wheeler, A multi-phase-field model of eutectic and peritectic alloys: Numerical simulation of growth structure, Physica D 138 (2000) 114

[26] W. J. Boettinger, J. A. Warren, C. Beckermann, A. Karma, Phase-Field Simulations of Solidification, Ann. Rev. Mater. Res. 32, (2002) 163

[27] Ch. Pequet, M. Gremaud, M. Rappaz, Modeling of Microporosity, Macroporosity and Pipe Shrinkage Formation during the Solidification of Alloys using a Mushy-Zone Refine-ment Method, Met. Mater. Trans. 33A (2002) 2095

[28] J. Campbell, Casting, Ed. O. Butterworth-Heinemann. (1991)[29] M. Rappaz, J.-M. Drezet, M. Gremaud, A New Hot-Tearing Criterion, Met. Trans. 30A

(1999) 449[30] J.-M. Drezet, M. M'Hamdi, S. Benum, D. Mortensen, H. Fjær, Hot tearing during the

start-up phase of DC cast extrusion billets, in Aluminium Alloys, Eds. P.J. Gregson and S. J. Harris (2002) Part 1, p. 59

[31] C. L. Martin, M. Braccini, M. Suéry, Rheological behaviour of the mushy zone at small strains, Mat. Sc. Eng. A325 (2002) 293

[32] M. M'Hamdi, A. Mo and C. L. Martin, Two-phase modelling directed towards hot tearing formation in aluminium direct chill casting, in Met. Mater. Trans. Vol. 33A, July 2002, p. 2081

[33] M. Rappaz, A. Jacot and W. Boettinger Last stage solidification of alloys: a theoretical study of dendrite arm and grain coalescence, Met. Mater. Trans. 34A (2003), p.467

[34] V. Mathier, A. Jacot, M. Rappaz, Percolation of equiaxed grains during last stage solidi-fication, Model. Simul. Mater. Sc. Eng. 12 (2004) 1

[35] S. Vernède and M. Rappaz: Transition of the mushy zone from continuous liquid films to a coherent solid, accepted for publication in Philosophical Magazine 2005

[36] O. Ludwig, B. Commet, J.-M. Drezet, C. L. Martin, M. Suéry, Rheological behaviour of partially solidified Al-Cu alloys: Experimental and numerical study, in Modelling of Casting, Welding and Advanced Solidification Processes X, Ed. D. Stefanescu et al (TMS, Warrendale, USA, 2003) p. 183

[37] O. Ludwig, J.-M. Drezet, P. Ménès and M. Suéry, Rheological behaviour of a commercial AA5182 aluminium alloy during solidification presented at ICASP, Stockholm 2005, and submitted to Mat. Sciences and Eng. A

162

Numerical Simulation of the Growth of Interacting, Equiaxed Dendrites with a Two-Scale Model

M. Jurgk1, H. Emmerich

2, R. Siquieri

2

1Max-Planck-Institut fur Physik komplexer Systeme, Dresden

2RWTH Aachen

1 Motivation

Simulating the solidification of a stable phase at the expense of a metastable phase is one of theelaborate multiscale problems of computational materials science. Starting from atomisticconsiderations of atom attachment over dendritic growth dynamics of a single crystal to graingrowth and nally to the inuence on material properties of macroscopic cast metals the phenome-non of solidification spans 9 orders of magnitude in length scale. For an understanding of themultiscale nature of solidification paths models of dendritic growth at the scale of micrometershave played a central role. These models include the atomistic dynamics via transport and at-tachment coeficients. Moreover, they allow for a coupling to macroscopic transport dynamicsvia scaling relations for the growth velocity of the dendritic tip depending on system parameterslike e.g. the surface-tension anisotropy or the strength of the undercooling. Apart from the studyof dendritic growth in order to gain insight into the underlying, fundamental, physical mecha-nisms, one goal is to improve macroscopic application-oriented solidification simulations intheir quantitative power of prediction as well. The growth dynamics of a single isolated dendritehas already been studied and understood quite well and within this contribution we will presentnew scaling relations for dendritic growth which account for the interaction of several dendriticcrystals in a melt.

Figure 1: Left: The original problem posed by alloy solidification in a characteristic domain of sample. Right: Illustration of the homogenised two-scale model.

163

2 Model

If an alloy solidfiies and any flows in the melt are neglected, the controling transport mecha-nisms for the solidification process are the diffusion of heat and the diffusion of solute. The heatdiffusion in alloys is usually several orders of magnitude faster than the solute diffusion. Thisfact motivates the introduction of two different scales in the modeling of alloy solidification: amicroscopic scale for the solute diffusion and a meso/macroscopic scale for the heat diffusion.The movement of the solid-liquid interface of a dendritic crystal takes place on the microscopicscale of the slowly diffusing solute while the interaction between several of these crystals takesplace on the macroscopic scale. The initial situation is shown in the left part of Figure 1 and isas follows:

• Everything is in 2D. • Equiaxed dendritic nuclei are homogeneously distributed in the melt of a binary alloy.• The process is modeled by a sharp-interface model and the equations for both diffusion fields

T and c have to be solved in the same, whole domain, but only in the liquid phase (one-sided model).

The separation into two scales is mathematically done by the introduction of the scale param-eter , which species the ratio of the solute-diffusion scale to the heat-diffusion scale,

and for the derivation of the final equations shall satisfy the condition 1. The final homoge-nised two-scale model is achieved by means of asymptotic expansions in

plugging these expansions into the modeling equations and collecting all terms with up to thesecond order [1].

The equations of the two-scale model and the situation after the homogenisation procedure(right part of Figure 1) are as follows:

Now, there is the macroscopic, homogenised heat-diffusion equation

, (1)

c

T

ll

20 1 2

20 1 2

20 0

( , ) , , , , , , ...

( , ) , , , , , , ...

( , ) , , , , ...e

x x xT t x T t x T t x T t x

x x xc t x c t x c t x c t x

x xt x t x t x

0sL

T K Tt c

164

which we complement with a boundary condition of Biot type,

, (2)

and there is the microscopic growth problem for each nucleus:

(3)

(4)

. (5)

The macroscopic problem (1)–(2) and the microscopic problems (3)–(5) are coupled in bothdirections: The term L/c · t s in equation (1) expresses the latent heat which is released atsome point of the macroscopic temperature field due to the microscopic growth of the phase in-terface at this point which is modeled by the change in the local solid volume fraction s. In theother direction the coupling is visible by the term –b1 T in equation (4) which expresses the de-pendence of the microscopic phase-equilibrium concentration on the macroscopic temperature.

3 Numerical Implementation

We have modified and extended a simulation for the computation of the growth of a single den-drite under control of a single diffusion field [2]. The numerical setup is illustrated in Figure 2a:The fast diuffsing temperature field is defined on the points of a coarse grid (thick, full lines)while the slow concentration field is defined on the points of a fine grid (thin, dashed lines).One nucleus is located at each point of the coarse temperature grid. The important features ofthe numerical code are:

• Spatial and temporal discretisation is done with finite differences. • For the temporal integration of the diffusion equations the explicit first-order scheme (Euler)

is used. • The microstructure part: • The phase interface is a polygon line, which moves with higher accuracy than grid spacing,

because it is discretised independently of the diffusion-field grid. (see Figure 2b) • Because a single grid causes artificial numerical anisotropy, a stack of rotated grids (usually

4) is used for the discretisation of the microscopic diffusion field. (see Figure 2c) • The measured dendrites’ properties are the tips’ radii and growth velocities.

A BB

TBi T T

n

0cD c

t

2 1I eqb c c bTL

nI

cD c

n

165

4 Results

The scale parameter can be related to the nuclei density : 1 With the computer simu-lation we have investigated• the dependence of the tip-growth velocity tip on the scale parameter or on the inverse of

the nuclei density 1/ , respectively, and • the inuence of the Biot number Bi which characterises the heat transport out of the domain

on the growth-mode selection.

The 1/r-dependence of tip is shown in Figure 3. There appears a transition between twogrowth modes at crit = 1/ cri = 0.0011. The existence of these two solution branches can also beseen in analytical investigations [3].

4.1 Solution branch

A smaller density of crystal nuclei and thus a larger value of means a larger scale for theheat diffusion. As long as this scale is that large, that the transport of latent heat away from thenuclei can still be accomplished completely, the solid fraction of the nuclei grows considerablyin time during the evolution. As a consequence the term L / c t s is large and dominates theheat-diffusion equation (1). The respective analysis results in kinetic dendrites with a kinetic co-ecient depending strongly on the crystal density (see [3]). That is why the velocities of the solu-tion branch decrease steadily with increasing nuclei density.

4.2 Solution branch

The solution branch appears for higher nuclei densities and thus smaller values of . Due tothe smaller scale for the heat diffusion the transport of the arising latent heat is not complete andthe nuclei grow slowly. Hence the solid-fraction term L / c · t s does not dominate the heat-diffusion equation (1) and consequently in the analysis the temperature term in the Gibbs-Thomson relation (4) does not depend on (like it does in the case). This is why the solid-frac-tion term and thus the scale parameter or the nuclei density , respectively, does not determine

1n

2n

2n

a) b) c)Figure 2: Illustrations of the numerical setup

166

the growth rate of the nuclei for this solution branch and the growth follows the laws of capilla-ry dendrites.

4.3 Biot Number

The analytic investigations show an inuence of the Biot number Bi on the critical value crit

where the transition between the solution branches appears [3]. In Figure 4 the -dependence of

Figure 3: The new numerically obtained scaling relation with two solution branches

Figure 4: The inuence of the Biot number on the scaling relation

167

the tips’ growth velocity tip is plotted for the two Biot numbers Bi = 1 and Bi = 10. The transiti-on is shifted to larger values of for a smaller Biot number.

5 Outlook

The directions of our current and future work are

• to check if these new scaling relations improve the quantitative validity of multiscale approaches like e.g. in [4] by replacing the empirical metallurgical relations,

• to study the two-scale model for other growth morphologies like peritectic growth, • to include melt ow as a transport mechanism in the model, and • to incorporate a phasefield model for the extension to 3D.

6 References

[1] C. Eck, P. Knabner, S. Korotov, J. Comp. Phys. 2002, 178, 58 [2] T. Ihle, H. Müller-Krumbhaar, Phys. Rev. E 1994, 49, 2972 [3] H. Emmerich, M. Jurgk, R. Siquieri, Phys. Stat. Sol. (B) 2004, 241, 2128 [4] C.Y. Wang, C. Beckermann, Metall. Mater. Trans. A 1993, 24A, 2787

168

Monte Carlo Simulation of Grain Growth in Three Dimensions

D. Zöllner, P. StreitenbergerOtto-von-Guericke-Universität Magdeburg, Fakultät für Naturwissenschaften, PF 4120, D-39016 Magdeburg

1 Abstract

A Monte Carlo Potts model algorithm for single-phase normal grain growth is presented, whichallows one to simulate the development of the microstructure of very large grain ensembles inthree dimensions. The emphasis of the present work lies on the investigation of the relaxationprocess. Different initial grain structures characterized by different initial size distributions aresubjected to grain growth via Monte Carlo simulation. The temporal development of 3D grainstructures reaches independent of the initial size distribution, after an initial period of time, aquasi-stationary self-similar coarsening regime where all scaled size distribution functions col-lapse to the single universal, time independent size distribution f(x). The relaxation process tothis universal self-similar state is studied by following the temporal development of quantitieslike the average grain size, the standard deviation of the grain size distribution and topologicalcorrelations.

2 Introduction

One method to simulate grain growth is the Monte Carlo (MC) simulation based on a numericalrealization of the Potts model [1]. In 1984 Anderson et. al. for the first time introduced the Pottsmodel to simulate the grain growth kinetics and to investigate grain size distribution and topolo-gy in two dimensions [2, 3].

Especially the last years opened up growing possibilities of MC investigations of graingrowth in three dimensions (cf. e.g. [4, 5, 6, 7] and the literature within).

Figure 1: Two shots of 3D grain growth (left: 0th MCS; right: 500th MCS) taken from a simulation run; the grains are shaded for a better presentation

169

The aim of the present work is the investigation of normal grain growth by studying the tem-poral development of quantities like the average grain size, the standard deviation of the grainsize distribution, the microstructure and topological correlations. It is shown that the simulatedself-similar volumetric rate of change of all grains can be fitted to an average scaled growth lawthat allows us to model the grain growth behavior by an analytic mean-field theory.

3 3D Monte Carlo Simulation

3.1 Monte Carlo Potts Model Algorithm

A Monte Carlo algorithm for single-phase normal grain growth in three dimensions has beenimplemented. Therefore the microstructure is mapped on a cubic lattice with 26 nearest neigh-bors and periodic boundary conditions. Each lattice point is called one Monte Carlo Unit(MCU). Each MCU has assigned an integer value representing the orientation Q of the latticepoint. While in earlier simulations the size of the 3D lattice was usually limited to 100 100 100MCU’s and therefore, those simulations were terminated to small grains we choose in the pre-sent 3D simulations the number of MCU’s up to 250 250 250.

The time unit of the simulation is called Monte Carlo Step (MCS) and is defined as N reori-entation attempts where N is equal to the number of MCU’s. That means, in our present 3D sim-ulation N = 15625000. Further details can be found in the literature [2–7, 8].

Each initial structure was subjected to grain growth 10 times for 1000 MCS. For the analysisof the simulation the data of each step were summed up over the 10 runs.

3.2 Growth Law, Scaling Regime and Size Distribution

The coarsening process (Fig. 1) has been investigated by following the temporal developmentof an initial Poisson Voronoi Tessellation and an initially rayleigh distributed grain structure.

Figure 2: Temporal development of the mean grain radius together with the fit of the growth law (Eq. 1): I – initial period; II – self-similar coarsening regime

170

Both 3D grain structures as they have been simulated by the MC procedure follow, after an in-itial period of time, the well-known growth law for the average grain radius (cf. Fig. 2)

, (1)

where the growth exponent of the initially rayleigh distributed grain structure is numerically gi-ven by = 0.4998. The constant b is given by b = 0.4322 corresponding to a growth parameter

k is the kinetic constant of the simulation.After the initial period the grain size distribution function (GSDF), F(R, t), is self-similar, i.e., itis characterized by the scaling law

. (2)

g(t) is a time-dependent factor and f(x) is the normalized GSDF with x = R / <R>, where R is de-fined as the radius of an volume equivalent sphere and <R> is the average grain radius. AllGSDF within the quasi-stationary self-similar (QSSS) coarsening regime collapse to the singleuniversal, time independent size distribution f(x) (cf. Fig. 3).

R b t const

12 6543k/ R R .

F(R,t) g(t) f(x)

Figure 3: Temporal development of the GSDF of the initial Voronoi Tessellation for different time steps (stars:

Figure 4: Comparison of the temporal development of the standard deviation of both initial grain structures

171

The temporal development of the GSDF is quantitatively shown through the development of

the standard deviation sd, which is defined by and .

In the QSSS coarsening regime sd stays nearly constant, while the temporal development of sd towards this state depends on the initial GSDF (Fig. 4).

3.3 Microstructure and Topological Correlations

In order to investigate the coarsening process of the microstructure to the QSSS state the tempo-ral development of the number of neighboring grains s of grains with relative size x has beenobserved. In Figure 5 s vs. x is shown for the initial state (Voronoi Tessellation) and the QSSSstate after 500 MCS.

The initial Voronoi Tessellation shows a certain correlation between the number of neigh-boring grains and their size in so far as larger grains have more neighbors than smaller grains(cf. Fig. 5a). During relaxation this size correlation becomes stronger and, within the QSSSstate, can be described by a self-similar quadratic function (Fig. 5b).

Because of the changes in the topology the mean number of neighboring grains or faces <F>changes with time. For the Voronoi Tessellation <F> has a value of <F> = 16.7993, while theinitially rayleigh distributed grain structure gives <F> = 13.55. During the simulation both val-ues change and yield within the QSSS state <F> = 13.9292 respectively 13.9339. The value<F> 13.93 is close to 14 which is the value of the mean number of faces of Kelvin’s model [9].Lord Kelvin’s model describes a possibility to fill a 3D space completely with objects of thesame size. These so called tetrakaidecahedron have each 14 faces (eight hexagonal and 6 quad-ratic faces) which are slightly curved.

3.4 Average Scaled Growth Law and Mean-Field Theory

Aside from the initial period of time the grain growth can be characterized by the volumetricrate of change , which depends only on the scaled grain size x and therefore is time

2 2sd x x n nx x f(x)dx

1/ 3 ~V V RR

Figure 5: Number of neighboring grains s of grains with size x (a – initial Voronoi Tessellation, b – QSSS state: in terms of the residual sum of squares (= sum of all squared residuals, abbreviated as rss) a quadratic relationship relating s to x yields clearly a better approximation to the simulation results than the linear fit)

a b

172

independent and self-similar [10, 11]. In terms of mean-squared errors a quadratic function (cf. Fig. 6a) gives the best approximation to the simulation results.

A mean-field theory based on a quadratic form of vs. x and the Lifshitz-Slyozov-Hillertstability conditions has been given in [8, 12, 13].

In this theory the size distribution function is given by the one-parameter family

, (3)

where a = a(x0) is defined by the scaling requirement <x> = 1 [12, 13]. The upper cut-off x0

serves as a free adjustable parameter. The solid line in Fig. 6a shows the relation as itfollows from a least-square fit of the theoretical expression [8]

(4)

to the simulation data of , yielding x0 = 4.1275 and consequently a(x0) = 16.795. The fitgives a good average description of the simulation results of vs. x. The resulting analyticGSDF Eq. 3 is in very good agreement with the simulated size distribution (cf. Fig. 6b).

For x0 = 9/4 and consequently a(x0) = 3 Eq. 4 reduces to a linear function and Hillerts GSDFresults [14], for Eq. 3 yields Louats GSDF [15], which are both clearly in contradictionto the present simulation results (cf. Fig. 6b). It is the relatively small quadratic term in Eq. 4that is responsible for the excellent agreement between the simulated and analytical GSDF inFig. 6b.

( )RR x

RR

00 02

00

( ) exp , ( )

a aa

x a xf x a x e x x

x xx x

( )RR x

22 0 0

0 0 0

3 6 3( ) 1( ) ( ) ( )

k x xRR x x x

a x a x a x

RRRR

0x

Figure 6: a – Average scaled growth law (stars: simulation data; line: Eq. 4); b – Simulated size distribution (stars)compared to analytic functions

173

4 Conclusions

In this paper MC simulation results on normal grain growth in three dimensions have been pre-sented. After a short initial period of time a quasi-stationary state of the evolving grain structureis reached, which is characterized by a parabolic grain growth kinetics and a self-similar GSDFthat is nearly time invariant. The local topology of the simulated grain microstructure can becharacterized by a quadratic size correlation function.

From the MC simulation results of 3D grain growth the average scaled growth law as a func-tion of the scaled grain size was determined. The analytic GSDF based on a mean-field descrip-tion of the simulated microstructure shows a very good agreement with the GSDF resultingfrom the 3D MC simulation.

5 Acknowledgements

The authors would like to thank the Professors Martin E. Glicksman, Günther Gottstein and Di-mitri Molodov for useful discussions and the Deutsche Forschungsgemeinschaft for financialsupport under GKMM 828.

6 References

[1] R.B. Potts, Proceedings of the Cambridge Philosophical Society Vol. 48, 1952, 106–109[2] M.P. Anderson, D.J. Srolovitz, G.S. Grest, P.S. Sahni, Acta Metallurgica Vol. 32, 1984,

783–791[3] D.J. Srolovitz, M.P. Anderson, P.S. Sahni, G.S. Grest, Acta Metallurgica Vol. 32, 1984,

793–802[4] P. Anderson, G.S. Grest, D.J. Srolovitz, Philosophical Magazine B59, 1988, 293–329[5] X. Song, G. Liu, Scripta Materialia Vol. 38, 1998, 1691–1696[6] Q. Yu, S.K. Esche, Materials Letters Vol. 57, 2003, 4622–4626[7] A. D. Rollett, P. Manohar, in Continuum Scale Simulation of Engineering Materials, edi-

ted by Dierk Raabe et.al., Wiley-VCH, Weinheim, 2004, 77–114[8] D. Zöllner, P. Streitenberger, in Proceedings of the 2nd Joint International Conference on

Recrystallization and Grain Growth, Annecy, France, TTP, 2004, 1129–1136[9] The Kelvin Problem, edited by D. Weaire, Taylor & Francis, London, GB, 1996[10] W.W. Mullins, Journal of Applied Physics Vol. 59, 1986, 1341–1349[11] M.E. Glicksman, Philosophical Magazine Vol. 85, 2005, 3–31[12] P. Streitenberger: Scripta Materialia Vol. 39, 1998, 1719–1724[13] P. Streitenberger: in Proceedings of the 1st Joint International Conference on Recrystalli-

zation and Grain Growth, Aachen, Germany, Springer Verlag, 2001, 257–262[14] M. Hillert, Acta Metallurgica Vol. 13, 1965, 227–238[15] N.P. Louat, Acta Metallurgica. Vol. 22, 1974, 721–724

174

Thermal Conductivity of Ternary and Multi-Component Aluminum Alloys up to and above the Melting Temperature

R. Brandt1, W. Bender2, G.-U. Grün2, G. Neuer1

1Institute for Nuclear Technology and Energy Systems (IKE), University of Stuttgart, 70550 Stuttgart, Germany, E-mail: guenther.neuer@ike.uni-stuttgart.de2 Hydro Aluminium GmbH, Bonn, Germany

1 Abstract

The determination of thermophysical properties in a broader temperature range is an essentialprerequisite for proper process optimization by numerical modeling techniques, which are nowcommonly used in the industry. Here, results of a measurement campaign on various ternary Al-SiMg and multi-component AlSiCu+ alloys using a new electric resistivity measurement unitare presented. The evaluated temperature dependent data are discussed regarding alloying ele-ment and equipment related influences.

2 Introduction

Thermophysical properties of metals in the melting range are crucial to optimize processes e.g.powder production by atomizing or casting by means of numerical models simulating solidifi-cation. Existing calculation programs model both the bulk heat transfer and the fluid dynamicprocess during solidification aiming to predict e.g. time dependent temperature profiles in thevarious states of the material: liquid, “mushy region”, solid, determine solidification rates etc.These numerical models are important tools for the better understanding of the physics of solid-ification and to improve quality of products, however, they are extremely sensitive against ther-mophysical properties. The demand and the availability of these properties is summarized byLudwig, Quested, and Neuer [1]. Beside viscosity and surface tension the knowledge of thethermal conductivity is of particular importance.

As the thermal conductivity is strongly related to microstructure of the solid material the de-termination during solidification is a key aspect, because the development of the microstructurehighly depends on the cooling rate. Therefore the measurement technique should enable tomeasure the thermal conductivity in the liquid state, in the “mushy” region, and in the solid statebefore and after melting. Only very limited thermal conductivity results of aluminum alloys arepublished till now [2,3,4,5,6,7]. Only one investigation of the thermal conductivity of aluminumalloys with systematic variation of individual alloy components has been found [8]. The Si con-tent of binary AlSi alloys had been varied between 0 - 3 -7 and 11 %Si.

In the frame of the European Network Program "Microstructural Engineering by Solidifica-tion Processing" (MEBSP) a number of aluminum based alloys has been produced by HydroAluminium Deutschland GmbH. Specimens of the binary system AlSi, of the ternary systemAlSiMg, both produced out of pure material basis, and of a AlSiCu+ alloy, representing a com-mercial multi-component system with fixed values for all other alloying elements but varied in

175

Si and Cu content, have been distributed to various laboratories in order to measure the density,surface tension, viscosity, thermal conductivity, and electric resistivity as functions of tempera-ture and of chemical composition. All samples were presented in the as-cast condition.The com-plete list of all 21 different materials has been presented by Brandt and Neuer [9] together withthe measured densities at room temperature and the electric resistivities measured at 25 °C,50 °C, and 75 °C. Thermal conductivity results of the AlSi alloys are published by Brandt andNeuer [10] at temperatures up to 800 °C. The data evaluated for the ternary and multi-compo-nent systems shall be discussed here.

3 Measurement Technique

The electric resistivity was measured by using the four probe technique, whereby a cylindricalsample of diameter D is loaded by a direct current I. The electric resistivity is determined bymeasuring the voltage drop U between two probes positioned at a distance of S using the fol-lowing equation:

(1)

Details of the measurement technique are described in [9,10]. For high accuracy measure-ments in the solid state the specimen is mounted on a vernier caliper with electrical insulated ra-zor blades on the flanks. This allows measurements with variable distance S. Specimens of 3mm to 10 mm in diameter and 100 mm to 200 mm length can be used. The whole arrangementis dipped in an oil bath to get a well defined and homogeneous temperature distribution. Thussteady state measurements up to about 100 °C are possible.

Measurements in the molten state of a metal are much more difficult than measurementsonly on solids because the sample must be kept inside of a ceramic tube. This involves that geo-metric data and sample temperature are more difficult to be determined and the complicate me-chanics of the electrodes may lead to increased measurement errors. Solutions of such problemsare described in [10]. For the investigation of the aluminum alloys described here specimen rodsof 5 mm in diameter and 100 mm long were inserted in a tubular crucible of aluminum nitridewhich was mounted in its vertical position in a vacuum chamber. Here the distance between thevoltage probes was fixed to S = 50 mm. The vacuum chamber is surrounded by a tube furnace,which enables measurements in vacuum or in each available gas atmosphere at temperatures be-tween 30 °C and 1600 °C. The apparatus is described in more detail in [10].

The measurements were carried out between room temperature and about 800 °C using aconstant electrical current of 2 amps. Higher currents give better voltage drop signals but thehigh contact resistance between current electrodes and specimen causes a rapid heating of thesample edges. All measurements were performed under high vacuum conditions. A heating rateof 360 K/h was applied in the solid and in the liquid state with interrupts before each measuringpoint, during melting and solidification the heating/cooling rate was reduced to about 100 K/hin order to get better resolution for the melting/solidification temperatures. The thermal conduc-tivity was calculated from the measured electric resistivity by using the Wiedemann-Franz-Lorenz law / = L T, which describes the relationship between thermal conductivity andelectrical conductivity = 1/ for metals. L is the Lorenz number, which depends on the type

2

4

U D

I S

176

of metal and also to some extent to its temperature, and T is the absolute temperature in Kelvin.For aluminum and its alloys it has been shown that by using the theoretical Sommerfeld valueof L0 = 2.445 * 10-8 V2/K2 the calculated conductivity data may be too high by about 13 % atroom temperature, but with increasing temperature this deviation becomes smaller, and above200 °C the deviation is lower than 2 % for the studied aluminum alloys [3, 10].

4 Measurements and Results

Table 1 and Table 2 compile together the sample designations, related chemical composition,density, and electric resistivity S at solidus and L at liquidus temperatures both during heatingand cooling cycles. A complete list of the electric resistivity measurements of all samples is pre-sented in [10]. On the basis of these data the thermal conductivity values discussed below arethen calculated as described in the section above.

Table 1: Sample designations and some characteristic values of the investigated ALSiCu+ alloys

The tables show that during melting the electric resistivity increases by a factor L/ S of

about 2.2–2.4 for all measured AlSiCu+ and AlSiMg alloys. This agrees with the values forpure aluminum (2.21) and binary AlSi alloys (2.3–2.4) measured before [10].

Fig. 1 exemplarily shows the resulting thermal conductivity development during heating andcooling for a AlSiCu+ alloy with 9 wt% Si and varying Cu contents, while in Fig. 2 the valuesfor fixed 3 wt% Cu and variable Si contents are displayed. It is evident that, compared to the bi-nary AlSi, an addition of 1% Cu lowers the conductivity considerably, 2% Cu causes a light fur-ther decrease, while an addition of 3%Cu has nearly no further influence on thermalconductivity. In the case of variable Si-content (Fig. 2) the conductivity decreases continuouslywith increasing Si-content.

sample No. AlSi-13 AlSi-15 AlSi-16 AlSi-17 AlSi-18 AlSi-19 AlSi-21

Si [%] 6 6 9 9 9 11 11 Cu [%] 1 3 1 2 3 1 3 others [%] 2.5 – 3 density [g/cm3] 2,748 2,783 2,730 2,747 2,765 2,710 2,755

heating cycle

Solidus 563 °C 538 °C 545 °C 550 °C 536 °C 550 °C 531 °C S [ *cm] 13,17 13,38 13,50 13,78 14,39 14,19 14,03 Liquidus 626 °C 611 °C 595 °C 602 °C 598 °C 580 °C 580 °C L [ *cm] 31,03 31,27 32,16 32,67 32,84 33,10 33,36 L/ S 2,36 2,34 2,38 2,37 2,28 2,33 2,38

cooling cycle

Liquidus 616 °C 603 °C 597 °C 593 °C 586 °C 575 °C 568 °C L [ *cm] 30,71 31,29 32,40 32,34 32,54 33,42 33,19 Solidus 541 °C 530 °C 537 °C 536 °C 533 °C 543 °C 539 °C S [ *cm] 16,59 18,22 17,89 18,06 18,71 19,59 19,06 L/ S 1,85 1,72 1,81 1,79 1,74 1,71 1,74

177

Table 2: Sample designations and some characteristic values of the investigated ALSiMg alloys

A common factor in both figures is the fact that the values for the solid sample in the coolingcycle after solidification are remarkably lower than those values measured for these temperaturelevels during the heating cycle. In order to better understand the reason of this behavior, addi-tional measurements at room temperature with the more accurate vernier caliper technique (VC)were carried out for the already processed sample AlSi-6 and compared with corresponding cal-ibration measurements before the high temperature measurements.

sample No. AlSi-6 AlSi-8 AlSi-10 AlSi-11 AlSi-12Si [%] 5 7 9 12 12Mg [%] 0,6 0,6 0,6 0,3 0,6density [g/cm3] 2,669 2,666 2,659 2,640 2,642

heating cycle

Solidus 567 °C 569 °C 567 °C 569 °C 580 °CS [ *cm] 11,71 12,75 13,08 13,73 14,96

Liquidus 634 °C 621 °C 599 °C 609 °C 590 °CL [ *cm] 27,79 28,83 29,05 31,61 31,55L/ S 2,37 2,26 2,22 2,30 2,11

cooling cycle

Liquidus 624 °C 609 °C 596 °C 576 °C 579 °CL [ *cm] 27,51 28,63 29,70 31,37 31,15

Solidus 538 °C 542 °C 549 °C 548 °C 545 °CS [ *cm] 12,78 13,98 15,02 18,23 16,98L/ S 2,15 2,05 1,98 1,72 1,83

Figure 1: Thermal conductivity of AlSi9Cu+ with varying Cu-content, calculated from electric resistivity mea-surements for heating and cooling cycles

178

This is shown in Fig. 3. Here, the resulting values of measurements at room temperature ofthe re-solidified rod for different distances and locations are compared with the results from thehigh temperature measurements. As it can be seen the value for the VC (S = 50 mm), whichmeans the vernier caliper measurement over the same distance as during high temperaturemeasurement nicely correlates with the final value of the high temperature measurement. But,the two values at a zone of 30 mm from the lower end (S = 30 mm bot.) and from the upper end(S = 30 mm top) demonstrate that there is a thermal conductivity gradient along the length ofthe rod after the resolidification.

This inhomogeneity in the results could have a variety of reasons. A gradual change inmicrostructure including segregation processes due to the solidification conditions could be apossible explanation. But also the development of shrinkage porosity inside the sample causedby hydrogen segregation or insufficient volumetric compensation within the probe crucible dur-ing the end of the cooling phase would result in such strong gradual differences.

For further clarification regarding the possible cause the processed sample rod AlSi-15 wascut along the longitudinal symmetry plane and porosity conditions and microstructure wereevaluated.

Fig. 5 clearly visualizes the most probable cause for the inhomogeneity. Quite obvious arethe large differences in porosity along the rod. Most probably due to gravity the upper part ofthe sample shows large pores, which in the cross section area nearly seem to separate both rodparts, although the sample was still intact when removed from the ceramic tube. So, still electri-cal contact existed throughout the duration of the experiment. In the section taken of the midpart of the sample the porosity is definitely lowest, while in the lower part a slight increase oc-curs again.

Figure 2: Thermal conductivity of AlSiCu3+ with varying Si-content, calculated from electric resistivity measu-rements for heating and cooling cycles

179

Figure 4: Thermal conductivity for AlSiMg alloys with varying Si-content, calculated from electric resistivity measurements for heating and cooling cycles

Figure 3: Thermal conductivity for sample AlSi-15 during heating and cooling phase. Single results on the left are conductivities before (run 1) and after (run 3) resolidification, measured with vernier caliper (VC) at different sections of sample.

180

Parallel evaluation of the microstructure revealed a more or less uniform structure along thewhole rod with no significant differences for an explanation of the variations in the thermal con-ductivity as shown in Fig. 3. Therefore, the most probable reason for the differences in the resis-tivity during heating and cooling cycle is the evolving porosity inside the sample during thesolidification phase of the measurement sequence.

This is supported by the results of the ternary AlSiMg alloy variants. As can be seen in Fig.4, the temperature dependent thermal conductivity for the 0.6 wt% Mg and variable Si contentsduring heating and cooling shows minor differences between the results achieved in both cy-cles. This coincides nicely with the related microstructure investigation. Fig. 6 displays in acomparable way to Fig. 5 the structure along the longitudinal cross section of the sample AlSi-6. It is clearly visible that here less porosity has developed and therefore a lower difference be-tween heating and cooling cycle should be reasonable.

Coming to an overall evaluation of the measurement campaign, the systematic investigationof temperature dependent thermal conductivity has led to the following general results:

In case of AlSiCu+ alloy material the conductivity does not show dramatic changes for fixedSi values with growing Cu content, except for the step from the binary to the multi-component

Figure 5: Microstructure of the multicomponent sample AlSi-15 (6% Si, 3% Cu, 3% others) at upper (top), middle and lower (bottom) position of longitudinal symmetry plane

Figure 6: Microstructure of the ternary sample AlSi-6 (5% Si, 0.6% Mg) at upper (top), middle and lower (bot-tom) position of longitudinal symmetry plane

181

variants. Typical for this material is the rather significant increase in conductivity from roomtemperature to approx. 300 °C in the heating cycle, which is less visible during cooling. Thisdoes not occur in case of zero copper content, which means the binary AlSi9 alloy sample.

• For the AlSiMg0.6 alloy variants a slight increase of the conductivity from room temperature to 300 °C only occurs for low Si values, for higher Si additions a continuous decrease is measured.

• For AlSiCu+ and AlSiMg alloys the increase of Si always leads to a decrease in conductivity level for both, a fixed Cu and Mg content.

• In those cases where strong deviations between heating and cooling cycle of the measure-ment could be linked to the occurrence of significant porosity material data only derived from the heating cycle should be introduced into simulation data bases.

5 Acknowledgements

The authors would like to thank the Deutsche Forschungsgemeinschaft DFG, Bonn, Germanyfor the financial support of the work presented here.

6 References

[1] A. Ludwig, P. Quested, G. Neuer, Advanced Engineering Materials 3, 2001, 11–14[2] J. Auchet, J. L. Bretonnet, Rev.Int.Hautes Temper.Refract. 26, 1990,181–192[3] L. Binkele, M. Brunen, Thermal conductivity, Electrical Resistivity and Lorenz Function

Data for Metallic Elements in the Range 273 to 1500 K, Report No. JÜL-3006, Forsc-hungszentrum Jülich, 1994

[4] R. E. Taylor, H. Groot, et al., High Temp.-High Press. 30, 1998, 269–275[5] H. Szelagowski, R. Taylor, High Temp.-High Press. 30, 1998, 343–350[6] Overfelt R A, Bakhtiyarov S I, Taylor R E, 2002, High Temp.-High Press. 34 401–409[7] J. Blumm, J. B. Henderson, L. Hagemann, High Temp.-High Press. 30, 1998,153–157[8] M. Rappaz et al., in: Modelling of Casting, Welding and Advanced Solidification Process

VII, Eds. M. Cross and J. Campbell; The Minerals, Metals & Materials Society, 1995, 449–457

[9] R. Brandt, G. Neuer, Advanced Engineering Materials 5 No. 1–2, 2003, 52–55[10] R. Brandt, G. Neuer, presented at the ECTP 2005 (http://www.ectp.sav.sk/) to be pub-

lished in High Temp.-High Press

182

FEM Simulation of Near Net Shape DC Billet of High Strength Al-Mg-Si Alloy

H. Nagaumi1, Y. Takeda1, Suvanchai P.2, T. Umeda2

1Nippon Light Metal Company Ltd., Kambara, Japan 2Chulalongkorn University, Bangkok, Thailand

1 Introduction

Aluminum casting and forging products are used for aircraft, vehicles, industrial machinery,precision machine, etc. Lighter automobiles are increasingly demanded in order to improve fuelefficiency and drivability. Many of suspension components require superior and reliable proper-ties such as high strength and toughness. Aluminum forging, with its high strength and ductility,is a promising alternative for such suspension parts since it can result in weight reduction of upto 40 % (compared with conventional steel parts). The high deformability of aluminum forgedparts is also highly suitable for maximum safety. But the higher cost of aluminum forgings has–up until now- been a barrier to fully capture the growing market for these applications.

Generally, in case of the aluminum suspension that is forged using the extrusion round bar, itleads to make the manufacturing cost rise because the homogenization, pre-heating and an ex-trusion processes are necessary. Recently a new DC casting process of near net shaped ingot(beam blank ingot), which aimed to reduce the manufacturing cost of suspension, was proposed[1, 2]. Since the shape of the near net shaped billet is complicated than usual round billet, manytroubles in the casting are originated from the ingot shape. For example, the peculiar problemssuch as the surface crack and the mold grasped by solidifying billet, which led to the interrup-tion of casting, resulted from the complicate shape and casting conditions. The mechanism ofthese peculiar phenomena had not yet been clarified. In this study, a coupled flow, solidificationand thermal deformation FEM analysis and casting experiments of near net shaped semi-contin-uously cast billet were carried out in order to clarify the peculiar problems. By comparing thecalculated results with the experimental ones, the cause of the surface crack and the moldgrasped by the solidifying billet during was clarified.

2 Casting Experiment

The material used in this study is the high-strength Al-Mg-Si alloy that was improved fromAA6061, and the chemical composition of the alloy is shown in Table1. The alloy was cast by alab-scale DC casting machine using the hot-top method. The schematic shape of mold used inthis experiment is shown in Figure1, and the casting conditions were shown as follows; Castingtemperature of 720 °C, Casting speed of 85–110 mm/min, Water supply of 150–200 l/min.

183

3 Analytical Method

3.1 Flow and Solidification Analysis

In this analysis, the followings were carried out; a coupled flow, solidification and thermal de-formation FEM analysis. The geometry used in the study of flow and solidification analysis isas shown in Figure 1. The calculation domain includes billet, mold and bottom block, where thecasting length is 500 mm. The flow and solidification analysis was calculated using a commer-cial solidification package, CAPFLOW. Enthalpy method was used to analyze the latent heatevolution of solidification. Considering the temperature dependency of density, specific heatand thermal conductivity, measured values, described later, were used. The physical propertiesused in the calculation are shown in Table 2. The thermal boundary conditions at the interfacebetween billet/water-cooled mold, billet/cooling water and billet/bottom block are prescribedusing the Newton’s law of cooling assuming h, heat transfer coefficient, as a function of tempe-rature as shown in Table 3.

Table 1: Chemical composition of the Al-Mg-Si alloy (in wt %) Element wt %

Table 2: Physical properties used in the calculation

Si Fe Cu Mg Mn Cr Ti Al

1.00 0.18 0.40 0.83 0.37 0.27 0.02 Bal.

Billet Bottom Block Mold

Al-Mg-Si A6061 Cu Density, kg/m3 2680 2700 8000Specific Heat, J/(kgK) See Figure 2 1080 385 Thermal Conductivity, W/(m/K) See Figure 3 150 394

Figure 1: Schematic geometry and boundary conditions

184

Table 3: Relation of heat transfer coefficient and billet surface temperature

h1: Heat transfer coefficient between billet and mouldh2: Heat transfer coefficient between billet and bottom block in the center of the billet. h3: Heat transfer coefficient from billet and bottom block to the intruded water in the gap

3.2 Thermal Stress Analysis

In this study, thermal stress during casting was calculated by a commercial structure analysispackage, ANSYS. The thermal histories obtained from the CAPFLOW were used as input datato an elasto-plasticity model which simulated the thermal stress and distortion of the billet. Thecalculation domain includes the billet, mold and bottom block, where mold and bottom blockare assumed to be as a rigid body and do not consider their thermal distortion. In order to simu-late the thermal distortion of the billet the element type was employed for a large deformationelement called as SOLID 45. The effect of friction forces which happened between the billetand the bottom block and also between the billet and the mold, was considered by using thecontact element called as SURFACE CONTACT 169 and 171, and the friction coefficient wasset to 0.1. A multi-linear isotropic work-hardening law was determined to fit the experimentaldata. Von Mises criterion law determined the yield criterion. Young’s modulus, linear expansi-on coefficient and the work-hardening were considered as a function of the temperature and the-se values were obtained by experiment as shown later. In the solid-liquid coexistence region, anaverage linear expansion coefficient was estimated from a volumetric shrinkage percentage of6%.

4 Thermal Properties

Figure 2 shows the specific heat of this alloy was continuously measured from room tempera-ture to 700

º

C. Solidus temperature (580 ºC), liquidus temperature (554 ºC) and the equilibriumsolidification temperature range (74 ºC) of the alloy were obtained by specific heat curve [3].The relation between fraction solid and temperature calculated by Gulliver-Scheil model usingThermo-Calc and specific heat curve is also shown in Figure 2. The solidification temperaturesfrom liquidus temperature down to solid fraction of 0.8 are almost same for the two methods,but the final stage of solidification is quite different. In this study, the Gulliver-Scheil modelwas used to evaluate solidification and distortion behaviors. The linear expansion coefficientwas continuously measured from room temperature to 600 °C using the vertical type of thermalexpansion apparatus. Density at each temperature was calculated from these data. The thermaldiffusivity was measured at a specified temperature using the laser flash method. Figure 3

T,ºC h1, W/m2 K T,ºC h2, W/m2 K T,ºC h3, W/m2 K

0 20 200 2000 50 3000 578 20 300 500 100 10000 580 1000 400 500 150 20000 650 2000 582 10000 200 600 700 2000 660 20000 500 4000

700 4000

185

shows the thermal conductivity which is determined from density, specific heat and thermal dif-fusivity, and the linear expansion coefficient, as a function of temperature. The thermal conduc-tivity increases from room temperature to solidus temperature and it tends to decrease when thetemperature exceeds solidus temperature, whereas the linear expansion coefficient increaseswith the rise in the temperature.

5 Tensile Testing during Solidification

In this study, tensile tests from room temperature to solid-liquid coexistence region were carriedout by using the high temperature tensile test equipment with induction heating. At first, thespecimen was heated up to temperature above its liquidus temperature, and then kept for a cer-tain period of time. Next, it was cooled down to the test temperature at cooling rate 1 °C/s. Bythis means, a solidification structure was obtained. Finally, tensile load was applied at strain rate10–1 s–1 and the load-displacement was recorded.

Figure 4 illustrates the relationship between Young’s modulus and temperature. With thetemperature rise Young’s modulus decreases and when the temperature exceeds the solidustemperature (580 °C , it rapidly decreases as shown in Figure 4.

Figure 5 shows the relationship between tensile strength and temperature. The tensilestrength decreases with the temperature increases. When the temperature reached 625 ºC, the

Figure 2: Specific heat-temperature curve Figure 3: Thermal conductivity and linear expansion coefficient as a function of temperature

Figure 4: Young’s modulus as a function of temperature Figure 5: Tensile strength versus temperature

186

value of tensile strength became zero, so the temperature 625 °C is ZST (Zero Strength Tempe-rature) of this alloy at which corresponding solid fraction is 0.74.

6 Results and Discussion

6.1 Casting Experimental Results

Typical fracture occurrence position of the billet is shown in RHS of Figure 6. The cracks arewell generated at D, E and F position of the billet as shown in Figure 6, and the cracks are gene-rated on the surface and then propagate inside the billet. The surface cracks seem to initiate inthe concave parts of the billet owing to constrained contraction. These cracks are different fromthe round billet cracks which occur in the center of the billet [4–5] and here we call the crack ofthe near net shaped billet as surface cracks. Figure 6 also indicates the fractured SEM images.The fractured surfaces of the shaped billet have a rupture structure where intergranular fractureswith remaining liquid around interdendritic regions were observed, namely internal crack oc-curs in the solid/liquid coexisting (mushy) state above the solidus. Many intermetallic com-pounds and precipitates around grains were observed, and it is conjectured that the crack occurseasily because the intermetallic compounds and precipitates become fragile among grains.

6.2 Distortion Profile

The comparison of cast and calculated profiles of the billet with designed mold shape isshown in Figure 7. The distortion profile of cast billet agrees well with the calculated ones. Bothlegs were shifted inside by the solidification contraction. A neck falling phenomenon due to so-lidification shrinkage is observed. A comparison between calculated and cast butt-curls wasmade. The calculated value of but-curl (14 mm) agreed well with the cast result (13 mm).

Therefore it is considered that this thermomechanical modeling in this study is appropriatefrom the comparison between cast and calculated distortion of the shaped billet.

Figure 6: Crack occurrence position and SEM images of the fracture surfaces of the billet

187

6.3 Mechanism of Surface Crack

Figure 8 shows the distribution of plastic strain along X and Y directions obtained from thermaldistortion calculation at start-up state. It is easily understood for crack to generate at the D posi-tion since the tensile strain of the X direction concentrates in the D position as shown in Figure8 (a). It is the reason why both legs were shifted inside and contacted with the mold by the soli-dification shrinkage. Consequently the reaction forces by the mold are generated and the D po-sition receives the tensile stress. It is also likely to generate the cracks at E and F positionsbecause the tensile strain of Y direction concentrate in the E and F positions. A neck fallingphenomenon occurred and then the billet contacted to the mold due to solidification shrinkageas shown in Figure 7, so the reaction force by the mold is generated. As a result, the cracks be-come easy to be generated since the E and F division receive the tensile stress. Crack initiationposition predicted by the thermal stress analysis reproduced well the real casting result.

6.4 Cause of the Mold Grasped

The mold grasped by the billet is a peculiar phenomenon always occurred in the shaped billetcasting. When the phenomenon occurred, the billet did not moved down with bottom block be-cause the mold was clipped by the billet. Figure 9 indicates the distribution of temperature inthe cross section of the billet, (a), and the contact situation between the billet and mold, (b) and(c). Solidification is uneven at each location of the billet as shown in (a). Solidification had fi-nished in the areas such as two legs and the neck areas, but the solidification had not yet finis-hed in the center area. The contraction during solidification forced the two legs to shift towardthe mold due to the uneven solidification. The mold, hence, is grasped by the billet and so thecontact pressure between the billet and mold is generated as shown in (b), (c). It is consideredthat the phenomenon of the mold grasped by the billet occurs when the contact pressure beco-mes bigger and exceeds the drawing force applied in the bottom block.

Figure 7: Comparison of cast and calculated profile of billet with designed mold shape directions

Figure 8: Distribution of plastic strain along X and Y the at start-up state

188

7 Conclusions

A coupled flow and solidification and thermal deformation FEM analysis and casting experi-ment of near net shaped semi-continuously cast billet were carried out. The causes of the pecu-liar problems such as the surface crack and the mold grasped by shaped ingot were clarified.The main results are given as follows:

1. The cracks of irregular shaped billet are different from the ones of round billet. They are generated on the surface and then propagate inside the billet. The fractured surfaces of the shaped billet have a rupture structure where intergranular fractures with remaining liquid around interdendritic regions are observed, namely internal crack occurs in the solid/liquid coexisting (mushy) state above the solidus.

2. The surface crack of the near net shape billet is originated from the reaction force which ari-ses by contacting the mold by the distortion of the solidifying billet.

3. The phenomenon of the mold grasped by the billet occurs when the contact pressure beco-mes bigger and exceeds the drawing force applied in the bottom block.

8 Reference

[1] M. Anderson, R. Bruski, D. Groszkiewicz and B. Wagstaff: Light Metal, 2001,847–853

[2] K. Sugita, E. Sagisaka, S. Ichikawa: J. P. Patent # 7-67598, July 26, 1995 [3] H. Nagaumi: J. Jpn. Inst. Light Met. 50, 2000, 49–53[4] I. Farup, J. M. Drezet and M. Rappaz: Acta Meterialia, 49, 2001, 1261–1269[5] H. Nagaumi, T. Umeda: Journal of Light Metals, 2 , 2002,161–167

Figure 9: Distribution of temperature in the billet and contact situation between billet and mold

189

Numerical Simulation of DC Casting; Two Ways to Interpret the Results of a Thermo-Mechanical Model

W. Boender, A. Burghardt, E. P. van KlaverenCorus RD&T, IJmuiden, The Netherlands

1 Introduction

The manufacture of aluminium products starts with the casting of ingots that are subsequentlyrolled to plate or sheet. Plates and sheets are used for a wide range of products, for example ae-roplanes and beverage cans. The most widely used casting technique is direct chill (DC) ca-sting, which is a semi-continuous process.

Cold cracking, which is initiated by mechanical stresses at temperatures significantly belowthe solidus temperature, is one of the problems that impede the efficiency of DC casting. Figure1 shows two types of cold cracks that may occur in an ingot [1]. In the 1990s, numerical simula-tions of the mechanical behaviour of an ingot during casting were added to the tools that areused to solve these problems [2]. Here, the interpretation of the results of such simulations to as-sess the likelihood of cold cracking is discussed.

In a thermo-mechanical model, both the thermal and mechanical phenomena in an ingot aresimulated numerically. The model developed within Corus RD&T has been successfully em-ployed several times to improve the casting operations in the plants of Corus Aluminium. It wasset up using the finite element package MSC.Marc.

To illustrate how the results of a thermo-mechanical model can be interpreted, the casting ofa 2000 510 mm2 ingot of an Al - 4.5% Cu alloy at a casting speed of 60 mm/min was simulat-ed. This alloy was chosen, as it is a representative of high-strength alloys.

The input of the model consists of the dimensions of the mould, the ingot and the bottomblock; the process parameters for the drop; the physical properties of the ingot and the bottomblock; and the boundary conditions. Because of symmetry, only a quarter of the ingot has to besimulated. Figure 2 shows a perspective drawing of this quarter.

The length of the ingot was 2000 mm in the simulation, which gives a complete picture ofthe start-up. The bottom block was made of steel. The flow rate of the cooling water per unitlength of the mould opening and the drop rate were similar to the ones Hannart et al. [2] hadused for casting AA2024. Information on the physical properties of the Al - 4.5% Cu alloy andof the material of the bottom block can be found in [3].

From top to bottom, the following areas are distinguished at the vertical sides: the primarycooling zone, the air gap and the secondary cooling zone.

Direct contact between ingot and mould occurs in the primary cooling zone. A falling film ofwater cools the ingot in the secondary cooling zone. In the air gap between these two zones, theingot is neither in contact with the mould nor with the cooling water. At the butt of the ingot,there may either be direct contact with the bottom block; a narrow gap between them filled withair and vapour; or a wide gap filled with cooling water. Based on the local distance between thebutt and the bottom block, the model determined which situation occurred. See [3] for more in-formation on the boundary conditions applied.

190

2 Normal Stresses

Figure 3 shows the temperatures inside the ingot at a cast length of 1.88 m. It covers the samequarter of the ingot as shown in Figure 2. Figure 4 shows, for the same instance, the normalstresses in the x-direction, i.e. the stresses perpendicular to the casting direction and parallel tothe rolling face. The picture on the left-hand side shows the inside of the ingot and the pictureon the right its exterior. Tensile stresses are positive and compressive stresses are negative. The-re are large tensile stresses in the lower interior part of the ingot, and there are large compressi-ve stresses at the lower part of the rolling face. The shell of the lower part of the ingot hasalready reached the temperature of the cooling water. However, the interior of the lower part isstill hot. As the interior continues to cool down, it shrinks, which the rigid shell counteracts. Thestress state in the lower part of the ingot results from this balance between tensile stresses in theinterior and compressive stresses in the shell. For the y- and z-direction, similar stress distributi-ons are obtained [3].

Figure 1: Types of cold cracks Figure 2: Mould, ingot and bottom block

Figure 3: Temperatures Figure 4: Normal stresses in x-direction

191

Only a few measurements of stresses in a billet [4, 5] or an ingot [5, 6] have been reported.The results above are in line with their findings; tensile stresses exist in the interior and there arecompressive stresses below the surface.

Stress is a symmetric 3 3 tensor, which contains three normal stresses and three shear stress-es. In the solidified part of the ingot, none of the three normal stresses is negligible. Therefore, areal triaxial state of stress exists.

A consequence of the symmetry of the stress tensor is that, at each point, three mutually per-pendicular directions exist along which there are no shear stresses. These directions are calledthe principal stress directions, and the corresponding normal stresses are called the principalstresses 1, 2 and 3, where 1 2 3.

To gain more insight into the stresses within an ingot, the principal stresses and their direc-tions were evaluated. This approach also enables an assessment of the maximum normal stresscriterion, which is assumed more applicable to brittle materials, like gray cast iron, than the oth-er theories [7, 8]. This criterion states that failure occurs when either the maximum principalstress, 1, reaches the uniaxial tensile strength or the minimum principal stress, 3, reaches theuniaxial compressive strength [9].

Figure 5 shows the directions of the major principal stress for the ingot's interior, i.e. thewide symmetry plane, and its surface, i.e. the rolling face. The major principle stress will equal

1 if 1 is greater than the absolute value of 3, else it will equal 3. The major principal stressesare directed horizontally in the centre region. Going towards the narrow side, these tensilestresses turn more and more to the vertical direction. Griffith postulated that a crack would openup in the plane normal to the direction of maximum stress [10]. Figure 6 shows the values of themajor principal stresses for the interior and the surface of the ingot. Assuming cracks only occurin regions of large tensile stresses, and that they propagate perpendicularly to the tensile stressdirections, one can conclude the following.

Areas with a high probability of crack formation are found about 100 mm below the surfaceof the narrow side or a few 100 mm above the butt. The cracks near the narrow side would thenpropagate following a path that resembles J-cracks. Those near the bottom block would propa-gate upwards like trouser cracks. The observation of the directions of the major principal stress-es, therefore, led to hypotheses for the start locations and the directions of trouser and J cracks.

Figure 5: Directions major principal stress Figure 6: Major principal stresses

192

3 Fracture Mechanics

The results of others [11, 12] for the maximum principal stress when casting AA7050 can becompared with the tensile strength that they had measured. The calculated stresses appear to beless than the tensile strength. This finding agrees with the experience of the authors. The factthat cold cracks nevertheless occur is, therefore, due to the presence of stress raisers.

To determine the effect of the size of an imperfection on an ingot's mechanical behaviour,fracture mechanics was applied. An existing crack propagates only if the total energy of the sys-tem decreases [13]. Following this energy approach, the maximum size of an imperfection thatcan be tolerated under a certain loading can be determined. For a through-thickness crack,which is ideally sharp, in an infinite plate under remote tensile loading, a simple expression wasderived:

(1)

where KIc is the plane strain fracture toughness, is the tensile stress, a is half the length of thecrack and the subscript c stands for critical. Cracks longer than 2ac grow unstoppably.

According to Chang and Kang [11], the minimum value of KIc for AA7050 in the as-castcondition at room temperature is 8.54 MNm–3/2. The fracture toughness of AA2024, which so-lidified at 0.77 °C/s, is about 14.63 MNm–3/2 [14]. Using these values and equation (1), the crit-ical crack lengths at a few tensile stresses were calculated to demonstrate the application offracture mechanics. The selection of the stresses is based on the tensile stresses in Figure 6. Ta-ble 1 shows the results. These guesses at the critical crack lengths can be compared with sizes ofdefects in an ingot. For example, Gauckler et al. [15] mentioned the sizes of a few types of in-clusions in aluminium. Especially oxide skins, whose maximum size may be 5 mm, may begreater than the critical crack length. Hence, they may cause cracks in areas with high stresses.

Table 1: Critical crack lengths for AA7050 and AA2024

Stress Half crack lengthAA7050 AA2024

Mpa mm mm

50 9.29 27.25100 2.32 6.81150 1.03 3.03200 0.58 1.70250 0.37 1.09300 0.26 0.76

2

2

Icc

Icc

Ka

Ka

193

4 Conclusions

Two methods were used to interpret the results of a TM-model. The application of the majorprincipal stresses demonstrated where cracks might be initiated, and how they might propagate.The application of fracture mechanics produced guesses at the critical crack length, i.e. fromwhich minimum size cracks would grow. Internal defects, like voids and inclusions, should besmaller than this critical size. If imperfections are distributed homogeneously in an ingot, a lar-ge volume with high tensile stresses will contain more of them than a small volume. Hence, inthe former case, it is more likely that a crack will grow than in the latter. Therefore, to avoidcracks it is not only necessary to lower the stress levels but also to decrease the sizes of volumeswith high stresses. The thermo-mechanical model is a tool to find ways to achieve this.

5 References

[1] J. Du et al. in Light Metals 1998 (Ed.: B. Welch), TMS, Warrendale, PA, USA, 1998, p. 1025

[2] B. Hannart, F. Cialti, and R. van Schalkwijk in Light Metals 1994 (Ed.: U. Mannweiler), TMS, Warrendale, PA, USA, 1994, p. 879

[3] W. Boender et al. in Light Metals 2004 (Ed.: A.T. Tabereaux), TMS, Warrendale, PA, USA, 2004, p. 679

[4] A. Roth, M. Welsch, and H. Röhrig, Aluminium 1942, 24, p. 206[5] G. Seeger, Gießerei 1951, 38, 14, p. 325[6] S.A. Levy, R.E. Zinkham, and J.W. Carson in Light Metals 1974, (Ed.: H. Forberg), The

Metallurgical Soc. of the AIME, New York, NY, USA, 1974,Vol. 2, p. 571[7] J.H. Faupel and F.E. Fisher, Engineering Design, (New York, NY, USA: John Wiley &

Sons, Inc., 1981), p. 242[8] N.E. Dowling, Mechanical Behavior of Materials, Prentice Hall, New Jersey, USA, 1999,

Chapter 7, p. 256[9] S. Timoshenko, Strength of Materials, D. van Nostrand Company Inc., Princeton, NJ,

USA, 1958, Part II, p. 444[10] G.C. Sih in Mechanics of Fracture 1 - methods of analysis and solutions of crack pro-

blems (Ed.: G.C. Sih), Noordhoff, Leiden, The Netherlands, 1973, Introductory chapter, p. XXIX

[11] K-M. Chang and B. Kang, Journal of the Chinese Inst. of Engineers 1999, 22, 1, p. 27[12] H-M. Lu, K-M. Chang and J. Harris in Light Metals 1997 (Ed.: R. Huglen), TMS, War-

rendale, PA, USA, 1997, p. 1091[13] M. Janssen, J. Zuidema, R.J.H. Wanhill, Fracture Mechanics, Delft University Press,

Delft, The Netherlands, 2002, p. 3[14] R.K. Paramatmuni et al., Material Science and Engineering 2004, A 379, p. 293[15] L.J. Gauckler et al., J. of Metals 1985, 37, 9, p. 47

194

Continuous Casting of Hypermonotectic AlBiZn Alloys: Experimental Investigations and Numerical Simulation

M. Gruber-Pretzler1, F. Mayer1, M. Wu1, J. Moiseev2, B. Tonn2, A. Ludwig1

1University of Leoben, Leoben, Austria 2Clausthal University of Technology, Clausthal-Zellerfeld, Germany

1 Abstract

Hypermonotectic alloys separate into two melts prior to solidification. Therefore, continuouscasting of hypermonotectic alloys is still a challenge that inhibits the use of these bearing mate-rials in industry. As the secondary phase distribution has a major impact on the wear propertiesof the final product, it is of utmost importance to understand its formation. The motion of the se-condary phase droplet during demixing of the melt is caused mainly by Marangoni transportand gravity-induced sedimentation. By a series of experiments with hypermonotectic AlBiZnalloys, a laboratory scale strip casting device has been adapted to the special behavior of thesealloys. The resulting droplet distributions were experimentally investigated by varying not onlythe alloy composition but also relevant process parameters. Beside this investigation the forma-tion of secondary phase distribution was also modeled with a two-phase volume averaging ap-proach. The Bi-enriched minority liquid phase is treated as the second phase whereas the parentmelt as first. The model solves the mass, momentum, species and enthalpy conservation equati-ons for both liquids including droplet nucleation, solute redistribution and monotectic reaction.

2 Introduction

Alloys with a miscibility gap in the liquid state, especially for those with gross concentrationabove the monotectic point (hypermonotectics), are potential bearing materials for the automo-tive industry [1, 2]. As hypermonotectic alloys separate into two melts prior to solidificationtheir continuous casting is still a challenge. Difficulties arise from the fact that the secondaryphase droplets move due to Marangoni forces and gravity-induced sedimentation. After a periodof less attention, the interest on the solidification of hypermonotectic alloys has increased again,because the EU has proscribed the use of Pb containing alloys. Therefore, the development ofnew Pb-free bearing materials is of great importance and so alloys based on Al-Bi are again ofparticular interest for materials research and development [3, 4].

Strips of hypermonotectic alloys are produced at the TU Clausthal on a vertical continuouscasting unit at a laboratory scale. This unit insures the possibility of high cooling rates that arenecessary for the casting of hypermonotectic alloys in order to cross the demixing interval veryrapidly. Due to consequent prevention of minority-phase sedimentation, a fine distribution of Bidrops in the Al matrix can be produced. However, the influence of additional alloying elements,like Zn, or of differences in process parameters, like casting speed, casting temperature andcooling conditions on the distribution of the minority phase is still open and thus addressed inthe present paper.

195

Theoretical descriptions of the demixing of hypermonotectic Al-based alloys were alreadysuggested in [5–6]. A numerical treatment of this process was presented by two of the authors in2003 [7, 8]. Their work addressed the influence of Marangoni force and of gravity-induceddroplet sedimentation on the droplet distribution. However, the studies presented in [7, 8] weredone for 2D square geometry which was assumed to be cooled from all sides. In the presentstudy, a first numerical description of a strip casting process for the solidification of a hyper-monotectic binary AlBi10 alloy is given. The temperature field and the distributions of dropletfraction and density are presented and discussed.

This paper consists of two parts, the first is the experimental part and the second the numeri-cal part.

3 Experimental Part

3.1 Description of the Continuous Casting Unit

The vertical continuous casting machine at the TU Clausthal consists of the mold with pouringnozzle and tundish and the mechanical-hydraulic strand withdrawal mechanism. The ready-as-sembled continuous casting copper mold is made out of two side frames and two spacers. Thespacers confine the strand laterally and keep the mold gap at a constant width of = 10 mm. Asteel pouring nozzle is set into a notch at the top of the mold and a tundish symmetrically on topof that. Figure 1 illustrates schematically the assembled mold ready for use. The tundish (1) andthe steel pouring nozzle (2) are pre-heated up to the pouring temperature. The melt is heated upin a separate furnace also to that temperature and then poured into the tundish in order to startthe casting.

There are two cooling water circuits. In the first circuit the cooling water is flowing throughthe cooling water supply (5) towards the cooling channel (6) where the water directly impingesthe strand (7). This cooling channel acts as secondary cooling, while the upper part of the moldoperates as primary cooling. In each half of the mold there are a total of 17 cooling channels andassociated cooling grooves. With the cooling circuit (4) an additional heat removal from the

Figure 1: Scheme of the middle cross sec-tion through the mold (detailed informa-tion in the text and see ref. [9])

Figure 2: Bi droplets distribution in an AlBi8Zn6 strand

196

mold is realized. The amount of water flowing through the two cooling water circuits can bevaried depending on the required cooling intensity. With the maximum cooling possible a tem-perature gradient at the solidification front of G = 50 K/mm [3]. Steady-state casting conditionsare found to be reached after a few seconds. Therefore, a uniform phase distribution along thelength and width of the strand can be established.

3.2 Evaluation and Discussion of Results

In the primary cooling zone which extends over the first L = 23 mm of the copper mold, themelt is cooling rapidly. When the temperature falls below the binodal temperature, droplets ofthe minority phase start to form. From the experimental point of view, it is obvious to assumethat the main nucleation happens in areas close to the walls. Further growth the nucleated dro-plets is then governed by diffusion. Due to the Marangoni effect, which is caused by the tempe-rature-dependence of the interfacial tension, the continually growing particles migrate towardsthe middle of the strand. This favors their collision and coagulation. The Marangoni force isproportional to the temperature gradient and proportional to the square of the droplet diameter,whereas the Stokes forces increases only linearly with the diameter. The occurring droplet moti-on is the overlay of casting flow, Marangoni motion and sedimentation.

Figure 2 gives a qualitative impression of the Bi-droplet distribution in the strand ofAlBi8Zn6. The structural constituents occurring in this case were produced at a casting velocityof Vcast = 450 mm / min, a casting temperature of about Tcast = 960 °C and a cooling water flowrate of about Q = 4000 l/hour for each mold side frame. In all ternary Al-Zn-Bi alloys the edgestrand zones are depleted in Bi-droplets up to a depth of 4 mm. In the Al-Bi system the monote-ctic composition occurs at cm = 3.4 wt.% of bismuth. This composition corresponds approxi-mately to the contents measured in the edge zones. In opposite, the core zone of the strand isenriched in droplets and reveals around twice as much as Bi content compared to the initialcomposition.Figure 3 to 6 give an overview on the influence of variations in alloy composition and processparameters like casting speed, casting temperature or cooling water flow rate on the maximumand the mean Bi-droplet size. Figure 3 shows how the mean and maximum droplet size in theenriched middle zone by varying the Zn content for the two alloys, AlBi6Znx and AlBi8Znx. Fi-gure 4 shows the influence of the casting speed on the mean and maximum droplet size for theAlBi8Zn6 alloy.

It is visible that increasing Zn content increases the Bi-droplet size to a certain level for both,mean and maximum values. However, if the amount of Zn in the alloy is larger than 10 wt.%the droplet size decreases again. On the other side, high Bi amount leads to a clear increase ofthe droplet size. The influence of the Bi-content on the droplet size is substantially stronger thanthe one of Zn and higher radial components.

If a casting is desired which reveal only small size maximum droplets, the optimum castingspeed for an AlBi8Zn6 alloy would be around Vcast = 550 mm/min. This finding is true for acasting temperature of Tcast = 960 °C and a cooling water flow rate of Q = 4000 l/hour. A smallcasting speed would favor the drop growth already in the steel feeder. Evidently, a larger cast-ing speed would lead to a deep melt pool and thus higher temperature gradients in front of thesolidification front. This increases the importance of the Marangoni convection and hence influ-ences the appearing droplet distribution.

197

Figure 5 shows the influence of the cooling water flow rate on the maximum and mean Bi-droplet size. For alloys which should reveal only small size maximum droplets, the optimumcooling water flow rate is around Q = 2000 l/hour for Tcast = 960 °C and Vcast = 550 mm/min. Asmaller flow rate/cooling rate favors the droplet sedimentation, whereas a higher flow rate/cooling rate Marangoni motion is encouraged.

Figure 6 shows the influence of the variation of casting temperature on the maximum and themean Bi-droplet size. It is obvious that small casting temperatures lead to smaller temperaturegradients ahead of the solidification front and thus higher viscosity of the parent melt. There-fore, the droplet motion is made difficult.

Figure 3: Influence of Zn- and Bi-content on the mean and maximum Bi-droplet size in AlBi8Znx and AlBi6Znx

Figure 4: Influence of casting speed on the mean and maximum Bi-droplet in AlBi8Zn6

Figure 5: Influence of quantity of cooling water on the mean and maximum Bi-droplet in AlZn6Bi8 alloy

Figure 6: Influence of casting temperature on the mean and maximum Bi-droplet in AlZn6Bi8 alloy

198

4 Simulation Part

4.1 Model Description

Because of the fact that in the present paper process modeling is of importance we skip a detai-led model description. The reader is referred to the original papers of corresponding authors [5,6]. Nevertheless, a brief description of most important model assumptions is given.

The used two phase model considers the parent alloy as the first, L1, and the forming dropletsas the second phase, L2. During the monotectic reaction the monotectic matrix is transformeddirectly from L1. Therefore, the solidification of monotectic matrix is modeled by increasing theL1 viscosity and releasing latent heat on reaching the monotectic temperature. The decomposedL2 droplets approaching the monotectic reaction front are modeled to be entrapped in the mono-tectic matrix by applying the same enlarged viscosity at or below the monotectic point. A simi-lar approach is used by [1,6,10–12].

The model is based on solving the mass, momentum, species and enthalpy conservationequations for both liquids including droplet nucleation, solute redistribution and monotectic re-action [7]. Both Stokes and Marangoni motion is taken into account.

In addition to the above mentioned model considerations the following assumptions aremade:

• Gravity-induced sedimentation is modeled with the Boussinesq approach.• Both liquid phases are assumed to have the same viscosity.• Collision and coagulation of droplets are not yet taken into account.• Diffusion in a single droplet is thought to be infinite.

4.2 Geometry and Boundary Condition

For the process simulation of a binary AlBi10 alloy a casting velocity of Vcast = 828 mm/min anda casting temperature of Tcast = 792 °C is considered. Due to the geometry of the casting a 2Dsymmetry has been chosen for the simulation. The mold is schematically shown in Figure 1 andin a 3D view in Figure 7. Here ( ) indicates the copper mold, ( ) indicates the used steel pour-ing nozzle on the top, ( ) and ( ) show the primary cooling zone, where ( ) indicates the partof the mold where ideal contact is assumed and ( ) the lower part where already solidificationshrinkage and thermal contraction of the solid shells is considered. ( ) indicates the position ofthe secondary cooling zone. In Figure 8 the applied boundary conditions are shown. ( ) givesthe position of the inlet, where a velocity inlet with the casting speed Vcast is considered. A heattransfer coefficient (HTC) of h = 100 W/m²K and a temperature of Tfeed = 682 °C is consideredfor the steal feeder ( ). ( ) is divided in two parts: the upper part where h = 1000 W/m²K andTmold = 202 °C and a lower part where h = 100 W/m²K and Tmold = 202 °C is used in order tomodel the start of the secondary cooling. For the rest of the secondary cooling ( ) we have ap-plied h = 10000 W/m²K and Twater = 23 °C. For the outlet ( ), outflow is considered. A grid of15960 cells and 16523 nodes is used.

The applied material properties are the same as used in [7]. Tab. 1 gives the values for theused thermodynamic properties of the system. As initial conditions, we start with hot melt(Tinit = 792 °C) at rest (Vinit = 0 m/min). Cooling and inflow are then resulting in an acceleration

199

of the melt in the strand, the formation of a solidifying shell and nucleation and growth of Bi-droplets.

4.3 Results and Discussion

Figure 9 shows three properties in the upper part of the casting, namely temperature field, volu-me fraction and density of the Bi-droplets 44s after switching on flow and cooling. The picturesare overlaid with two isothermal lines: binodal (Tb = 786 °C) and monotectic (Tm = 657 °C).

Table 1: Thermodynamic phase diagram used for the simulation

The temperature field displayed in Figure 9a shows the temperature evolution from the upperborder of the steel feeder ( ) and to the end of the copper mold ( and ). According to theexperimental conditions, a gap between casting and copper mold is considered 15 mm belowthe steel feeder ( ). This gap is indicated in Figure 9 by the white gap between mold and cast-ing. The secondary cooling zone starts with that gap and moves on fare below the copper mold.

t = 10 s after switching on cooling und inflow, the temperature distribution reaches steady-state. The complicated boundary conditions applied for the considered strip caster lead to a tem-perature distribution (Figure 9 a) in the solidifying strand that has a strong temperature gradient

Monotectic temperature Tm 930 K 657 °CMonotectic concentration cm 0.47 at.% 3.526 wt.%L2 monotectic concentration cL2 83.4 at.% 97.493 wt.%Critical Temperature Tc 1310 K 1037 °CMelting point of Al Tf

A 933 K 660 °CMelting point of Bi Tf

B 543 K 270 °CGross concentration c0 1.415 at.% 10 wt.%Slope of liquidus at c0 m 148.1 K per at.% 20.42 °C per wt.%Partitioning coefficient k 51.72 9.55

Figure 7: Casting Mold (described in the text) Figure 8: Geometry and boundary con-ditions (described in the text)

200

in the upper part of the copper mold ( ). As the Marangoni force is proportional to the temper-ature gradient, the droplets which form along the wall are forced to move towards the centre ofthe strip (Figure 9 b). The movement of the droplets towards the centre of the strand is overlaidby Stokes motion and is therefore at the origin of two small vortices which occur at the wall( ).

Figure 9c shows the density, n, of the Bi-droplets. The black curve overlaid is the value of nalong the strand at a depth of 25 mm (n = 0 is at the center of the figure). Due to the small tem-perature gradient in the steel feeder nucleation appears visible slightly below the binodal tem-perature (Figure 9c). It can be seen that the nucleation of the droplets is somewhat oscillatory innature ( ). At present, the origin of these oscillations is unclear. Zhoa [5] mentioned that oscil-lations appear in numerical calculations for describing the microstructure evolution in hyper-monotectic alloys. The reasons therefore are changes in the supersaturation of the liquid causedby Marangoni and Stokes motion. Detail studies on that issue are ongoing.

5 Concluding Remarks

First results of investigations with ternary Al-Bi-Zn alloys have been presented in the first partof the paper, intended mainly to examine the influence of additional alloying elements, like Zn,and various process parameters. The results intend to show the following facts:• Zn content variation has minor effect on the maximum and the mean Bi-droplet size compa-

red to Bi-content variations.

Figure 9: steel pouring nozzle, (ideal contact) and start of the secondary cooling (air gap). (a) Tempera-ture field with a strongest gradient in ; (b) volume fraction: indicates the position of high Marangoni motion towards the centre of the cast; (c) density of droplets: indicates oscillatory in the droplet density. The black curve overlaid is the value of droplet density along the strand at a depth of 25 mm.

201

• For the alloy AlBi8Zn6 the casting conditions which would result in the smallest maximum droplet size for Tcast = 960 °C have been found to be: a casting velocity of Vcast = 550mm/min combined with a cooling water flow rate of Q = 2000 l/hour. For lower casting temperature the maximum droplet sizes can even been reduced.

• Further investigations are planed to evaluate the influence of different grain refiners, especially based on Al-Ti-C and Al-Ti-B, on the fineness of the minority phase distribution.

In the second part of the paper preliminary simulation results for a strip casting of AlBi10have been presented. The complexity of the demixing process of hypermonotectic alloys togeth-er with the different boundary conditions particular for the considered caster make the qualita-tive interpretation of the results difficult.

However, the following statements can be made:

• The temperature field shows the highest gradient in the upper part of the copper mold.• The Marangoni force increases with increasing temperature gradient which causes the

motion of the Bi-droplets towards the centre of the strip.• Oscillating droplet nucleation appears spatially in the strand. At present the reason for that is

not clear. Maybe changing supersaturation in the liquid caused by Marangoni motion and Stokes sedimentation are of importance. Detail studies on that issue are ongoing.

6 Acknowledgements

This work was financially supported by the ESA-MAP project ‘Solidification Morphologies ofMonotectic Alloys-MONOPHAS’.

7 References

[1] L. Ratke, S. Diefenbach, Mater. Sci. Eng. 15, 1995, p. 263[2] B. Predel et al, Decomposition of Alloys : The Early Stages, ed. Walter H. U., Ashhy M.

F., Berlin: Springer, p. 517[3] J. Moiseev, S. Vogelgesang, H. Zak, H. Palkowski, B Tonn, Metall 58, 2004, p. 289[4] J. Moiseev, H. Zak, H. Palkowski, B. Tonn, Aluminium 81, 2005, p. 92[5] J. Zhao, L. Ratke, Scripta Mater. 39, 1998, p.181[6] J. Zhao, L. Ratke, Z. Metallkunde, 89, 1998, p. 241[7] M.Wu, A. Ludwig, L. Ratke, Modell. Simul. Mater. Sci. Eng. 11, 2003, p. 755[8] M.Wu, A. Ludwig, L. Ratke, Metall. Mater. Trans. 34A, 2003, p. 3009[9] B. Prinz, DE Patent 40 03 018 A1, 1991[10] J. Alkemper, L. Ratke, Z. Metall. 85, 1994, p. 365[11] S. Diefenbach, Ph. D. Thesis Ruhr-Univercity Bochum, 1993[12] L. Ratke et al., Materials and Fluid Under Low Gravity, ed L. Ratke et al. Berlin: Sprin-

ger, 1995, p. 115

202

Numerical Simulation of the Upward Continuous Casting of Magnesium Alloys

A. Landaberea1, P. Pedrós1, E. Anglada1 and I. Garmendia1

1INASMET Foundation, San Sebastian, Spain

1 Abstract

The continuous casting of magnesium alloys in vertical upward direction is a novel technologywhich can be employed for the production of semi-finished materials circumventing the maindisadvantages of using conventional casting processes since the risks of burning and explosionare practically eliminated. The present investigation deals with the simulation of the upwardcontinuous casting of round billets of magnesium alloys. The equations for the flow field withheat transfer are numerically solved by a finite volume method and the solidification is accoun-ted via an enthalpy-porosity formulation where the mushy region is modeled as a pseudo porousmedium. The obtained temperature distribution is then used as input for a thermo-mechanicalanalysis to determine the stress field in the billet during the casting process. Several configurati-ons have been simulated and comparison of computed results with available experimental datais provided.

2 Introduction

Continuous casting is the key technology for cost effective manufacturing of semi-finished ma-terials for further processing to obtain high quality final products and has been the state of theart in steel and aluminum processing for decades. However, due to the specific properties ofmagnesium alloys, i.e. tendency to burn of the melt when exposed to air, the danger of explosi-on when the melts come into contact with water and the low strengths at temperatures above300 ºC, the application of conventional gravity continuous casting techniques with magnesiumalloys is only possible assuming high risks.

In order to prevent these difficulties, the Upward Direct Chill (UDC) casting principle hasbeen proposed at IW – University of Hanover for magnesium alloys where the continuous cast-ing is operated vertically against the gravity. Initial works have been further developed withinthe EU – 5th Framework funded “EuroMagUpCaster” project (see [1] for further information).

Over the years, several mathematical models have been developed to help to understand theprocesses occurring during continuous casting most of them focused on steel and aluminium al-loys (see for example [2, 3]). As part of the “EuroMagUpCaster” project, a model has been de-veloped to study the UDC process which is described in the present article.

203

3 Model Description

The model calculates the temperature distribution and the associated thermal stresses in the bil-let during the upward casting process. To do this, a combination of commercial together withself developed codes has been adopted. Thus, the commercial Computational Fluid Dynamics(CFD) code FLUENT [4] has been used to calculate the flow field, temperature distribution andsolidification of the billet. FLUENT is based on the Finite Volume Method (FVM) and uses anenthalpy-porosity method for modelling the solidification. The thermo-mechanical analysis hasbeen performed with MSC.Marc [5] which is a general purpose code based on the Finite Ele-ment Method (FEM) with well recognized capabilities for nonlinear problems. A series of trans-lation codes have been developed to pass the mesh and thermal solution from FLUENT toMSC. Marc so the temperature history at each node is applied as thermal load in the stress ana-lysis.

3.1 Computational Domain

Due to the geometry and configuration of the casting process, the problem can be considered asaxially symmetric and simplified to a two dimensional formulation. The computational domainfor the thermal problem is sketched in Figure 1 and includes both the mold and the billet. In theresults presented in this work, the mold has not been considered for the mechanical calculations.FLUENT considers the horizontal axis as the axial direction and the vertical as the radial.According to this, the gravity vector has been specified to act in the negative horizontaldirection. The domain has been meshed with quadrilateral cells in FLUENT which convert tolinear quadrilateral elements in MSC.Marc

3.2 Boundary Conditions

Regarding the flow and thermal problem, at the inlet section a uniform velocity and temperatureare imposed according to the casting process parameters. In the mold, convection to the primarywater cooling loop is assumed with a heat transfer coefficient of h = 4500 W/m2/K and a watertemperature of 20 ºC. At the external surfaces of the mold natural convection to ambient air(h = 10 W/m2/K, T = 20 ºC) is applied. After the mold, a combination of natural convection

Figure 1: Computational domain

204

(h = 10 W/m2/K) and radiation ( = 0.8) to ambient air (T = 20 ºC) is applied at the surface ofthe billet. In order to include the effect of gap formation between the billet and the mold due toshrinkage during solidification and further cooling a contact resistance Rc has been applied atthe interface between the billet surface and the mold. This contact resistance is assumed to varywith the liquid volume fraction being zero (perfect contact) when the billet is in liquid state andRc when the billet has solidified. The value of the contact resistance has been modified until theresults of the model have been adjusted to the experimental data provided by IW. At the outletsection, a mass flow balance is applied.

Afterwards, the temperatures computed with FLUENT are used as input to MSC.Marc forthe stress analysis where an essentially unconstrained billet is assumed since, together with thesymmetry conditions along the axis, only the longitudinal displacement of the node in the axislocated at the outlet section has been restricted.

3.3 Material Properties

The model uses thermally dependent physical properties of the materials for both the thermaland mechanical problems. The properties have been obtained mainly from [6, 7]. The model as-sumes a elasto-plastic behavior of the billet material during solidification. It must be pointed outthat the mechanical properties above 300 ºC are not available so they have been extrapolated tonearly null values at solidus temperature. In addition, the thermal expansion coefficient for tem-peratures above the liquidus temperature has been assumed to be zero to account for the liquidphase in the model.

4 Results

The model has been applied to the stationary continuous casting of 90 mm diameter billets ofpure magnesium and AZ80 alloy. Results are presented hereafter.

4.1 Pure Magnesium

The computed temperature and streamline distributions for pure magnesium at a casting speedof 95 mm/min are presented in Figure 2. As expected, the billet cools down in the mold due tothe action of the water cooling and some reheating occurs at the surface of the billet after themold due to the heat stored in the central region and to the lack of a secondary cooling unit. Thetemperature at the surface of the billet just after the mold is around 550 ºC. A large recirculationgenerates after the step of the melt distributor due to the abrupt change of section.

Figure 3 compares the shape of the solidification front obtained in casting trials with thatpredicted by the simulation after final adjustment of the model showing a good correlation. Themodel predicts that the billet solidifies before the melt reaches the mold insert. This is in agree-ment with the surface cracks generated on the billet (see Figure 3) which can be attributed to thestepwise filling and solidification of material in this zone. As can be observed in the tempera-ture distribution, the corner region between the melt distributor and the mold insert is refrigerat-ed by the water cooling so a better insulation between these parts should be introduced to avoid

205

the creation of the surface cracks. Several simulations have been performed with different cast-ing conditions to try to solve this problem and modifications to the casting assembly have beenproposed which are currently under preparation.The contours of circumferential stresses developed in the billet together with the deformedshape are given in Figure 4. The billet exhibits a contraction in length and in diameter due to thesolidification and further cooling. In the figure the residual stresses in a cross section in the bil-let away from the mold is also included. As can be seen, compressive residual stresses appear atthe outer zone which turn into traction inside the billet and again into compression near the axis.Some oscillations are originated in the distribution due to numerical difficulties in the solutioncoming from the highly non linear properties of the material and the very low introduced resi-stance at high temperatures. Figure 5 shows the radial displacement at the billet surface in themold region. Initially, a dilatation of the billet is predicted which is not constrained since nocontact algorithm has been included in the calculations. Later, the billet shrinks due to the soli-dification and further cooling with a radial contraction of 0.113 mm at the outlet of the moldwhich gives an overview of the size of the gap between the mold and the billet surface.

Figure 2: Pure magnesium. Contours of a) temperature (ºC) and b) streamlines

Figure 3: Pure magnesium. Experimental and computed solidification front and billet surface quality

206

4.2 AZ80

Figure 6 shows the temperature distribution and streamlines computed for AZ80 alloy at a ca-sting speed of 60 mm/min. The temperature contours follow a similar distribution as in the case

Figure 4: Pure magnesium. Contours of circumferential stress (Pa) in the billet and radial distribution of circum-ferential stress at a section away from the mold

Figure 5: Pure magnesium. Radial displacement (m) of the billet surface in the mold region

207

of pure magnesium with cooling in the mold region and some reheating after the mold. Thetemperature at the surface of the billet when it leaves the mold is around 400 ºC, lower than inthe previous case mainly due to the lower casting speed so the material has more time to cool in-side the mold. In addition, the recirculation zone after the step of the melt distributor has almostdisappeared also due to the reduced velocity.

The calculated liquid volume fraction is given in Figure 7 where the wider mushy zone of thealloy can be clearly observed. The simulation is compared with the solidification front obtainedin casting trials showing also a good agreement. Again, the solidification takes place before themelt reaches the mold insert leading to the generation of the surface cracks observed in the trials(see also Figure 7). Also, from the results of the simulation, it can be said that the materialpresents a coherency temperature and develops some strength in the mushy zone before reach-ing the solidus temperature.

Figure 6: AZ80. Contours of a) temperature (ºC) and b) streamlines

Figure 7: AZ80. Experimental and computed solidification front and billet surface quality

208

5 Conclusions

A mathematical model has been developed to analyze the fluid flow, heat transfer and stress ge-neration during the upward continuous casting of magnesium alloys. The predictions have beenfound to be in good agreement with available experimental data.

The model has shown that, for the selected casting conditions, the solidification occurs be-fore the melt reaches the mold insert leading to the formation of cracks at the surface of the bil-let. The main reason for this effect is the low efficiency of the thermal insulation between thewater cooling and the melt distributor. Through a series of simulations, several modificationshave been proposed to the casting assembly to overcome this problem and to obtain higher qual-ity billets.

The model has proved to be an efficient tool to help in the design and optimization of thecasting process. It has been also applied to other magnesium alloys and to an upscaled mold as-sembly with d = 203 mm for which preliminary casting parameters have been defined.

6 References

[1] Fr.-W. Bach, S. Schacht, A. Rossberg, DGM Conference, Continuous Casting 2005, Neu-Ulm

[2] J.E. Kelly, K.P. Michalek, T.G. O’Connor, B.G. Thomas, J.A. Dantzig, Metall. Trans. A 1988, 19A, 2589–2602

[3] J.M. Drezet, M. Rappaz, Metall. Mater. Trans. A 1996, 27A, 3214–3225[4] FLUENT´s Manual version 6.2, Fluent Inc., 2005.[5] MSC.Marc 2005, User’s Manual, MSC.Software Corporation, 2005[6] R.D. Pehlke, A. Jeyarajan, H. Wada, Summary of thermal properties for casting alloys

and mold materials, University of Michigan, 1982[7] Magnesium and Magnesium alloys (Ed.: M.M. Avedesian, H. Baker), ASM International,

USA, 1999

209

New Possibilities in the Simulation of Continuous Casting Processes with WinCast-Conti

H. Ricken1, C. Honsel2

1Technische Universität München, Institute of Metal Forming and Casting (utg), Walther-Meißner-Straße,D-85747 Garching2RWP GmbH, Am Münsterwald 11, D-52159 Roetgen

1 Abstract

The casting-simulation WinCast was adapted to the continuous casting process. The thermaland mechanical equations are coupled and solved for every time step. Thus the material flowand the shrinkage of the billet can be predicted. The interactive mesh-generator enables the userto change geometry very freely. Boundary conditions and process parameters may be varied in awide range. This freedom in variation enables the user to change parameters and add specialgeometrical features to existing or planned system configurations. Special material propertiesmay be integrated into the model. The Institute of Metal Forming and Casting(utg) at the Tech-nische Universität München has developed a test facility to predict heat transfer coefficientsagainst gap length in the mould. Now a lot of parameter studies with the real system can be re-placed by simulations. The ability of the user to improve the model is a major step to upgradeproduction facilities.

2 WinCast

It was in 2002 when the Institute of Metal Forming and Casting at the Technische Univer-sität München began continuous casting simulations with the RWP software package WinCast.

Figure 1: Structure of the WinCast-Conti software package

210

At this time was difficult to calculate the transient heat transfer between strand, mould andcooler. Especially the complex phenomenon the of the gap-building between mould and coolerwas yet to be solved.

The advantage of freedom in modelling with pentaeder elements and the convenience in set-ting various boundary conditions revealed the foundations to create a powerful simulation forthe modelling of continuous casting processes.

Figure 2: Thermal results

Figure 3: Stress and distortion

211

These attributes were the most important properties of the programm package giving the im-pulse to develop a special continuous casting module.

In 2005 the transport of material during the process and the coupling of thermal and stresscalculation were implemented.

The transport of matter is calculated not only for molten metal and strand, but also for thecooling water.

With the new module there is a direct coupling between the thermal results, the resulting ge-ometrical deviation of the whole system and the changing heat transfer caused by gaps openingand closing.

For calculating the complete thermal behaviour of the mould it is also essential to determinenot only the cooling and solidification of the melt, but also the heating of the cooling water.

With the new WinCast-Conti module it becomes possible to calculate the temperature of theeffluent water in dependency on the influent water and the current heat transfer through the

mould. This is a major step to calculate the thermal behaviour and heat conductivity of mouldand cooler in an early stage of design.

So with the heat balance between strand and coolant the complete thermal behaviour of theheat exchanger can be calculated. With the ease of temperature measurement in cooling wateras well as in melt and strand surface outside the mould the consistency between simulation andreality can be proved.

The graphic representation of the results provides a comprehensive insight to the process. It is possible to select materials or sections to evaluate special regions of interest.The distortion is represented with a user-defined superelevation. So the user is able to im-

prove the shape and the mounting of the parts relevant to the heat transfer process. It is also pos-sible to export values to a chart for further analysis.

Figure 4. Section of upper and lower coolant pipe

212

At present, utg operates a test facility to quantify heat-transfer coefficients between the dif-ferent materials used in the heat exchange system.

The configuration of this test facility is similar to a horizontal continuous casting machine. Inthe middle of this system there are uniaxial and quasi-statical conditions of heat transmission.

Different contact conditions and gaps are tested as well as special classifications of material.This new database increases the correlation between the new simulation and the reality in in-

dustrial casting machines.

213

Modeling Continuous Casting of Metal Wire Rods

Chang Hung-Ju1, Hwang Weng-seng1, Chao Long-sun2, Pan Wensen3, Lai Yi Lin3

1 Department of Material Science and Engineering, National Cheng Kung University, Tainan, Taiwan2 Department of Engineering Science, National Cheng Kung University, Tainan, Taiwan3 Metal Industries Research & Development Center

1 Introduction

Because the electrical industrial is developing fast, the requirements for the metal wire rod’squality and size become rigid especial for the metal wire of electrical packaging, which is madeof metal wire rod. Continuous casting is applied in the production of metal wire rod. The pro-cess is similar to that of the steel continuous casting, but the casting speed is slower than steel’s.For controlling the quality, it is necessary to build an analysis model to understand the pheno-mena of heat transfer and solidification in the continuous casting of metal wire rod.

The modeling of continuous casting systems is a problem of a great mathematical and indus-trial significance. Continuous casting involves many complex physical phenomena and up tonow no model can include all of the phenomena at once. In recent years, heat transfer modelsfor continuous casting have been developed [1–3], but most of them studied the continuouscasting of steel. This paper is to develop the heat transfer model of continuous casting for pro-ducing metal wire rods. The numerical method is the finite difference method. The effectivespecific heat/enthalpy method is used to handle the release of latent heat. Casting speed andcooling rate are very important variables for the continuous casting and the simulations for dif-ferent working variables can investigate their effects on the heat transfer behavior of the contin-uous casting. Simulations can give us important output data, such as temperature distributionand the profile of solidified shell. It can help the industrial to shorten the experiment time andseek the best operation condition.

2 Mathematical Model and Numerical Method

The physical model of the continuous casting machine of metal rods is shown in Fig. 1. The me-tal in the crucible is in liquid state due to the heat from the heater around the crucible. The liquidmetal enters the graphite mold from the bottom of the crucible. The graphite mold is surroundedby the cooling system. Because of the heat extraction of the cooling system, the liquid metal be-comes solidified soon after entering the graphite mold. The liquid metal and solid casting movedown together with constant casting speed. To build the mathematical model, the following ba-sic assumptions are made firstly:

1. The system is in the steady state.2. The system is the axial symmetry. 3. The natural convention effect is ignored since the diameter of the metal rod is small.4. The casting speed is constant.5. The thermal conductivity and heat capacity of metal are temperature-dependent.

214

The concerned area of this study is the region of fast temperature variation and solid/liquidtransformation after the liquid metal enters the graphite mold. Since the length of the graphitemold is much larger than its inside diameter, the metal temperature does not have any signifi-cant variation before leaving the graphite mold. Consequently, the graphite mold and the metalinside it are the computing domain. According to the previous statements and assumptions, thegoverning equation (energy equation) can be written as

(1)

where r is the density, V is the casting speed, k is the thermal conductivity. C peff is the effectivespecific heat, including the effect of latent heat. The equation is used in the metal and in the gra-phite mold, the convection term C peff V T/ z does not exist in the energy equation. In thisstudy, because the nonlinear effect of the latent heat makes the iterative computations difficultto converge, the transient equation is solved instead to ease off the nonlinear effect. In solvingthis equation, the effective specific heat/enthalpy method is used to deal with the latent heat.

The boundary conditions of the computing domain are given as follows:

1. At the top of metal, the temperature is fixed and is the temperature of the liquid metal in the crucible.

2. At the bottom of metal, there is no temperature variation, i.e., no temperature gradient.3. At the centerline of metal, it is symmetric.4. At the interface of metal and graphite, there exists a contact resistance and the heat flux is

continuous. The expression of heat flux can be given by

qmetal = qgraphite = hmg(Tmetal–Tgraphite) (2

1Veff T T TC p k r k

z r r r z r

Figure 1: The schematic diagram of continuous casting

215

where hmg is the effective heat transfer coefficient whose inverse is the contact resistance. qmetal, qgraphite, Tmetal and Tgraphite are the heat fluxes and temperatures of metal and graphite at the interface.

5. At the outer surface of the graphite mold, an effective convective boundary condition is app-lied.

q = heff(T – Tref) (3)

where q and T are the heat flux and temperature at the outer surface. heff is the effective heattransfer coefficient which is used to handle the effect of the cooling system. Tref is the referencetemperature and here is the water temperature of the cooling system.The effective specific heat/enthalpy method is the combination of the effective specific heatmethod and the enthalpy method. It owns the advantages of these two methods: fast conver-gence and good accuracy. At one time step, the relationship of temperature, effective specificheat and enthalpy can be written as

(4)

The numerical method is the finite difference method. The backward difference is used forthe time derivative and the centered difference is for the spatial derivatives. The control volumemethod is utilized to derive the difference equations for both the governing equation and bound-ary conditions. Because the difference equation is nonlinear, the iteration method is used tosolve the difference equations.

3 Results and Discussions

3.1 Verification of Numerical Analysis of Continuous Casting

In the literature, many numerical simulations of continuous casting only put the metal in thecomputing domain. This study is based on the experiments from Metal Industries Research andDevelopment Centre (MIRDC) in Taiwan. The temperature data available are from the measu-red points on the surface of the graphite mold. Accordingly, to compare the computing tempera-tures with the measured ones, it is necessary to include the graphite mold and the metal in thecomputing domain. In this research, the interface between the metal and the graphite mold is re-garded as non-perfect contact, which is handled by applying Eq. (2). According to the MIRDC’sexperiments, the working conditions are (1) the metal material is copper, (2) the metal tempera-ture at the inlet of the graphite mold is 1473 K, (3) the height, inside diameter and thickness ofthe graphite mold are 23.5 cm, 0.4 cm and 0.9 cm, (4) the casting speed is 0.667 cm/sec. The ef-fective heat transfer coefficient (heff) in Eq. (3) is 2700 W/m2 which is estimated by using theexperiment data (inlet and outlet temperatures of water and flow rate) from the cooling system.The computing temperature distribution is shown in Fig. 2 and Fig. 3 is the enlarged one nearthe inlet of the graphite mold. From these two figures, it can be found that the primary tempera-ture variation is along the axial direction and the variation in the radial direction is small.

To verify the feasibility of the numerical simulation, the computing results are comparedwith the experimental ones. Since the temperatures in the metal are not easy to measure, there

1

1

n neff

n ne e

C pT T

216

measured points were set on the outer surface of the graphite mold. Their locations are 1, 6 and12 cm from the top of the graphite mold. Table 1 shows the computed and measured tempera-tures of these three locations. The last measured point was arranged at the outlet of the graphitemold. The computed outlet temperature of the metal is 133, while the measured one is 150.From the comparison of these four temperatures, the computed temperatures are close to themeasured ones and therefore the feasibility of the numerical simulation can be verified.

Table 1: Comparison of computing and measured temperatures

3.2 Comparisons of Casting Speed

The working conditions set up in this section are similar to those in the last section, besides ca-sting speed. The casting speed which this research used are 0.4 cm/sec, 0.6 cm/sec and0.8 cm/sec. The temperature distributions along the central axis for different casting speeds areshown in Fig. 4. From this figure, the temperature gradient increases with the casting speed.When the speed is faster, the duration time of metal in the graphite mold is shorter. This makesthe amount of heat taken by the cooling system less and therefore the metal temperature goesdown slower.

3.3 Comparisons of Interfacial Heat Transfer Coefficient

The effective interface heat transfer coefficient can be decided according to the cooling water’stemperature and flow rate. The working condition in the first section is that the casting speed is0.6 cm/sec and the effective interface heat transfer coefficient is 2700 W/m2K. The situation

Numerical Solution Temperature()

Experiment Solution Temperature ()

Graphite surface 1(cm) 125 119Graphite surface 6 (cm) 77 52Graphite surface 12 (cm) 42 36Metal outlet 133 150

Figure 2: Computing temperature distribu-tion of the metal and graphite mold

Figure 3: Enlarged figure of the temperature distribution near the inlet of the graphite mold

217

that the effective interface heat transfer coefficient is raised or reduced is equivalent to that ofraising or reducing the cooling rate. To study the effect of effective heat transfer coefficient,another two heat transfer coefficients 1000 W/m2K and 4700 W/m2K were chosen. The tempe-rature distributions along the central axis for different heat transfer coefficients are shown inFig. 5. From the figure, it can be clearly found that the outlet temperature of the metal decreasesas the effective heat transfer coefficient increases, however, in the upper part of the graphitemold, the temperature curves are very close to one another.

3.4 Comparisons of Materials

The control variable this section is the casting material. The chosen materials are gold, silverand copper. The thermal properties of the metals are shown in Table.2. Temperature distributi-ons along the axial direction for these three metals are shown in Fig. 6. The temperature decrea-se of gold along the axial direction is the smallest since it has the smallest thermal conductivity

Figure 4: Temperature distributions along the axial direction for different casting speeds

Figure 5: Temperature distributions for diffe-rent effective heat transfer coefficient

Figure 6: Temperature distributions along the axial direction for different casting materials

218

and heat capacitances. The silver’s temperature gradient is smaller than either gold or copper,because the thermal conductivity of silver is the highest.

Table 2: Thermal properties of different metals

4 Conclusions

To investigate the heat transfer behavior of the continuous casting of metal rods, a mathematicalmodel is built in this paper. The numerical method is the finite difference method. In the begin-ning of the study, the computing temperatures are consistent with the measured ones and thusthe feasibility of the numerical simulation can be verified. From the computing results, it can befound that the temperature variation in the metal is primarily along the axial direction and mostof the temperature decrease occurs in the upper of the graphite mold. As the casting speedincreases, the temperature gradient increases. When the effective heat transfer coefficientincreases, the outlet temperature of the metal decreases, however, in the upper part of the gra-phite mold, the temperature curves are very close to one another. Among gold, silver and cop-per, the temperature decrease of gold along the axial direction and the temperature gradient ofsilver are the smallest.

5 References

[1] M. Bamberger and B. Prinz: Mathematical modeling of the temperature field in conti-nuous casting. Z. Metall. Vol. 77(1986), No. 4, p. 234–38

[2] S. Louhenkilpi: Simulation and Control of Heat Transfer in Continuous Casting of Steel, Doctoral Thesis, Helsinki University of Technology, Finland, 1995

[3] S. Louhenkilpi, E. Laitinen, R. Nieminen: Real-Time Simulation of Heat Transfer in Con-tinuous Casting.Met. Trans. B. Vol. 24B (1993), August, p. 685–693

[4] J. S. Hsiao, "An Efficient Algorithm for Finite Difference Analysis of Heat Transfer with Melting and Solidification," Numerical Heat Transfer, Vol. 8, pp. 653–666, 1985

[5] J. A. Dantzig, “Modeling Liquid-Solid Phase Changes with Melt Convection”, Internatio-nal Journal of Numerical Methods in Engineering V28, n8, pp. 1769–1785, August 1989

[6] J. P. Holman, “ Heat transfer,” 9th edition, Mcgraw-Hill company, New York, USA, 2002[7] M. Uoti, M. Immonen, and K. Harkki, “Theoretical and Experimental Study of Vertical

Continuous Casting of Copper.”[8] R. Wilson, “A Practical Approach to Continuous Casting of Copper-Based Alloys and

Precious Metals,” IOM Communication Ltd., London, UK, 2000

Material Au Ag Cu

Melting point, 1064.4 961.9 1084.8

CpS, 0.163 0.262 0.386

CpL, 0.166 0.310 0.494

Latent heat, 62.76 104.2 205

Thermal conductivity (W/m K) 317.9 356 334

219

Modeling of Macrosegregations in Continuous Casting of Sn-Bronze

M. Gruber-Pretzler, F. Mayer, M. Wu, A. LudwigChristian-Doppler Laboratory for Multiphase Modeling of Metallurgical Processes, Department of Metallurgy, University of Leoben, Franz-Josef-Str. 18, A - 8700 Leoben, Austria

1 Abstract

Macrosegregations in DC casting may be caused by various reasons. Besides sedimentation andflotation of equiaxed grains, feeding flow, thermal and solutal buoyancy driven flow, and inletflow are potential mechanisms for the formation of macrosegregations. However, the relativeimportance of these phenomena is difficult to estimate. With a two-phase volume averagingmodel the formation of macrosegregations for a DC casting of a Sn-bronze is simulated. Bycomparing the results of four different cases, it turns out that for the conditions applied, feedingflow creates the strongest macrosegregations, namely positive segregations at the casting sur-face and negative segregations in the center. On the other hand, negative surface segregationsform if only thermal buoyancy flow is considered.

2 Introduction

For almost all practical solidification processes inhomogeneous distributions of alloy elementsat the scale of the whole casting, known as macrosegregations, are found. In order to predict theformation of those undesired ‘defects’ numerical methods have been developed intensively du-ring the last years [1–12]. In a recent publication of two of the present authors [13] the co-lumnar-to-equiaxed transition has been modeled with a three-phase volume averagingapproach, where the motion of grains, the melt flow caused by shrinkage and thermo-solutalbuoyancy, and the growth of a columnar front were considered.

In the present paper we have used only the columnar solidification part of the above men-tioned three-phase approach [13]. In this part, the permeable mushy zone is thought to be com-posed of cylindrical ‘dendrites’ with a given primary dendrite arm spacing, 1. Feeding flow,through the mushy zone, as well as thermal and solutal buoyancy flow are accounted for. In or-der to investigate the role of feeding flow as well as of thermal and of solutal buoyancy flow inthe formation of macrosegregations in DC casting of Sn-bronze four different simulations werecompared: (i) considering only feeding flow; (ii) only thermal buoyancy flow; (iii) only solutalbuoyancy flow; and (iv) without considering feeding flow and thermo-solutal convection.

220

3 Simulation Description

3.1 Model Description and Assumptions

Here, only a short outline of the used columnar solidification model is given. For more detailsthe reader is refereed to the original publications [11–13]. The model considers two phases, themelt and the growing columnar dendrites. For these two phases the conservation equations ofmass, species and enthalpy are considered. In addition, the momentum conservation equationfor the melt is solved. The main assumptions of the model can be gathered as follows:

• The thermodynamic for the binary CuSn system is considered by assuming a simplified phase diagram. A constant redistribution coefficient, k, and a constant liquidus slope, m, was used. At the peritectic temperature the solid fraction reaches already 95–98%. Therefore, and because of the fact that the model for the peritectic reaction is still under development, we assume that the remaining liquid solidifies over a small and arbitrary temperature interval.

• Nucleation and growth of equiaxed grains are ignored.• Columnar dendrites are thought to start growing at the mold wall as soon as the temperature

drops below liquidus.• The columnar dendrites are approximated by growing cylinders.• A shell-wise growth driven by diffusion around the cylinder is assumed. • Mechanical interaction between the mush and the flow is calculated via Darcy’s law and the

Blake-Kozeny permeability model.• Corresponding source terms to account for feeding flow and thermo-solutal buoyancy driven

flow is introduced.

3.2 Geometry Information and Boundary Conditions

For the process simulation a casting velocity of Vcast = 1.92 mm / s and a casting temperatureof T0 = 1389 K is considered. Due to the cylindrical shape of the mold an axis symmetric simu-lation has been chosen. The mold is schematically shown in Figure 1a where ( ) gives the po-sition of the nozzle, ( ) indicates the free surface on the top, ( ) shows the upper part of themold which is assumed to be insulating, ( ) shows the lower part of the graphite mold which issurrounded ( ) by a copper–steel mold including a water cooling. In Figure 1b the boundaryconditions are shown. ( ) gives the position of the inlet, where a pressure inlet is considered. Aheat transfer coefficient (HTC) of h = 50 W/m²K and a nozzle temperature of TSEN = 1292 K isconsidered for the submerged entry nozzle (SEN) region. For ( ) HTC and temperature have avalue of h = 50 W/m²K and Tsurface = 325 K. For ( ) almost ideal insulation is assumed withh = 10 W/m²K and Tmold = 1292 K. ( ) has h = 3000 W/m²K and Tmold = 550 K and ( )h = 1000 W/m²K and Twater = 300 K. A constant velocity Vcast is taken for the outlet ( ). Forthe nozzle and for the free surface a slip condition is used. The mold wall is assumed to movewith the casting velocity. Here, a slip condition for the liquid phase and a non-slip condition forthe columnar phase have been chosen. A grid of 9016 cells and 9296 nodes is used. The chosenconditions are the same as used in [14]. As initial conditions, we start with hot melt(Tinit = 1292 K) in rest (Vinit = 0 m/s). The presented results are taken when steady state were re-ached.

221

To estimate the relative importance of feeding flow, thermal and solutal buoyancy flow, andinlet flow the simulation results for four different cases are compared:

• Case A: No thermal and solutal buoyancy flow and no feeding flow is considered. We have used equal densities for the liquid and the solid.

• Case B: Here only solutal buoyancy flow is considered. The solutal expansion coefficient has been chosen to be c = 0.11 wt.%–1.

• Case C: Here only thermal buoyancy flow is considered. The thermal expansion coefficient has been chosen to be bt = 8.6 · 10–5 K–1.

• Case D: For this simulation only feeding flow is considered. So we assumed the liquid and solid densities to be independent of temperature and concentration, but different. During solidification the higher solid density leads then to a shrinkage-induced feeding flow.

4 Results and Discussion

The temperature distributions of the four different cases are shown in Figure 2. In Figure 3 thevelocity fields for Cases A and C are shown. In Figure 4 the mixture concentration defined as

(1)

is compared, again for the four different cases. Here, cl and cs stand for the concentration, fl andfs for the volume fraction and l and 2 for the densities of liquid and solid. In all figures the li-quidus (TL = 1289 K), the solidus (TS = 1230 K) and the temperature where the solidification isthought to be completed (TP = 1072 K) is shown by iso-lines. Within the columnar mushy zone,located between TL and TP, the volume fraction of solid varies from 0 to 1. As we considered themush to be permeable, melt-flow occurs through the mush. However, the flow velocities in themush are much smaller compared to the inlet flow. The incoming melt reveals a velocity as highas Vin = 25 mm/s. This large value is a consequent of the constant outlet velocity, Vcast, and the

l l l s s smix

l l s s

c f c fc

f f

Figure 1: (a) Continuous casting: nozzle; free surface; graphite mold with isolation; graphite and cop-per mold; steel mold with water cooling. (b) Grid and interfaces for boundary conditions (details given in the text).

222

overall mass conservation. The inlet ‘jets’ hit the mold close to the region where first solidforms, bend inwards and form corresponding vortices (Figure 3). Due to the assumed permeabi-lity of the mush, these vortices penetrate into the mushy zone only moderately. Nevertheless,some solute which has segregated close to the dendrite tips is removed from these ‘tip’ regionsand replaced by not segregated ‘fresh’ melt from the jet vortices. The washing away of segrega-ted melt from the ‘jet-mush’ interaction zones (labeled with I in Figure 4) leads to a somewhathigher solute concentration in the bulk of the casting. As a result, weak negative macrosegrega-tions form in those ‘jet-mush’ interaction zones and a hardly noticeable positive macrosegrega-tion forms in the rest of the strand (labeled with II in Figure 4). This ‘jet-mush’ interactionphenomenon takes place in all considered cases. Its strength depends on the strength of the ‘jet-

Figure 2: Steady-state temperature distributions (in K) for the four different cases: (a) without feeding and thermo-solutal buoyancy flow (Case A); (b) only solutal buoyancy flow (Case B); (c) only thermal buoyancy flow (Case C) and (d) only feeding flow (Case D)

Figure 3: Velocity field at the inlet region (scaled in m/s) (a) for the case without feeding and thermo-solutal buoyancy flow (Case A) and (b) the case where only thermal buoyancy flow is considered (Case C)

223

mush’ interaction. If a higher mush porosity would have been chosen or if the jet would havemet the mush more intensively, the amount of washed away segregated melt would have beenhigher and with that the resulting negative surface macrosegregations and positive ‘bulk’macrosegregation would have been stronger. In Case A the described ‘jet-mush’ interactionphenomenon is not overlaid by other phenomena. Here, the flow from the jets and the flowcaused by the withdrawal of the strand are the only machanisms present. However, even in thiscase the weak negative surface macrosegregations and hardly any positive ‘bulk’ macrosegre-gation can be seen with the color scale applied in Figure 4. This scale was chosen to optimallydepict the differences in macrosegregations for the four considered cases.

By comparing the temperature distributions (Figure 2), the mixture concentration distribu-tions (Figure4) and the flow pattern for Case A (Figure3a) and B (not shown) it becomes obvi-ous that for the considered situation solutal buoyancy driven flow is negligible. However,thermal buoyancy driven flow seems to be of more importance (see Figure 2c, 3b, 4c). The meltclose to the mold is cooled down and thus increases its density, which in turn results in a down-wards acceleration and with that in a strengthening of the vortices (Figure 3b). The accelerationby the denser and cooler melt and the strengthening of the vortices has three important conse-quences. Firstly the vortices are moved upwards; secondly cooled down and thus denser meltleaves the vortices and flows further downwards and thirdly melt from deep down in the pool ispressed/sucked upwards in the centre of the middle axis of the cylindrical casting (labeled withIII in Figure 4c). This upwards flow is the reason for the fact that with thermal buoyancy flowthe liquidus temperature in the middle of the casting is located somewhat closer to the SENcompared to the case without this flow (see Figure 2).

It is worth to mention that the ‘jet-mush’ interaction is strengthened by the thermal buoyancyflow and thus more segregated melt is washed out which leads to slightly more pronounced neg-ative surface macrosegregations (Figure 4c). The washed out segregated melt is accumulated

Figure 4: Steady-state distributions of the mixture concentration for the four different cases: (a) without feeding and thermo-solutal buoyancy flow (Case A); (b) only solutal buoyancy flow (Case B); (c) only thermal buoyancy flow (Case C) and (d) only feeding flow (Case D). Bright green represents the initial alloy concentration, yellow positive and blue negative macrosegregations. Flow patterns are also indicated by arrows.

224

mainly in the mush in the centre of the casting, which is the reason for the weak positive mac-rosegregations appearing there (labeled with II in Figure 4c).

Up to now, the surface macrosegregations are predicted to be negative and macrosegrega-tions in the casting center are predicted to be more or less positive. However, if shrinkage-in-duced feeding flow is included and thermal buoyancy driven flow is ignored, the surfacemacrosegregations turned out to be positive and the macrosegregations in the center stronglynegative, so just the opposite! Feeding flow is always directed from the dendrite tip towards itsroots and thus carries segregated into the mush. Since the early work of Flemings in 1967 [15,16], this phenomenon is known to result in a positive macrosegregation at the surface of a cast-ing, the so-called inverse segregation1. Exactly this happens in Case D (labeled with IV in Fig-ure 4d).

At the centre of the cylindrical casting the dendrite tips approaches each other and so theyform a ring which is going to be closed while solidification proceeds. On the other hand, a rela-tively large mush area at the center is solidifying and thus a large amount of melt is needed tofeed the corresponding shrinkage. This melt is sucked into the solidifying mush via the closing‘ring of dendrite tips’ and so a relative strong downwards feeding flow occurs in the center ofthe casting (labeled with V in Figure 4d, see ref. [14]). With that the solidifying dendrites arefed with less- or no-segregated ‘fresh’ melt from the melt pool and therefore a negative mac-rosegregation in the center region of the casting forms. In addition, heat is carried with thedownwards feeding flow which leads to the somewhat lower position of the isotherms, especial-ly at the casting center (Figure 1d). Again, the predicted macrosegregations depend on the per-meability of the mush. If larger mush porosity would have been chosen, feeding through themush would have been easier and the center downwards flow would have been significantlybroader. This would have caused a broader negative center macrosegregation which, however,would not have been so strong.

5 Conclusions

The formation of macrosegregations in DC casting of a columnar solidifying Sn bronze iscaused by the ‘inlet jet-mush’ interaction, the thermal buoyancy driven flow and the shrinkage-induced feeding flow. Conclusions gained from the study of the different macrosegregation for-mation mechanisms can be gathered as follows:

• The ‘jet-mush’ interaction leads to weak negative macrosegregations on the surface and hardly any positive macrosegregations in the bulk of the casting.

• For the investigated situation solutal buoyancy driven flow is negligible.• With thermal buoyancy driven flow the ‘jet-mush’ interaction is stronger, and thus the nega-

tive surface macrosegregations and the positive bulk macrosegregation is more pronounced.• Shrinkage-induced feeding flow causes macrosegregations which are opposite to those

which forms with and without thermal buoyancy driven flow. Namely, positive surface macrosegregation and a strong negative center macrosegregation.

1 Nowadays, inverse (surface) macrosegregation is known to be also caused by an interdendritic flow towards the sur-face accompanied with reheating and remelting during the local formation of a gap between the casting and the mold.

225

• The strength of the macrosegregations is thought to be dependent on the considered mush permeability. Here, further investigations are highly needed.

Although, we did not present predictions on macrosegregations for the case with both ther-mal buoyancy driven flow and shrinkage-induced feeding flow, it is obvious from Figure 4 thatfor the considered conditions the effect of feeding flow will dominate the formation of mac-rosegregations. However, for different conditions the opposite might be true.

6 Acknowledgements

This work was financially supported by the Austria Christian-Doppler Society (CDG) and bythe Wieland-Werke in Germany for which the authors kindly acknowledge.

7 References

[1] M. Rappaz, Int. Mater. Rev. 34, 1989, p. 93[2] J. Ni, C. Beckermann, Metall. Trans. 22B, 1991, p. 349[3] C. Beckermann, R. Viskanta, Appl. Mech. Rev. 46, 1993, p. 1[4] J. Ni, F.P. Incropera, Inter. J. Heat Mass Transfer 38, 1995, p. 1271[5] J. Ni, F.P. Incropera, Inter. J. Heat Mass Transfer 38, 1995, p. 1285[6] C.Y. Wang, C. Beckermann, Metall. Mater. Trans. 27A, 1996, p. 2754[7] C.Y. Wang, C. Beckermann, Metall. Mater. Trans. 27A, 1996, p. 2765[8] C.Y. Wang, C. Beckermann, Metall. Mater. Trans. 27A, 1996, p. 2784[9] C. Beckermann, JOM 49, 1997, p. 13[10] A.V. Reddy, C. Beckermann, Metall. Mater. Trans. 28B, 1997, p. 479[11] A. Ludwig, M. Wu, Metall. Mater. Trans. 33A, 2002, p. 3673[12] M. Wu, A. Ludwig, A. Bührig-Polaczek, M. Fehlbier, P.R. Sahm, Inter. J. Heat Mass

Transfer 46, 2003, p. 2819[13] M. Wu, A. Ludwig, Metall. Mater. Trans., 2005, submitted[14] A. Ludwig, M. Gruber-Pretzler, F. Mayer, M. Wu, Mat. Sci. Eng. 2005 in print[15] M.C. Femings, G.E. Nereo, Trans. Metall. Society AIME 239, 1967, pp. 1449–1460[16] M.C. Femings, R. Mehrabian, G.E. Nereo, Trans. Metall. Society AIME 242, 1967,

pp. 41–49

226

Continuous Casting Simulation: From Solidification and Fluid Flow to the Calculation of Grain Structures

R. EberleWieland-Werke-AG, 89079 Ulm, Germany

1 Abstract

Nowadays the simulation of the continuous casting process is an important tool for optimizingproduction in the foundry. Simulation tools become more and more powerful. The possibilitiesstart from simple calculations of the solidification and the fluid flow in the melt. The next stepis the consideration of mechanical stresses and the combination of thermal and mechanical cal-culations called thermo mechanical coupling. An even more complex task will be the calculati-on of segregation effects or grain structure.

A selection of the above mentioned tasks is presented and the results are illustrated, analysedand compared with plant trials. Thereby the growing demand for an excellent knowledge of thecasting process with increasing complexity of the problem will be demonstrated.

It is important to estimate the validity of the models for an optimization of the casting proc-ess. The prediction power of the different models will be pointed out. Furthermore difficultiesand limits of this ability are discussed.

2 Introduction

The continuous casting process represents a very complex system with a lot of process parame-ters. Often, the parameters are correlated and usually the kind of correlation between them isunknown. These facts make it very difficult to optimize a continuous casting process by planttrials only. For a systematic proceeding one has to make a series of plant trials with the conse-quence of high costs.

This problem can be solved by using simulation techniques to understand the interaction be-tween the different process parameters and to find optimized parameter combinations to be test-ed in a few plant trials.

In this paper typical examples are treated starting from simple tasks like solidification andfluid flow to more complex ones like macrosegregation. The possibilities to influence and to op-timize fluid flow are illustrated in Chapter 3. Chapter 4 deals with the calculation of gap forma-tion between the mould and the strand. For this purpose an iterative procedure is used. Finally inChapter 5 results of macrosegregation calculations are presented.

3 Optimizing Fluid Flow in a Rectangular Shaped Mould

The kind of fluid flow has a large influence on shell formation, heat distribution in the liquidpool and the surface quality of the cast strand. With an unfavourable fluid flow shell formation

227

could be disturbed. Additionally slag from the top of the casting could be drawn into the meltand lead to a bad surface quality. Thus, it is important to optimize fluid flow in the mould.

3.1 The Status of the Process at the Moment

First a typical process is analyzed with Magmasoft®. Several steps have to be done before thecalculation can be run.

The geometry of the mould and the slab together with the nozzle and an inlet has to be con-structed. Then the geometry is meshed. In a third step the process data are programmed. Thedifferent material properties for the mould and the slab are defined, parameters like the amountand the temperature of the water cooling are specified and different boundary conditions are de-termined. Casting velocity and calculation parameters terminate the collection of process data.

A typical geometry is shown in Figure 1. The melt enters the mould through a T-shaped noz-zle and leaves in horizontal direction towards the narrow side of the mould.

The velocity at the outlet of the nozzle is determined through casting velocity and outlet areaof the nozzle.

In Figure 2 results of fluid flow after a calculation time of 150 seconds are shown. The fluidflow can be visualized by tracer particles.

The melt leaves the nozzle towards the narrow side of the mould. At this side it is guided tothe corner of the mould and flows back to the middle of the broad side. While flowing back themelt cools down and descents, preferential in the middle where two streamlines from oppositedirections meet. Thus, in the middle of the broad side a zone with descending melt forms. Thereit is probable that some slag is transported from the surface to the solidification zone and is in-cluded in a region near the surface of the slab.

3.2 Solutions Avoiding Descending Zones

According to the statements above it is advisable to avoid zones with a strong descent. For thispurpose, nozzles of different shape and orientation were tested with the simulation. As one ex-

Figure 1: Geometry of mould and slab Figure 2:. Corresponding fluid flow for T-shaped nozzle

228

ample a nozzle was inclined towards the broad side of the mould. The results from the fluidflow calculation are shown in Figure 3.

The streamlines are reflected at the broad and narrow sides of the mould so that the meltmoves in an ellipse-like way. There is no descending zone any more and the melt is distributedhomogeneously. This configuration should avoid slag to be included into the slab.

4 Gap Formation Between the Mould and the Strand

Considering the continuous casting process, the behaviour between the strand and the mould isan often discussed question. In the upper region of the mould the liquid metal fully wets themould. Then the strand starts to solidify and shrinks. This leads to the formation of a gap bet-ween strand and mould which lowers the heat transfer. The solidified shell reheats, the strandexpands towards the mould and the heat transfer increases again. With unstable casting parame-ters this process can occur several times until the strand leaves the mould.

It is a very interesting task to describe this behaviour within a simulation model.

4.1 Model for the Heat Transfer

First of all we need a model to describe the heat transfer between the strand and the mould. Forthe casting process some kind of parting compound is used which forms a film on the mouldsurface. At first there is no gap formation and the film fills up the space between strand andmould completely. Then the heat transfer coefficient (HTC) is determined by the heat trans-mission through the film. If dfilm is the thickness of the film and film its thermal conductivity weget a HTC of

Figure 3: Fluid flow results with an inclined nozzle

229

(1)

With a larger gap the film cannot fill the space completely and an air gap forms. Then thecomplete HTC composes of three different parts. Through the air gap the heat is conducted byradiation and the heat conductivity of the air. Finally it has to pass the film. Posing all thesemechanisms together we get

(2)

rad describes the radiation, air and film represent the heat conductivity, dair and dfilm the thick-ness of the air gap and the film respectively.

In Figure 4 the HTC depending on the thickness of the gap is illustrated.

At a thickness up to 1 mm where the gap is fully filled with the parting compound, the HTCis large, determined by the width of the film and its thermal conductivity. If the width increasesabove 1 mm, an air gap forms. Then the HTC drops rapidly to low values determined by radia-tion and the heat conductivity of air.

4.2 Iterative Calculation of Gap Formation

With this model the gap formation for a known casting process can be calculated. Because thereis a coupling between the heat transfer and the gap formation an iterative procedure is chosen.

Starting with an arbitrary HTC between strand and mould the solidification and the shrinkingof the strand are calculated. From the results of the shrinking a new characteristics of the HTCcan be determined. This HTC is then used for the calculation of solidification and shrinking in

film

filmd

11 1

air filmrad

air filmd d

Figure 4: Model for the heat transfer coefficient with a maximum film thickness of 1 mm

230

the next loop. That procedure is repeated until a stable configuration for HTC and shrinking isachieved.

This proceeding will be exemplary shown by the continuous casting of billets. As startingconditions a constant HTC of 1000 W/m²K and a maximum film thickness of 1 mm were cho-sen. After some iterations the HTC gets stable. The starting value and the final result of theHTC are shown in Figure 5.

At the top of the casting there is a region with a high HTC. Here is the liquid pool and theshell starts to solidify. After the strand gets a solid shell, shrinkage begins and the HTC is dis-tinctly lower. The reduced HTC in this area causes a reheating of the shell due to the cooling ofthe core. Thus the strand approaches the mould. At the end of the mould a secondary coolingbecomes important, shrinkage is increased (see Figure 6) and a further reduction in the heattransfer is caused.

Figure 5: HTC for the beginning and final results after the iteration procedure

Figure 6: Shrinkage of the strand after the final iteration

231

The development of the air gap is regarded in Figure 6. In principle it is vice versa to the be-haviour of the HTC. Near the top of the casting there is no air gap. The liquid pool and themushy shell completely touch the mould. After solidification a small gap of 0,2 mm developswhich remains nearly constant. The air gap decreases in this case at 80 % of the mould lengthby the repulsing of the strand to the mould.

Although there are some simplifications in the model, this is a suitable approach to describethe dynamic behaviour of a strand in the mould.

5 Calculation of Macrosegregation in CuSn Alloys (Bronzes)

A well known effect for the continuous casting of bronze is the so-called inverse macrosegrega-tion. This means that during the casting process tin enriches in the outer areas of the strand andat the same time decreases in the core. To optimize the quality of a cast product it is importantto achieve a tin distribution over the whole strand which is as homogeneous as possible. Againsimulation provides the possibility to predict element distribution within the strand and optimi-ze process parameters.

5.1 The Model

In this example the tin concentration for the continuous casting of slabs was calculated withCalcosoft-2D. For this purpose a cross section along the narrow side of a rectangular geometrywas chosen.

The relative movement between the solid and the liquid phases is important for the calcula-tion of macrosegregation. For the model two main effects are taken into account. These are thenatural or forced convection and the shrinkage due to solidification. Therefore, first of all solid-ification and fluid flow have to be calculated.

Furthermore additional material data are required to perform the calculation. It is necessaryto know several parameters from a phase diagram like melting point of the base material, eutec-tic temperature, liquidus slope and eutectic concentration.

5.2 Results for Solidification, Fluid Flow and Macrosegregation

Feeding all this into the calculation results for solidification, fluid flow and tin distributionalong the slab were obtained. These results are shown in Figure 7.

All pictures above represent a cross section through a slab. On the right side it is cooleddown by a mould. The edge on the left represents the middle of the slab. There a symmetricboundary condition was applied.

The left part of Figure 7 shows the fraction solid. The change in colour from white to bluedescribes the solidification front. In the middle part the fluid flow is illustrated. It is assumedthat the whole top of the casting represents the inlet. Below the surface the melt flows towardsthe cooled side and descends along the solidification front. A small part of the melt moves backtowards the corner. Below this main flow a curl forms turning in clockwise direction.

232

The last part on the right side shows the tin distribution in the slab. The contour of the solid-ification front, which plays an important role for macrosegregation, can clearly be recognizedwhen compared with the fraction solid. Within the solid region the tin distribution is fixed.Caused by macrosegregation tin distribution differs by –0,14 to +0,26 % from the nominal con-centration. As indicated by the line in Figure 7 a profile of the tin distribution has been extractedwhich is shown in Figure 8.

Figure 7: Results for fraction solid, fluid flow and macrosegregation

Figure 8: Final tin distribution in the solid region of the slab

233

The horizontal line represents the nominal tin content. In the core of the slab the tin content islowered. A few centimetres further the nominal concentration is reached again. Then it staysconstant just slightly above this value and increases close to the surface by 0.25 %. This beha-viour arises from inverse macrosegregation often observed in continuous casting.

6 Summary

In the preceding three chapters a brief survey of some topics which can be treated with simula-tion was given. It could be shown that simulation gives a lot of hints which effects during theprocess could occur. Nevertheless these results strongly depend on the quality of the materialdata used for the simulation. Especially concerning macrosegregation for the calculations oftensome material parameters are not known. Thus, the results must be regarded critically and haveto be carefully tested in plant trials.

234

Mould Temperature Fields during Continuous Casting of DHP-Copper

M. Mäkinen1, M. Uoti2

1Helsinki University of Technology, Espoo, Finland2Outokumpu Copper R&D, Pori (Finland)

1 Introduction

In continuous casting processes the mould is the most complex and critical part, which controlsinitial solidification and surface quality. The amount of heat transfer and its uniformity affect onthe solidification front and liquid pool depth. This front is also influenced by other factors, suchas casting speed, mould design and superheat.

During casting, mould may distort due to the steep thermal gradients and air gap formed be-tween strand and mould due to shrinkage of casting. This air gap formation starts, as solidifyingmetal gets strong enough to withstand the metallostatic pressure. This air gap has been deter-mined to be the greatest single resistance to heat removal. Oscillation marks, which are formedduring every mould oscillation cycle, are small depressions on the surface of the cast metal.They may cause cracking and they can act as a small air gap in the mould area [1–5].

As higher volume and better quality of production is needed, heat flux between strand andmould is investigated by measuring the inflow and outflow temperatures and the throughflowvolume of the mould cooling water. Thermal variations are measured via embedded thermocou-ples. The aim of mould thermal monitoring is to predict breakouts and casting quality, to devel-op casting practise for new grades and casting powders and to determine the influences ofdifferent casting variables on mould heat transfer, heat flux profiles, shell growth rate in themould and temperature uniformity in the mould [6–9].

In this research during continuous slab casting of deoxidized high phosphorus (DHP) copper,the temperature field in the mould plates were measured via embedded thermocouples to getmore precise information how operating parameters, such as casting speed, superheat, coolingwater amount, oscillation of the mould and slab size affect on heat transfer, quality and produc-tivity. Mould plate temperatures at steady state were also compared between different locationsin the mould due to changing temperature field and its fluctuations during operation. Due to thistemperature fluctuation, there is also average temperature and standard deviation counted for in-vestigated thermocouples. The adjustable trial mould of our vertical casting machine consists ofcopper mould and a graphite die.

2 Experimental Procedures

In the industrial scale experiment we have cast DHP-copper (Deoxidized High Phosphorus,99.9 wt-% Cu) slabs with adjustable trial mould. The depth of liquid pool was measured by steelrod. Casting speed, superheat, cooling water amount and temperatures, oscillation of the mouldand slab size were selected as casting parameters. While comparing differences of one casting

235

parameter, others are kept constant unless anything else has been mentioned. These selected ca-sting conditions are presented in Table 1.

Table 1: Casting conditions

Between the strand and the copper mould there are graphite linings, which act as a lubricantand molten copper is covered by carbon powder to prevent oxidation. Therefore no mould fluxis needed. In order to evaluate heat transfer and temperature field in the water-cooled adjustablecopper mould, there were 18 thermocouples embedded in both wide copper plates at several lo-cations. The uppermost thermocouples are 16 cm below meniscus and following rows are 10 cmfrom each other. In this research, only six of them were used due to the scope of this article.These thermocouples locations in wide copper plates with three different mould sizes are illus-trated in Figure 1 a. Figure 1 b shows mould copper plate temperatures and standard deviationfor different thermocouples during casting of DHP-copper. The standard deviation of thermo-couple temperatures particularizes how stable casting conditions are. For the convenience of thecaster, different temperatures are shown in different coloured cells. Mould size in this experi-ment has been 155 1025 mm2 and casting speed was 210 mm/min.

3 Results

The temperature field in the mould wall during casting process varies a lot as can be seen fromFigures 2 a–b. The width of the mould has also some effect on this phenomenon. Temperaturesare slightly higher using wider mould. One reason for this is that with constant casting speed li-

Casting speed 145–225 mm/min

Mould size 120*1085 mm2, 155*1025 mm2, 200*850 mm2

Mould Water cooled Cu mould with graphite liningsCasting temperatures / TS 1140–1150 °C /1084 °CSecondary cooling Narrow face 13–15 m3/h

Wide face 27–53 m3/hOscillation 50–100 rpmOscillation amplitude 20 mm

Figure 1: a) A top and a side view of three different mould sizes with embedded thermocouples, b) Mould tem-perature field and standard deviation of thermocouples during casting with mould size 155*1025 mm2 and casting speed 210 mm/min.

236

quid flow into the mould is greater, and more heat has to be extracted. Highest temperatures onthe mould wall have been gauged usually from thermocouple 6 b. This is due to liquid flow andso called dog bone effect. This dog bone effect means that the slab is a bit thicker on the narrowfaces than in the centre of the wide face.

Broken lines represent wider mould (mould size 200*850 mm2) and it is this size where ther-mocouples 6 b and 6 c are closer to the mould narrow face (see Figure 1a.), where the liquidflow hits. Temperature fluctuation is more visible from thermocouples located near the narrowface (thermocouples 6 b and c in Figure 2 b.) especially with wider mould, but temperature fluc-tuation decreases in the middle of the mould wall (Figure 2 b.). The measured depth of liquidpool was few centimetres higher with wider mould as can be expected. Average temperaturesand standard deviation for these measurements are illustrated on Table 2.

Table 2: Average temperature fields and standard deviations for selected thermocouples with mould sizes 155 1025 mm2 and 200 850 mm2. Casting speed was 180 mm/min.

Casting speed is always optimised to the best yield and quality and it plays vital role for un-interrupted production. Due to possible shell breakout, the evaluation of temperature fields andespecially temperature fluctuations during changing casting speed is essential.

In Table 3 average temperatures and standard deviation for three different casting speeds aredepicted. As results show, increasing casting speed increases average temperatures of thermo-couples. Standard deviation, on the other hand, decreases which means that temperature fieldchanges less than before. The former is due to that more hot metal is coming into the mould aswell as growing shell thickness is decreased. The latter is due to better contact between strandand mould. Here mould size is 120 1085 mm2 and amount of mould cooling water is 25 % lessthan results shown in Figure 2 and Table 2.

155*1025 mm2 N4a N4b N4c N5b N6b N6c

Av. temp. °C 123.6 99.5 61.0 118.1 126.6 89.2

St. Dev. 2.7 1.9 1.2 1.3 1.8 1.7

200*850 mm2 W4a W4b W4c W5b W6b W6c

Av. temp. °C 133.9 101.7 58.4 122.9 134.0 100.9

St. Dev. 1.2 1.8 1.1 2.4 2.2 2.8

Figure 2: Measured temperature as a function of time for mould size 155*1025 mm2 and 200*850 mm2, respec-tively. Casting speed has been 180 mm/min in both cases.

237

Table 3: Average temperatures and standard deviation for three different casting speeds.

Temperature fluctuation during low casting speed is due to air gap formation between mouldand strand that is due to slab local shrinkage. This air gap decreases heat removal from slab sur-face and affects on solidification. After increasing casting speed by 15 %, the liquid pool depthwas also increased approximately by 20 %. After a certain value, where liquid pool depth wasabout 10 centimetres below mould there were no measured changes. The strong secondary cool-ing did strongly influence the temperatures at this point. The effect of casting speed on grainstructures can be seen from cross section samples in Figures 3 a–b. Increasing casting speedchanged orientation of columnar grains upwards and near the narrow face they turned a bitaway from the edge of the slab. The direction of heat extraction is perpendicular to the sumpprofile. With higher casting speed the direction of heat flow will turn more vertical directionand the cooling effect of narrow face at the local point is smaller.

Superheat affects on the solidification structure as well as midway and centreline crack for-mation. This has been investigated earlier with round DHP-copper billet [10, 11]. By examiningthe effect of superheat to the response of thermocouples, the finding was pretty obvious: de-creasing superheat approximately 10 °C decreases average temperature field. Also standard de-viation seemed to decrease, except thermocouple 4 b. With thicker mould and lower castingspeed temperature fluctuation was a bit more with thermocouples 4 b and 4 c. Obviously strandis thicker when less superheat is used. To avoid clogging of the entry nozzle, superheat can notbe reduced too much. The air gap formation is not strongly affected by superheat. It should belimited by other casting parameters such as casting speed and the amount and temperature ofmould cooling water.

Amount of mould cooling water as well as temperature of inflowing cooling water has partic-ular significance to the air gap formation, the shell growth and the solidification structure.These have been also confirmed for round DHP-copper billet [10]. It was visible that increasingthe amount of mould cooling water decreased temperatures of thermocouples that were embed-

4a 4b 4c 5b 6b 6c

145 mm/min Av. temp. °C 105.0 112.8 83.6 109.0 104.7 83.5St. Dev 17.9 15.0 5.2 12.8 10.3 3.6

160 mm/min Av. temp. °C 111.0 133.1 94.7 125.2 120.6 91.6St. Dev 6.8 5.8 4.1 5.0 5.8 3.6

170 mm/min Av. temp. °C 112.7 138.0 96.3 123.9 129.9 94.0St. Dev 1.1 1.8 3.4 2.2 1.9 2.3

Figure 3: Grain structures on cross section samples where casting speed was a) 190 mm/min, b) 220 mm/min

238

ded in the mould plates. Thermocouple 4a is an exception, where temperatures are even highercompared to the situation where there was less mould cooling water. The effect of the amountof mould cooling water was more visible looking at standard deviation values. They were high-er with greater amount of cooling water. After a certain amount of mould cooling water air gapstart to form between strand and mould. This has been noticed from the surface of the slab thatis more or less wavy. The surface of the slab always consists of some amount of small rippleson it. These ripples are oscillation marks and they act as small air gaps.

Oscillation marks are formed on every cycle where mould is moving up and down i.e. oscil-lating. Oscillation frequency has been noted to affect on depth of those marks as well as to thedistance between them [12]. Within our earlier studies it was noted that by increasing oscillationfrequency from 50 rpm up to 100 rpm the average depth of oscillation mark decreased approxi-mately from 0.45 mm to 0.15 mm resulting in a very smooth surface. The response of thermo-couples for different oscillation frequencies was not so clear. From 50 rpm up to 90 rpmaverage temperatures increased while standard deviation decreased. By increasing oscillationfrequency up to 100 rpm situation changed a bit; average temperatures decreased and standarddeviation increased. The former phenomenon was due to better contact between strand andmould with smaller oscillation marks.

4 Summary

Mould temperatures and its deviations during continuous casting of deoxidized high phospho-rus (DHP) copper slab for different casting speed, superheat, amount of mould cooling water,oscillation and slab sizes were determined. With wider mould average temperatures were higherthan with narrow one as well as temperatures standard deviation near the narrow face of mould.The wider mould was narrower than the other ones, which meant that thermocouples are locatedmore close to the narrow face and therefore average temperatures were higher. Liquid pooldepth was increased applying wider mould. Increasing casting speed increases average tempera-tures whereas standard deviation is decreased. Former is due to that more liquid metal is cominginto the mould and thickness of strand is decreased, whereas latter is influenced more via de-creased air gap between mould and strand. Liquid pool depth increased approximately by 20 %when casting speed was increased by 15 %. After a certain value there were no measurablechanges in the liquid pool depth because of heavy secondary cooling. After a certain point theamount of mould cooling water has effect on temperature fluctuation as air gap has started toform. Oscillation of the mould has been noted to influence to the surface quality as every oscil-lation cycle leaves small ripple on to the surface of the slab. These oscillation marks act as smallair gaps. Increasing oscillation frequency decreases the depth of these marks and average tem-peratures are higher with less deviation of temperatures. Liquid flow can be verified with ther-mocouples as can be seen especially from thermocouple 6b where liquid flow hits. It was alsopossible to see if submerged entry nozzle was starting to clog or was misaligned, when left andright hand sides were compared to each others.

239

5 References

[1] J. K. Park, B. G. Thomas, I. V. Samarasekera, U. S. Yoon, Thermal and mechanical beha-vior of copper molds during thin-slab casting (I): plant trial and mathematical modeling, Metal. Mater. Trans. B, 2002, Vol. 33B, pp. 425–436

[2] J. M. Rodriquez, A. Esteva, S. Meza, A note on the control of the solidification front in the continuous casting of copper tubes, J. Mater. Proc. Tech., 1999, 96, pp. 42–47

[3] Ozgu, M. R., Continuous Caster Instrumentation: State-of-the-Art Review, Canadian Metallurgical Quarterly, 1996, Vol. 35, No. 3, pp. 199–233

[4] J. Elfsberg, Oscillation mark formation in continuous casting processes, Licentiate thesis, Stockholm, Sweden (2003)

[5] J. Kron, Air gap formation and hot tearing in solidification processing of Al- and Cu-base alloys, Doctoral thesis, Stockholm, Sweden (2004)

[6] A. S. Normanton, P. N. Hewitt, N. S. Hunter, D. Scoones, B. Harris, Mould Thermal Monitoring: a Window on the Mould, Ironmaking and Steelmaking, 2004, Vol. 31, No. 5, pp. 357–363

[7] M. W. Nichols, Measuring variation in heat transfer in a slab casting mold using embed-ded thermocouples, 1990-PTD Conference Proceedings, pp. 45–52

[8] M. Yamamoto, T. Mizuguchi, Mold temperature measurements during semi-continuous slab casting of copper alloys and influence of operating variables, Proc. Merton Flemings Symp. Solidification and Materials Processing, TMS, 2001, pp. 407–410

[9] G. Xia, H. P. Narzt, Ch. Fürst, K. Mörwald, J. Moertl, P. Reisinger, L. Linderberger, Investigation of mould thermal behaviour by means of mould instrumentation, Ironma-king and Steelmaking, 2004, Vol. 31, No. 5, pp. 364–370

[10] M. Mäkinen, Effect of superheat and cooling on grain structure and crack formation in the continuous casting of DHP-copper, Licentiate’s thesis, Helsinki, Finland, (2004)

[11] M. Mäkinen, M. Uoti, The effect of superheat on micro- and macrosegregation and crack formation in the continuous casting of low-alloyed copper, Will be published by Trans. Tech. Publications in a special issue of Materials Science Forum in spring 2005

[12] M. Mäkinen, The effect of oscillation on the surface of copper slab, TKK-report 2003, pp. 25

240

Simulation of Heat Transfer and Solidification in Continuous Casting of Copper Alloys and the Effect of Fluid Flow

S. Vapalahti, S. Louhenkilpi, M. Mäkinen, P. VäyrynenHelsinki University of Technology, Laboratory of Metallurgy, Espoo, Finland

1 Abstract

A lot of mathematical models have been developed to simulate heat transfer and solidificationin continuous casting. In these models, it is usually assumed that the strand is withdrawn throu-gh the machine with a constant casting speed and the convective heat transfer generated by thefluid flow is taken into account by using an effective thermal conductivity method. At HelsinkiUniversity of Technology, this kind of heat transfer model is developed (TEMPSIMU3D). Itconsists of two separate models, the mould and the strand model. These two models are coupledtogether using a so called gap heat transfer coefficient as a function of the strand surface tempe-rature. A dynamic version, DYN3D, is also developed. The required material data for the simu-lations are calculated using an in-house model called CASBOA. It calculates the material datafor copper binary alloys. These models are presented in this paper.

Coupled models calculate the turbulent fluid flow, heat transfer and solidification simultane-ously. These kinds of models are generally subjected to convergence difficulties and long com-puting times. In the present study, a commercial FLOW-3D package is used to make coupledsimulations to investigate the effects of fluid flow on heat transfer and shell growth to estimatethe justification of the effective thermal conductivity method.

2 Introduction

Continuous casting involves many physical phenomena. The main phenomena are: fluid flow,heat transfer and solidification. Most of the thermal models developed are not calculating thefluid flow at all. In these models, it is assumed that the strand (solid and liquid) is withdrawnthrough the machine with a constant velocity field (= casting speed). The convective heat trans-fer generated by the fluid flow is taken into account by using an effective thermal heat conduc-tivity method. The equation to be solved now is the basic partial differential equation of heatconduction including the removal of latent heat of solidification and other phase transformati-ons. These kinds of models are increasingly being used to simulate continuous casting proces-ses. For steels, mostly one- or two-dimensional models have been developed and successfullyused but in the case of copper, the model must be three-dimensional. Today, more attention hasbeen paid on developing three-dimensional steady state and three-dimensional real-time heattransfer models.

In this paper, a three-dimensional steady state heat transfer model called TEMPSIMU3D anda three-dimensional real-time heat transfer model called DYN3D are presented. The thermo-physical material data for the models are calculated using in-house model CASBOA. These arealso shortly presented in the paper.

241

Coupled models calculate the fluid flow, heat transfer and solidification simultaneously. Thecoupled models include three main equations: the energy equation for temperature, the Navier-Stokes equation for velocities (momentum) and the mass equation for mass. If the flow is turbu-lent, a turbulent model must be added. The most commonly used turbulence model is so-calledk- turbulence. This increases again the complexity of the system, because two additional equa-tions must be solved. One example of the coupled calculation is presented and the effects offluid flow on heat transfer and shell growth is discussed.

2 TEMPSIMU3D and DYN3D - Heat Transfer Models for Continuous Casting

TEMPSIMU3D and DYN3D are 3-dimensional heat transfer models for continuous casting.The first is for steady state and second for transient casting conditions with a capability to beused in real-time. The model equations include the transient term but in the case of steady statemodel, this term is neglected. The models consist of two separate modules: the mould modeland the strand model. They are running iteratively and so called gap heat transfer coefficient asa function of the strand surface temperature is used to couple them. Models are based on finitedifference method where upwind scheme is used.

The strand model simulates the strand temperature field and the temperature related datathree-dimensionally using Eq. 1 (the term H/ t is neglected in steady state model).

(1)

Here H is the enthalpy, t is time, keff is the effective thermal conductivity, T is temperature, is density and is the actual casting speed. Enthalpy function H includes also the effect of allphase transformations.

The convective heat transfer in the liquid pool due to the liquid flow is described with the ef-fective thermal conductivity according to Eq. 2.

(2)

A is constant and fs is solid fraction in the mushy zone. The boundary condition for the strandin the mould is given by Eq. 3.

(3)

Here hgap is the gap heat transfer coefficient, T is the strand surface temperature and Text is theexternal temperature, in this case the temperature of the mould surface. Eq. 3 is also used as theboundary equation for the hot side of the mould, where hgap is the same gap heat transfer coeffi-cient as for the strand model but Text is now the temperature of the strand surface, which is cal-culated by the strand model. For the colder side, similar equation is used, but Text is thetemperature of the cooling water and h is a constant. It is assumed that there is no heat fluxthrough the top and the bottom surfaces of the mould.

( ) ( ) ( )eff eff effH H T T T

v k k kt z x x y y z z

(1 )eff S Sk k f A k f

( )gap extT

k h T Tn

242

The mould surface temperature is calculated by the mould model using Eg. 4 (the termc T / t is neglected in steady state model).

(4)

The model allows defining three different material layers in the mould. This is especially im-portant in the case of continuous casting of copper where there is usually at least a copper jacketwith a graphite die.

In the secondary cooling zone after the mould the model has been divided into calculationdomains between a pair of support rolls. This domain has been divided into four different cool-ing regions that are: roller contact area, pre-nozzle area, spraying area, after spray and pool wa-ter area (post-nozzle area).

Roller contact area indicates heat conduction from strand to roll. Roller heat transfer dependson the roll contact length in casting direction as well as the roll cooling mode. Eq. 5 is theboundary equation for the roll cooling.

(5)

Pre-nozzle and post-nozzle areas account for the indirectly cooled space between roller pairs.The boundary equation for these areas is defined in Eq. 6.

(6)

Here h is the heat transfer coefficient for air convection, which is the minimum value, usedfor all exposed surfaces in the machine. Text is the external temperature, is the emissivity, andTair is the air temperature. For post-nozzle area heat transfer coefficient for flowing water isused. This takes into account the heat transfer to the flowing water on the strand after nozzles.

In the spraying area, the strand is usually cooled by a spray of water or water-air mixture.Nozzle parameters like air and water flow have an effect on the cooling efficiency. The mecha-nism of cooling can be subdivided into heat transfer by radiation and convection. The radiationheat flux parameters, emissivity and the air temperature are the same as defined for the pre-noz-zle area. The heat transfer coefficient for the water spraying area is defined by the Eq. 7.

(7)

Here h is the water spray heat transfer coefficient, W the water flow rate, and a, b and c(T)are parameters. The parameter c(T) is a function of temperature and it takes into account the ef-fect of the strand surface temperature on the heat transfer coefficient (Leidenfrost effect).

2 2 2

2 2 2T T T T

c k k kt x y z

( )extTk h T Tn

44( ) ( )ext airT

k h T T T Tn

( )bh a W c T

243

3 CASBOA Software for the Calculation of Material Data

To obtain reliable results from the heat transfer simulations, accurate thermophysical materialdata are needed. Usually these data are obtained from literature but very seldom all requireddata is found. The use of inaccurate material data can lead to considerable errors in calculationsas presented in [1]. Typical data needed are the density, the thermal conductivity, the specificheat, the phase transformation temperatures and the corresponding latent heat with the informa-tion how the latent heat is released during the phase transformations.

CASBOA (Copper Alloys Solidification for Binary One-solid-phase Alloys) is a thermody-namic-kinetic solidification model for the simulation of solidification phenomena of copper al-loys [2]. Currently the calculation can be done for pure copper and for 14 binary copper alloysfor Cu-X, where X is Ag, Al, Cr, Fe, Mg, Mn, Ni, P, Si, Sn, Te, Ti, Zn or Zr. Multicomponentversion is under development.

4 Results

The parameter A in the Eq. 2 generally has a value between 1 and 8. The effective thermal con-ductivity method also assumes that fluid flow increases liquid pool heat transport isotropically,which is not the case, because fluid flow only contributes to heat transport in the flow direction.In this work the effect of the value of parameter A was studied in a copper billet casting ma-chine. The study was carried out using a commercial CFD package FLOW-3D. Both fully cou-

Figure 1: Liquidus and solidus isotherms for flow, A = 1 and A = 7 respectively

244

pled and effective thermal conductivity method calculations were performed in 2D cylindricalcoordinates. In the coupled calculations RNG (renormalized group theory) -model is used as theturbulence model.

Test calculations were carried out using A = 1, and A = 7 in Eq. (2) and these results werecompared with the fully coupled calculations. The results are presented in the Fig. 1. and Fig 2.

As can be seen, the parameter A has influence on the results, but more on the liquidus iso-therm than on the solidus isotherm. This means that accurate results within the liquid pool cannot be obtained. From the results it is possible to see a trend that if a constant value is used, toohigh a value leads generally to a thinner shell in the mould and to shorter pool length. Too smalla value leads to thicker shell in the mould and too long a pool length, correspondingly. The pa-rameter A has a minor influence on the solidus isotherm or on the temperature of the solid strandif the value is not badly over or under estimated as can be seen from the results. To over esti-mate the value to 7 leads to shell thickness to decrease by 15 centimeters near the mould wall.Models using an effective thermal conductivity method can be applied to study the temperaturesin the solid shell and related data such as the shell growth and the location of the liquid pool endposition. The value of A depends strongly on the geometry of caster, casting speed and it is dif-ferent in the mould than in the lower part of the machine.

Figure 2. Solid fraction profiles at the centre of the strand and next to the mould wall respectively

245

5 Conclusions

At the Helsinki University of Technology, 3-dimensional steady state (TEMPSIMU3D) and dy-namic heat transfer model (DYN3D) for continuous casting are developed. The models consistof two separate modules: the mould model and the strand model. They are running iterativelyand so called gap heat transfer coefficient as a function of the strand surface temperature is usedto couple them. The models are validated by industrial measurements. Models using an effec-tive thermal conductivity method can be applied to study the temperatures in the solid shell andrelated data such as the shell growth and the location of the liquid pool end position, but thetemperatures in the liquid pool cannot be calculated very accurately.

6 References

[1] S. Louhenkilpi, M. Uoti, H. Kytönen and S. Vapalahti, Effect of Thermophysical Material Data on Heat Transfer in Continuous Casting, Modeling of Casting, Welding and Advan-ced Solidification Processes X, Destin, Florida, USA, May 25-30, 2003, p. 733–740

[2] J. Miettinen, Simulation of Solidification and Calculation of Thermophysical Properties in Binary FCC Copper Alloys-Revised version, Helsinki University of Technology Publi-cations in Materials Science and Metallurgy, TKK-MK-118, 2001

246

247

Micro / Macro Structure

248

249

Spray Forming and Post Processing of Superalloy Rings

V. Uhlenwinkel1, N. Ellendt1, M. Walter2, J. Tockner3

1 University Bremen, Bremen, Germany2 Böhler Edelstahl, Kapfenberg, Austria3 Böhler Schmiedetechnik, Kapfenberg, Austria

1 Introduction

Superalloys are mainly used in airplane engines and power plants (for gas turbines) due to theirsuperior properties at high temperatures relative to conventional alloys. The powder metallurgyroute is the standard manufacturing process for high-performance aircraft engine components.However, high cost, risk of contamination, and high oxygen content are disadvantages of thisroute. Spray forming as an alternative production route was investigated in the past using diffe-rent superalloys (IN100, IN718, Rene 95, Rene 88DT) [1-8].

Nitrogen or argon was used as the atomization gas. With nitrogen a lower level of residualporosity in the as-sprayed material was achieved because the nitrogen is not inert and reactswith some alloying elements creating nitrides or carbonitrides [1], which affect the precipitationbehaviour and finally lead to poor mechanical properties. Therefore argon is preferred as the at-omization gas even though the use of argon leads to a higher porosity. It is important to mini-mize the argon porosity because the initial porosity can affect the material properties of the finalproduct. Therefore, this paper is focused on the optimisation of porosity in spray formed super-alloys rings and to improve the understanding of the process.

2 Experimental Procedure

In principle, the spray forming unit used is shown in fig.1. The molten metal (IN718 or U720)was superheated (app. 150 K above liquidus temperature) in the crucible by induction. Subse-quently, the melt (150 kg) was pulled into the preheated tundish and left the nozzle as a cylind-rical melt stream at the bottom of the tundish. This molten metal stream was atomized with ahigh velocity gas flow generated by a scanning free fall atomizer (frequency 15 Hz). The sub-strate was preheated to app. 1150 °C by an induction coil inside the rotating substrate (rotationfrequency 1.2 Hz). To optimize the porosity more than 30 spray runs were carried out varyingthe parameters shown in Table 1.

During the spray run several process parameters were recorded. The surface temperature ofthe substrate and the deposit were measured using a two colour pyrometer (company: Bartec,type: R 2520 - 80). Density measurements were carried out with cubes (app. 1000 mm3) usingthe buoyancy method (DIN EN 6018) in order to calculate the porosity. Additionally, porositydata was obtained using image analysis of micrographs, which gave a better spatial resolution.Both measurements gave similar results despite the first method being based on a volume andthe second on an area.

250

Table 1: Variation of process parameters

The as-sprayed rings were machined, divided into four parts and post processed following

three different routes: a) Hot isostatic pressing (HIP) b) Forging c) Hot isostatic pressing andforging to eliminate the residual porosity. The process parameters are summarized in Table 2.

Table 2: HIP and Forging parameter of the thermomechanically processed quarter-rings

Atomizer gas Argon, nitrogen

Nozzle diameter mm 6.0–6.5 Melt flow kg/h 800–1570 Gas pressure MPa 0.27 to 0.43 Gas flow kg/h 580–900 GMR 0.43 to 1.01 Scanning angle ° ±8.5 to ±10.5

Alloy IN718 U720

HIP temperature °C 1140 1140 HIP pressure MPa 100 100 HIP Soaking time H 3/6 3/6 Forging temperature °C 1000 1100 Soaking time H 2 2 Ram speed mm/s 10 10 No. of heats – 1 2

Figure 1: Sketch of the spray forming plant

251

3 Results and Discussion

The use of nitrogen as atomization gas led to pick-up of app. 500 ppm nitrogen in IN718. Evenif it would be possible to get lower porosity with nitrogen this was not followed up because thehigh nitrogen content was not acceptable.

Also it was observed that process parameters must be changed if argon is used instead of ni-trogen because the cooling effect of argon is much lower (more details are reported in [10]).

The scanning of the atomizer led to a typical ring shape displayed in fig. 2. The picture onthe left side shows a ring in the as-sprayed condition and on the right side a cross-section isshown. High porosity areas at the left and right margins were not investigated. The core of therings exhibited a low porosity (see fig. 3). The success of the optimising process is demonstrat-ed by fig. 4 which shows the decrease of the average porosity together with the standard devia-tion. The average porosity includes measurements across the whole thickness of the ring. Stepby step the porosity and the standard deviation was reduced by appropriate selection of sprayparameters and finally the average porosity of the argon sprayed IN718 was app. 0.75 vol.%.

Figure 2: (a) Spray-formed ring (IN718) and (b) cross-section

Figure 3: Porosity distribution in an IN718 ring (run 341, buoyancy method)

252

The boundary conditions of the droplets and the deposit during the impact are responsible forthe as-sprayed porosity. Therefore a correlation between process parameters (e.g. Melt flow,gas flow, …) and porosity is not very meaningful. Investigations with other base metals have al-ready indicated that the deposit surface temperature during the spraying process is the key pa-rameter concerning the porosity [11]. Therefore, the measured deposit surface temperature wasplotted against the local porosity in fig. 5. A low deposit surface temperature caused a high po-rosity (cold porosity) and a minimum porosity was observed in the temperature range between1240 and 1270 °C. The deviation of the porosity at a constant surface temperature may be in-

Figure 4: Optimization of porosity in argon sprayed IN718 rings (image analysis)

Figure 5: Dependency between local porosity (buoyancy method) and deposit surface temperature of spray-for-med rings (IN718 and U720 using argon or nitrogen), melt flow is given in the legend

253

duced by other parameters but this could not be differentiated. Finally, it can be established thatthe deposit surface temperature is the most important parameter to control the porosity.

Reproducibility of the process is a major need for industrial application. During the wholespray run of approximately 10 minutes the surface temperature was kept constant. This is dem-onstrated in fig. 6 which shows the surface temperature for different spray runs. The substratetemperature is monitored from the beginning of the operation. It decreases due to the turn on ofthe atomization gas. The temperature increased rapidly after the spray starts and is nearly keptconstant in a range of 1250 to 1260 °C, which finally led to a uniform porosity distribution ver-sus the thickness of the ring.

To exemplify, this porosity distribution versus ring thickness is shown in fig. 7 in compari-son to some data from literature [9]. The melt was IN718 and the atomizer gas was argon. Theporosity close to the substrate is completely different due to the preheating of the substrate. Inthis study, cold porosity in the vicinity of the substrate was inhibited totally. The area with a lowamount of porosity after [9] was still double as high as the porosity achieved in this investiga-tion. This can make a large difference for the final material properties after post processing.

To improve the material properties the porosity must be closed completely. Forging with theparameters mentioned in table 2 was not sufficient to close the porosity.

In contrast, hot isostatic pressing (HIP) led to a successful reduction of porosity for theIN718. The as-sprayed sample before HIP had an average porosity of 0.85 Vol.%, after HIP ofthree hours it was nearly zero and after 6 hours it was totally dense (see fig. 8). At the moment itis not clear what happened to the argon in the gas pores. A few ppm of argon were measured indifferent samples, but this did not affect the tensile properties (UTS, 0.2%YS, elongation) as re-ported in [12].

The U720 was HIPed as well using the same parameters, but the densification of this materi-al was not enough to eliminate the porosity completely. The reason for this result is not clearand new experiments are prepared with varying HIP parameters.

Figure 6: Reproducibility of substrate and deposit surface temperature during different spray runs with IN718

254

4 Conclusion and Outlook

Superalloy rings IN718 and U720 with a maximum diameter of 500 mm were spray formedsuccessfully. Using nitrogen for atomization led to an increase in nitrogen content (up to 400-500 ppm) which was not acceptable. Therefore, argon was used for the optimization process.After several optimization steps an average porosity of 0.75 vol.% was achieved. The porositywas uniformly distributed versus the thickness of the ring and cold porosity in the vicinity of thesubstrate was avoided due to the preheating of the substrate. A strong correlation between thedeposit surface temperature and the porosity with an optimal range between 1240 and 1270 °Chas been established. The reproducibility of the deposit surface temperature was extremely

Figure 7: Local porosity in as-sprayed IN718 rings from [9](marked as Whi-96) and this study (run 354), atom-izer gas is argon

Figure 8: Porosity of spray formed IN718 before and after HIP (parameter s. table 2, run 307)

255

good which is important into achieving low porosity. The porosity of the IN718 could be closedby HIP with moderate parameters. The same HIP parameters were not sufficient to close the po-rosity of the U720. New experiments are in progress to investigate the influence of the HIP pa-rameter to close the porosity.

Due to the high impact of the porosity in superalloys for aero engine applications, further in-vestigations are necessary. In the future, a more detailed characterization of the porosity interms of size and shape are planned. Moreover, there is an open question about the habitation ofthe argon which is measured in the HIPed material but cannot be seen optically.

5 Acknowledgement

The authors are grateful to the European Union which supported the work under the Growthprogram with the project number G4RD-CT 2002-00762

6 References

[1] M.G. Benz, et al., Proc. ICSF-II (1993), Swansea, UK, 171–181[2] Zhang et al., Proc. SDMA 2000, Bremen, Germany(2000), S. 161–170[3] M.D. Barratt, A.L. Dowson, M.H. Jacobs, Mat. Sci.& Eng. A 383(2004) 69–77[4] W.D. Cai, E.J. Lavernia, Mat. Sci.& Eng. A226-228 (1997), 8–12[5] M.K. Hedges, A.P. Newbery, P.S. Grant, Proc. SDMA 2000, Bremen, Germany,

379–394[6] R.M. Jones, et al., Proc. ICSF III (1996), Cardiff, UK, 71–78[7] R.S. Minisandram, et al., Mat. Sci.& Eng. A326 (2002) S. 184–193[8] S.D. Cai, J. Smugeresky, E.J. Lavernia, Mat. Sci. & Eng., A241 (1998), p. 60–71[9] E.D. Whitton, P.S. Grant, D. Bryant, Proc. ICSF III, 1996, Cardiff, UK, p. 89–99[10] V. Uhlenwinkel, R. Attwater, L. Achelis, M. Walter, Proc. PM 2004, Wien, Austria,

Vol.5, S. 33-38[11] V. Uhlenwinkel, M. Buchholz, R. Tinscher, A. Schulz, J. Fischer, R. Schröder, Proc. 4th

Int. Conf. On Spray Forming, 11.-13. Sept. 99, Baltimore, USA[12] O. Caballero, D. Fournier, W. Smarsly, Proceedings of Superalloys 718, 625, 706 and

Various Derivatives, 02-05 Oct. 2005, Pittsburgh, PA, USA

256

Macro- and Microstructure of Spray-Formed Tin-Bronze

D. V. Kudashov, H.R. Müller, R. ZauterWieland-Werke AG, Ulm

1 Abstract

The first step in the production of metallic materials is melting and casting. Macrostructure,microstructure, and segregation are influenced by the production process. For special materialswith strong tendency to segregation powder metallurgical processes are suitable methods withcomparatively high production costs. Less sophisticated systems can be cast in permanentmoulds or by continuous casting, which is considerably cheaper than powder metallurgy. Sprayforming is a process between these extremes considering both aspects. For example the produc-tion of Nb3Sn-super-conductors needs homogeneous material with minimized segregation.Spray-formed bronzes with tin content up to 17 % fulfil these demands.

Macro- and microstructure of spray-formed tin-bronze is compared with conventionally castalloys. Possible defects like porosity, the phenomenon inverse segregation, and micro-segrega-tion are presented.

2 Introduction

Copper tin alloys, bronze, are one of the eldest technical alloys invented by mankind. Figure 1shows a phase diagram of copper tin [1]. In the liquid phase tin is completely soluble in copper.In the solid state the maximum solubility of tin in copper is 15.8 % at 520 °C. With decrease intemperature the solubility is decreasing which does not have practical consequences because ofthe low diffusion coefficient of tin inside the copper matrix the equilibrium content of solutedtin is only achievable with extreme long annealing periods. During solidification of bronze, the-re are many factors disturbing the equilibrium state. Therefore, a technical two phase diagramhas been found to describe suitable phases (dotted line in Figure 1). According to this diagramcopper tin alloys with more than 6 % tin in the solid state at room temperature contain -Phaseand -Phase. This microstructure is due to the slow transformation kinetics. With long annealingperiods it is possible to achieve the equilibrium microstructure. This annealing process is notable to remove the inverse segregation which appears after continuous casting. Its effect is ahigher tin concentration at the edge of the billet than in the centre which does not correspond tothe phase diagram. Due to the inverse segregation, the further cold and hot working of coppertin alloys is restricted to bronzes with tin content less than 8 %. With the spray forming processbronzes with tin contents up to 17 % Sn are produced which can be further hot and cold formedwithout any prior homogenization process. The microstructure of spray-formed high tin bronzediscussed in the following chapters. It is shown that the optimum microstructures are onlyachievable in a small range of process parameters during spray forming.

257

3 Spray Forming

3.1 Process

The Wieland-Werke AG produces billets of high-tin bronzes by spray forming (scheme s. Figu-re 2). The melt is prepared in vacuum furnace with additional stirring coil. Thus the melt is ho-mogeneous and poor of soluted gas. The melt jet is then dispersed to small droplets by an inertgas. The average droplet diameter is about 60 μm. Comparing to the metal powder production,the mass flow rate is significantly higher, namely up to 35 kg/min with a single atomizer and upto 70 kg/min with a twin atomizer. Growth direction is vertical. The Wieland spray forming de-vice and process are described in detail in [2].

3.2 Influence of Process Parameters on Quality of the Spray Formed Billet

The spray forming process is specified by numerous parameters, which influence shape, micro-and macrostructure, porosity, segregation of the billet. Additionally they interact in differentways. For example the diameter of the casting nozzle and the melt level in the tundish define themelt flow rate (Torricelli formula). Figure 2 outlines the most important process parameters.The preset parameters are defined by the design of the plant or can not be changed during therun. In principle, the control parameters can be changed, but practically only the gas mass flowrate and the withdrawal speed w are controlled. At the Wieland plant the melt temperature andthe gas to metal flow rate ratio (GMR) are automatically controlled. Normally scan frequency

Figure 1: Phase diagram Cu-Sn [1]

258

and substrate rotation are kept on a default value. The operator adjusts only the withdrawalspeed to keep the billet diameter constant. This is necessary, if the metal flow rate or the com-pacting rate changes.

The parameter GMR influences the atomization (droplet size and velocity distribution in thespray cone), the compacting rate (ratio of atomized melt mass to deposited mass), the cooling ofthe droplets and the cooling of the deposit. In the following the effect on structure of high al-loyed tin bronze billets is discussed.

3.3 Porosity

A higher GMR causes the reduction of the mean particle diameter in the spray cone. Thereforethe fraction of solidified particles is increased and the heat entry into the deposit is reduced. Atthe time when the particles hit the target only a small percentage of particles is still liquid. Thispercentage of particles is too small to fill up all gaps between the solid material with liquidwhich causes pores. So with increasing GMR porosity does increase (compare Figure 3a, b).

Figure 2: Spray forming process parameters

Figure 3: CuSn13,5 a)GMR 0,53, porosity: 3,6 %, b) GMR=0,47, porosity: 1,2 %

259

Another important factor is the chemical composition of the melt. Figure. 4a shows the struc-ture of CuSn15 with some pores. At similar spray conditions, but with 0.2 % Ti the structure isnearly 100 % dense (Figure. 4b). Influence of reactive elements on the porosity are analysed indetail in [4].

So it seems to be very easy to avoid porosity by decreasing gas-flow rate and reactive ele-ments – if there would not be segregation.

3.4 Segregation

The driving force for the macro-segregation phenomenon is assumed to be the centrifugal forceof the rotation of the billet during the spray forming process [4].

Figure 5 demonstrates the Sn – concentration pattern in the cross section of CuSn12 spray-formed billets. The lower curve shows the segregation across the diameter at a low, and the up-per at a high gas flow rate at the same melt flow rate. The amplitude of tin content variationsacross the cross section is decreased by increased GMR.

On the other hand the increase of GMR causes an enhanced microsegregation. This micro-segregation is present in the form of -Phase agglomeration (Figure 6a). In the microscopic viewof a microetched billet cross section they are visible in shape of concentric circles inside thespray-formed material. In the longitudinal section the circles turn out to be 3 dimensional dome

Figure 4: Influence of Ti-content on porosity in CuSn15,5 at constant gas-metal-ratio G/M = 0,5; a) without Ti; b) with Ti

Figure 5: Sn-concentration in the cross section of CuSn12 billets spray-formed with different gas flow rates G in m3/s and constant melt flow rate

260

shaped agglomerations (Figure 6b). This dome shape obviously is the image of the surface ofthe billet in process at a certain time during the spray forming process. By optimizing the proc-ess parameters all these micro-segregation phenomena are avoidable.

4 Differences from Conventional Casting Process

4.1 Macrostructure

Figure 7a and 7b illustrate the macrostructures of spray-formed and cast bronzes. While spray-formed bronze shows fine homogeneous grains in the longitudinal direction (Figure 7 a), thegrains in continuous cast bronze are coarse and inhomogeneous (Figure 7, b).

4.2 Microstructure

The difference in microstructure between cast and spray-formed bronze with 15.5 % tin is illu-strated in Figure 8. The permanent mould cast shows a dendritic structure. Between the dendri-tes the tin-rich -phase is enriched (Figure 8, a). The high fraction (about 30%) of this brittle, lowmelting phase prevents cold and hot forming. In the spray-formed structure (Figure 8, b) the

Figure 6: Spray-formed 13,5 %-tin bronze with microsegregation a) microstructure b) macrostructure

a) b)

Figure 7: a) spray-formed 13,5 %-tin bronze with fine and homogeneous grain structure, grain size about 60 μm, b) continuous cast 12 %-tin bronze with course and inhomogeneous grains [3]

b)a)

261

fraction of -phase amounts to approx. 5 % and particles are not connected. This structure al-lows cold and hot forming.

4.3 Segregation

In continuous casting of high tin containing copper alloys the so called inverse segregation iswell known. In early days of the continuous casting development various publications dealtwith this phenomenon [6, 7, 8, 9]. Sucking by volume contraction and metallostatic pressure aremost likely explanations. The strong influence of metallostatic pressure was impressively de-monstrated by Ohm [10]

Figure 9 shows a plot of tin-concentration in dependence of the location inside a continuouscast bronze billet with 8 % tin and with a diameter of 7 inch. The tin concentration at the billetsurface is considerably higher than in the center. The tin concentration differences reach valuesup to 9 %. This macrosegregation is not removable by heat treatment [1].

During spray forming process hydrostatic pressure does not exist. If spray conditions are op-timized, macrosegregation is reduced to a concentration difference of about 1 wt-% inside thebillet. Figure 5 shows a concentration plot of a spray-formed 12%-tin bronze at optimized spray

Figure 8: Microstructure of 15.5 %-tin bronze produced by a) spray forming and b) die casting. [3]

a)

Figure 9: Tin concentration gradient in continuous cast 8 %-tin bronze billet, diameter 178 mm, cross section [11]

b)

262

parameters. Macrosegregation is limited to 1 wt-% tin. The driving force for this small segrega-tion phenomenon is assumed to be the centrifugal force of the rotation of the billet during thespray forming process [5].

5 Application

5.1 Superconductor

Superconducting wire [2, 3] with magnetic field strength up to 22 Tesla are usually low tempe-rature superconductors which run at liquid helium temperature. The super-conducting materialis Nb3Sn, produced by the bronze method. This type of magnet is applied in Nuclear MagneticResonance (NMR) spectrometers for analytical applications and in magnets for nuclear physics.

Nb3Sn is a brittle inter metallic phase, and so it is not a metallic material which can be pro-duced and brought into the shape of a wire by normal industrial melting, forming and machin-ing processes. It must be produced by co-working of the metals Niobium and bronze until thecompound has the shape of a wire. Finally a diffusion annealing process generates the Nb3Snsuper-conducting phase. The bronze serves on one hand as carrier for tin, on the other hand asheat and current carrying material in case of super-conductivity break-down.

The manufacturers of superconducting wire need a good workability of the tin-carryingbronze. By means of forming processes the tin is brought as near as possible to the niobium.The final heat treatment lets tin diffuse to the niobium and transfers both into the Nb3Sn phase.

In order to show the long route from material production to the final product, the productionprocess of superconducting wire is illustrated in Figure 10 in six separate steps:

1. Spray forming of bronze billet2. Extrusion and cold drawing of bronze tubes and rods3. Assembling bronze tubes and niobium rods

Figure 10: Production process of bronze for super-conducting Nb3Sn wire (schema) [2, 3]

263

4. Co-working of Niobium and bronze (extrusion, cold-drawing, heat treatment) into the shape of a wire

5. coiling of wire into final shape inside the magnet6. Generation of Nb3Sn by final annealing process.

During this process a multi-filamentary composite material is generated. Figure 11 shows across section through a stabilized type of superconductor wire in a production stage cold-draw-ing of the niobium-bronze-composite (production step 4). At a first glance 55 hexagonal greydots surrounded by bronze are visible. Each dot consists of 85 niobium filaments which also aresurrounded by a bronze matrix. The dimensions of niobium filaments and bronze channels fi-nally reach the size of 5 to 20 μm. The final production step is diffusion heat treatment.

5.2 Machinable Materials for Connectors

Between the standard machinable copper alloys and the machinable, high-strength but expensi-ve beryllium copper CuBe2Pb exists a gap in term of mechanical properties and in price, whichcan be closed by alloys of the system Cu-Sn. Wieland-Alloy CuSn13,5Pb shows an unusualcombination of high yield strength (about 900MPa ) and low modulus of elasticity (85 GPa).This combination of properties allows employing the alloy for the spring application (connec-tor). Good free cutting characteristics are achieved by the addition of Pb. Due to the spray for-ming process a fine and homogeneous lead distribution is obtained. Free cutting is substantiallyimproved in comparison to the lead free alloy CuSn13.5. The lead containing alloy exhibits amore favourable chip shape (see. Figure12). Mechanical properties are not influenced by thelead addition.

Figure 11: Stabilized superconductor with 55 x 85 = 4675 niobium filaments in Cu-Sn-matrix, wire diameter 1 mm (by courtesy of EAS/Hanau)

264

6 Conclusion

The spray forming process is specified by numerous parameters, which influence the propertiesof the materials. The influence of GMR on micro- and macrostructure, porosity and segregationof high alloyed tin bronze billets is discussed. The process parameters have to be set very preci-sely to get minimized segregation on one hand and limited porosity on the other hand.

Spray-formed high-tin bronzes have various advantages compared to cast high-tin bronzes.The grain structure of spray-formed material is fine and homogeneous. Macrosegregation isminimized to 1 % tin concentration difference across the billet. Owing to these advantages thespray-formed bronze with tin content up to 17 % can be hot extruded and cold drawn withoutany prior homogenization.

Spray forming is a modern production process which is able to shift the classical border be-tween wrought and cast alloys to considerably higher tin contents.

7 References

[1] K. Dies, Kupfer und Kupferlegierungen in der Technik, Springer-Verlag, Berlin/Heidel-berg/New York, 1967, 514

[2] H. R. Müller, R. Zauter, Erzmetall, 2003, 56, Nr. 11, 643–650[3] R. Zauter, K. Ohla, H.R. Müller, J. Maier, Intern. Conf. On Spray Deposition and Melt

Atomization, SDMA 2003, Bremen – Germany, 2003, 5-113–5-122[4] H. R. Müller, K. Ohla, R. Zauter, M. Ebner, Materials Science and Engineering A, 2004,

383, 78–86[5] H.R. Müller, S. Hansmann, K. Ohla, Intern. Conf. On Spray Deposition and Melt Atomi-

zation, SDMA 2000, Bremen – Germany, 2000, 205–218[6] H. Voßkühler, Z. Metallkunde, 40, 1949, 8, 305–311[7] W. Roth, Z. Metallkunde, 40, 1949, 12, 445–460[8] H. Kästner, Z. Metallkunde, 41, 1950, 8, 193–205[9] H. Kästner, Z. Metallkunde, 41, 1950, 8, 247–254[10] L. Ohm, Metall 43, 1989, 4, 520–524[11] DKI - Deutsches Kupferinstitut, Berlin – Düsseldorf (1965), 14[12] K. Ohla, H. R. Müller, M. Keppeler, A. Bögel, Metall, 55, 2001, 4, 213–215

Figure 12: Different shape and length of machining chips of high tin bronze a) without lead, b) with lead [12]

a) b)

265

Influence of the Crystallization Conditions on the Microstructure and Mechanical Properties of TiAl- and Ti3Al-Based Alloys

B. A. Greenberg, N. V. Kazantseva, A.E. Volkov, Yu. N. AkshentsevInstitute of Metal Physics, Ural Division, Russian Academy of Sciences, Ekaterinburg, Russia

1 Introduction

Titanium aluminides are of practical interest alloys providing unique combination of physicaland mechanical properties. The TiAl-based alloys are characterized by good strength at tempe-ratures up to 650 °C ( 600 MPa), but poor room-temperature plasticity ( 1–2 %). The latterproperty limits their commercial applications. The efforts aimed at the improvement of the pla-sticity of these alloys include both the design of new alloys and new methods of their produc-tion. The optimum combination of properties can be attained by the formation of a specificstructure, such as a fully lamellar two-phase TiAl/Ti3Al structure with controlled content of and 2 phases [1–2]. However, the ultimate strength, plasticity, and fracture behavior of such al-loys are very sensitive to the orientation and microstructure of lamellae. In many works the pro-perties were found to be anisotropic, depending on the interlamellar spacing, domain size,which in turn depended on the crystallization condition of and chemical content of the alloys[3–5].

This work presents a comparative study of influence of the crystallization conditions on themicrostructure and mechanical properties of the TiAl- and Ti3Al- based alloys with the addi-tions of V, Nb, and Mo. All of these alloys were prepared by the special method of pulsed vol-ume pressing (PVP) [6].

2 Experimental Procedure

The alloys in the PVP method were melted in a vacuum of 10 Pa by an electric arc formed bet-ween the initial sample and a consumable electrode of the same composition. The melt wasoverheated by about 10–15 degrees above the melting temperature (this instant was detected byan optical sensor), then the copper (or steel) mold was pressed to the sample, and simultaneous-ly the melt mirror was subjected to gas (Ar) pressure pulses. This imparts a high velocity (5–20m/s) to the melt, which rapidly fills in the mold. In addition, a punch installed in the bottom partof the mold moves toward the melt and presses it, providing additional pressure on the mold. Atthe instant of casting, the melt undergoes a vibration at a frequency of 25–50 Hz.

The Ti-48.%Al-1.%V alloy was prepared in a form of rods of 8 mm in diameter by variousregimes such as casting into copper and steel molds under an additional pressure and castinginto a steel mold under normal pressure. All of others alloys were prepared by casting into cop-per mold under additional pressure.

X-ray diffraction examination was performed using a DRON-3 diffractometer (Cu K radia-tion). The microstructure was examined using a Neophot-2 optical microscope and JEM-200CXelectron microscope. Mechanical tests of samples 3 3 4,5 mm in size were performed in air at

266

room temperature by compression using an INSTRON machine at strain rate of 0,05 mm/min.The pores in the samples were studied using the ultrasound method.

3 Results and Discussion

3.1 Ti-48 at.% Al-1 at.%V

According to X-ray diffraction data the alloy prepared by all the regimes used consists of twoordered phases such as TiAl ( ) and Ti3Al ( 2). The alloy that was cast in a copper mold underpressure has a very fine lamellar structure. The grains in this sample are rather uniform, andtheir average size is 40 m. The TEM study shows a fine lamellar structure with a high disloca-tion density in the lamellae. The 2-phase is present as fine lamellae at twin boundaries in the-phase. The width of and 2 lamellae is 0.06–0.2 and 0.02–0.06 m, respectively (Figure

1a–b). The alloy that was cast into the steel mold without pressure consists of grains both con-taining lamellae and free from them. The grains in this alloy are less uniform in size, which is

Figure 1: Microstructure of the Ti-48%Al-1%V alloy, TEM: (a)–(b)- cast into a copper mold under pressure, (a) dark-field image in the (1-11) -phase reflection, (b)- electron-diffraction pattern for (a), zone axis [10-1] II [11–20] 2; (c) cast into a steel mold without pressure, dark-field image in the (111) reflection; (d) cast into a steel mold under pressure, dark-field image in the (1-10) 2 reflection

267

also 40 m on average. The alternating and 2 lamellae are 0.05–0.4 and 0.05–0.1 m wide,respectively (Figure 1c). Like the alloy cast into the copper mold, the alloy cast into steel moldunder pressure has a lamellar structure (fig1d). The grains are widely differing in size (from 20to 200 m). Both and 2 lamellae are wider (0.2–0.5 and 0.03–0.1 m).

3.2 Ti-46 at.% Al-1,3 at.% V

According to X-ray diffraction data this alloy, as the previous one, consists on two phase: and

2. Parameters of the crystal lattices are: in 2 phase – a = 0.5726 0.0005 nm,c = 0.4615 0.0003 nm; in phase – a = 0.3998 0.0001 nm, c = 0.4024 0.0003 nm.

The structure of this alloy consists on the grain with different form. Some part of the samplecontents the elongated grains. The average size of the grains is about 60 m. The lamellar struc-ture forms in the middle of the grains; and 2 lamellae are 0.3 and 0.08 m wide, respec-tively (Figure 2a).

3.3 Ti-45 at.% Al-1at.% V

X-ray diffractogram of this alloy has also the lines of two phases: and 2. Parameters of thecrystal lattices are: for 2 phase a = 0.57086 0.0005 nm, c = 0.45626 0.0003 nm; for phasea = 0.3982 0.0001 nm, c = 0.4117 0.0003 nm.

Zone of the elongated grains occupies the most part of the sample. The central part of thegrains has thin lamellae: for 0,2 m and for 2 0,03 m in wide (Figure 2b)

3.4 Ti-34 at.% Al-1,6 at.% Nb-0,5 at.% Mo-0,3 at.% Cr

According to the equilibrium diagram this alloy must have 2 single phase content. However,the addition of the beta stabilized elements and high rate of cooling allowed serve high temper-ature 0. The diffractograms of this alloy content the lines of two phases: 0 (2) and 2 (D19). Pa-

Figure 2: Microstructure of the alloys, TEM: (a) Ti-46%Al-1,3%V, dark-field image in the (0-12) 2-phase reflection, (b) Ti-45%Al-1%V, dark-field image in the (002) reflection

268

rameters of the crystal lattices are: for 2 phase a = 0.574 0.0005 nm, c = 0.459 0.0003 nm;for 0 phase a = 0.319 0.0001 nm, c = 0.4117 0.0003 nm.

The structure of this alloy consists on the grains with Widmanstatten structure inside ofthem. The sizes of grain are from 50 up to 200 m. It is found that little grains form in the dif-ferent places of the sample, not only in the middle of the sample as it can be seen in the case ofusual cast ingot. The samples don’t have a texture or a dendrite structure characteristic. On theTEM pictures one can see the short and thick 2 plates and thin 0 plates between them. Manydislocations are seen inside of the 2 plates. We also found the small particles of -phase (B82)inside of the 0 plates (Figure 3a–c). After the additional aging at 900 °C–5 h., the -phase re-flects disappeared and quantity of the dislocations were reduced (Figure 3d).

The techniques of the preparation of titanium aluminides with lamellar structure determinethe grain size, the orientation of lamellae and their thickness. The alloys prepared in the labora-tory by zone melting and having oriented lamellar structure (polysynthetically twinned crystals,PST) have very high mechanical properties at a grain size varying between 25 and 50 m. Thestrength of such alloys reaches 1100 MPa [2]. The maximum strength of PST TiAl alloys is

Figure 3: Microstructure of the Ti-34%Al-1,6%Nb-0,5%Mo-0,3%Cr alloy, TEM: (a) bright-field image; (b) dark-field image in the (011) reflection; (c) electron-diffraction pattern for (b), zone axis [100] ; (d) micro-structure of the alloy after aging 900 °C–5 h, bright-field image

269

1800 MPa at plasticity of 23 %, the interlamellar spacing in the alloy is 1.4 m. In turn, thestrength of polycrystalline TiAl samples prepared by complex thermomechanical treatment andhaving a nonoriented lamellar structure is as small as 600 MPa, and plasticity of the alloy isabout 6 % [7]. Compared to the “pure” PST TiAl alloys, the PST alloys containing vanadiumexhibit a higher plasticity upon both tension and compression, the plasticity of the Ti-48,4at.%Al-0,6at.%V alloy upon compression reaches 28 %. The vanadium-containing alloysare characterized by a more uniform distribution of 2 lamellae and by a thickness of lamellaeof 0.2–2 m [8].

Table 1: Strength characteristics of the alloys prepared by PVP method.

So, according to the experimental results obtained in this work we could figure out some ofthe external factors that responsible to the successful structure of alloys with high mechanicalproperties:

1. Vibration and addition pressing under crystallization are beneficial for small grain structure, which form by the dendrite breakage.

2. The rough and very cold mold, lowed temperature of casting, and convection flows are also beneficial for small uniform equiaxed grains formation.

3. The cooling rate varying in the mold with different heat capacities (0.385 cal/g K for copper and 0.12 cal/g K for steel) also substantially affects the grain size, the uniformity of its distri-bution and the lamella thickness. In the Ti3Al-base alloy high rate of cooling allows serve “soft” plates of 0 phase.

4 Acknowledgments

This study was supported financially by the Program "National technological basis", grant47/04/768-2004, and the Russian Fund for Basic Research Ural, grant 04-03-96008.

N Alloy % b, MPa 0,2, MPa

1 Ti-48 at.% Al-1 at.% V a) casting into copper mold under pressureb) casting into steel mold under pressurec) casting into steel mold without pressure

32,626,528,6

10881041972

685734568

2 Ti-46 % Al-1,3 % V. 18 1030 8443 Ti-45 at.% Al-1at.% V 17 1184 10074 Ti-34 at.% Al-1,6 at.% Nb-0,5 at.% Mo-0,3 at.% Cr

a) casting into copper mold under pressureb) + aging 900°C–5 h

1019

19201700

1314872

270

5 References

[1] Kim Young-Won, JOM. 1989, 41, 24–30[2] S. Naka, M. Tomas, and T. Khan, Mat. Sci. and Technol. 1992, 8, 291–298[3] M. Yamaguchi and Y. Umakoshi, Progress in Materials Science. 1990, 34, 60–73[4] N.V. Kazantseva, A.E. Volkov, B.A. Greenberg, A.A. Popov, V.V. Yurovskikh, Phys.

Met. Metallogr. 2001, 91, 173–178[5] B.A. Greenberg, N.V. Kazantseva, A.E. Volkov, etallography and Temperature treating

of the Materials. 2005 (in print)[6] A.E. Volkov, A.V.Frolov, V.N.Boiko, RU Patent 2 106 226, 1996[7] T.Maeda, M.Hosomi, M.Okada. in The Proc. Symp. Sponsored by the Structural Materi-

als Division (SMD) of TMS (Ed.: Young-Won Kim et.al), Warrendale, Pa, 1995, 771–778

[8] K.F.Yao, H.Iniu, K.Kishida, and M.Yamaguchi, Acta Metall.Mater. 1995,43, 1075–1086

271

Effects of Process Parameters on the Characteristics of the Billet Sump and Related Defect Formation during DC Casting of Aluminum Alloys

D. G. Eskin1, L. Katgerman2

1Netherlands Institute for Metals Research, Delft, The Netherlands2Delft University of Technology, Dept. Materials Science and Engineering, Delft, The Netherlands

1 Introduction

There are three main process parameters that one can change upon direct-chill casting, i.e. ca-sting speed, water flow rate, and melt temperature. Water flow rate should typically assure ef-ficient cooling (nucleate boiling conditions), and its increase above this sufficient level does notaffect much the structure and defect formation [1] The other two process parameters are extre-mely important, e.g. in occurrence of casting defects [1, 2, 3, 4, 5]. Casting speed and castingtemperature are also known to affect the structure formation during solidification. This is becau-se of their influence on cooling conditions, melt flow and geometry of liquid and semi-liquidparts of the billet [1, 3, 6]. The increase in casting speed results in proportional deepening of theliquid sump, increasing of the mushy zone thickness, and in overall acceleration of solidificati-on [3]. The melt temperature, however, has not received much attention from the viewpoint ofits influence on the structure and defect formation in DC cast billets. In a very good review ofDC casting [7], the authors mention the melt temperature only once and then among other para-meters that can also affect the structure and quality of a billet.

As for the casting defects, two are most frequently encountered in practice, i.e. macrosegre-gation and hot tearing.

Macrosegregation patterns generally depend on the distribution coefficient k and are obvi-ously linked to the morphology of the forming solid phase, permeability of mushy zone, magni-tude of solidification shrinkage, ratio between the shrinkage velocity and the direction andvelocity of melt flow, level of solute rejection to the melt, and movement of the solid phase inthe liquid and slurry regions [1, 8]. Macrosegregation is known to increase with the castingspeed [1, 3]. There are controversial data on the effect of melt temperature on macrosegregation[9, 10]. Shestakov [11] showed analytically that the increase in melt temperature narrowed thetransition region in the billet and accelerated the solidification (local solidification time is short-ened).

The most important casting parameter that affects hot tearing is also the casting speed, andthe main reason for that is believed to be connected to greater temperature gradients and, there-fore larger thermal strains [3, 5, 12]. The optimum casting speed is a compromise between pro-ductivity, alloy composition, billet size, and quality (structure and defects). The effects of melttemperature on hot tearing was previously studied only for shape castings [13]. It was shownthat higher melt temperature results in higher hot tearing susceptibility. However, direct appli-cability of these results to direct-chill casting is unclear.

The analysis of available literature sources reveals, therefore, gaps in published experimentaland modeling information. Only few experimental data are reported on the interrelation be-

272

tween the sump and mush dimensions, and melt flow patterns on one side and the macrosegre-gation and hot tearing on the other side.

The aim of this paper is to report the results of systematic examinations of billets of binaryAl–Cu alloys cast under different process conditions. Effects of casting speed and melt temper-atures on macrosegregation and hot tearing are examined and correlated to computer simulatedsolidification patterns.

2 Experimental and Numerical Procedures

A series of experiments was performed in a pilot direct-chill casting installation in Delft Uni-versity of Technology. Round billets 195 mm in diameter were cast in a 200-mm hot-top mould.Binary Al–Cu alloys containing 1 to 5% Cu were prepared using 99.85% pure aluminum andAl–47.7% Cu master alloy. Two types of experiments were performed, i.e. upon stationary andtransient stages of casting. In the first case, the casting parameters were changed stepwise withproducing at least 200 mm of a billet at each casting regime. In the second case, the castingspeed was ramped up and then down at a constant rate, and the structure was examined at diffe-rent casting speeds during the ramping. The detailed description of experimental equipment andexperiments can be found elsewhere [1, 4, 5]

The structure of the billets was examined in an optical microscope Neophot 30. The chemi-cal composition was measured using a spark spectrometer SpectroMax. The hot-cracking sus-ceptibility was estimated as the ratio of the billet cross-section affected by cracks to the totalbillet cross-section.

The flow patterns and sump profiles were assessed numerically using the Flow3D and CFXsoftware [1, 4] and, the case of sump profile, also with the MSC.Marc software [5]. Measure-ments of temperatures in the hot top and sump of the billet and measurements of the sump depthwith a rod during DC casting experiments were used for validation of the numerical results.

3 Results and Discussion

3.1 Macrosegregation and Floating Grains

The magnitude of macrosegregation is clearly affected by the casting speed as shown in Fig. 1a.Casting speed also greatly influences the geometry of the billet sump as demonstrated in Fig.1b, c. One can notice a distinct correspondence between the extent of macrosegregation in theradial direction of the billet and the width of the transition region in the billet. This correlation isillustrated in Fig. 1d. One can see that the relative copper concentration normalized to the verti-cal distance between solidus and liquidus isotherms is not dependent on the radial position inthe billet cross-section, except for the subsurface region where additional mechanisms ofmacrosegregation, e.g. shrinkage-driven flows, may act complementary to the thermo-solutalconvection [1].

273

Figure 1: Effect of casting speed under steady-state casting conditions on macrosegregation (a) and parameters of the sump (b, c) in the Al–4.3% Cu billet; and (d) macrosegregation of copper (as in a) normalized to the width of the transition region (as in c). Numbers in b relate to the centre of the billet: 1, calculated total depth of the sump; 1’, measured total depth of the sump; 2, calculated depth of the liquid pool; and 3, calculated distance bet-ween liquidus and solidus isotherms.

Figure 2: (a) Effect of melt temperature in the furnace on the macrosegregation and (b) the flow pattern at a melt temperature at the inlet to the hot top of 725 °C (corresponds to the melt temperature in the furnace 760 °C) in the Al–2.8% Cu billet cast at 200 mm/min (transient casting stage). Water flow rate 150 l/min.

274

The melt temperature does not affect much the segregation pattern in the bulk of a billet,which is in line with the opinion of Tarapore [9] and Reese [10]. The subsurface segregation,however becomes more pronounced on increasing the melt temperature as illustrated in Fig. 2a.Computer simulations show that the transition zone in the center of the billet becomes narrowerwith increasing the melt temperature whereas the transition zone at the periphery of the billetexpands [4]. Therefore the correlation between the dimensions of the transition region and theextent of macrosegregation shown in Fig. 1d holds here as well. In addition, the computer simu-lation of flow patterns at high melt temperatures show strong currents (of solute-enriched liq-uid) directed towards the billet surface, Fig. 2b.

The appearance of floating grains is frequently linked to macrosegregation patterns, assum-ing that solute-depleted grains brought to the center of the billet by melt flow contribute to thenegative segregation [2]. Floating grains are characterized by a coarser internal structure indica-tive of their longer solidification time. Recently, based on the numerically assessed flow pat-terns in the slurry part of the billet, we suggested a mechanism of floating grains formation [1].This mechanism implies that floating grains are formed in the upper part of the slurry region

Figure 3: Effects of casting speed and melt temperature of the distribution of floating grains in the horizontal cross-section (a, b) and on the calculated flow patterns in the Al–2.8% Cu billet (c–f): a, c, e, casting speed 100 mm/min, steady-state stage; b, 200 mm/min, transient stage; and d, f, 200 mm/min, steady-state stage. Melt tempe-ratures are (c, d) 700 °C and (e, f) 760 °C. Water flow rate is 150 l/min. Temperatures are the melt temperatures in the furnace.

275

and closer to the periphery of the billet, and then are transported by the flow to the central partof the billet. The effective cooling solidification time for these grains is thus much longer than itcan be presumed just from the vertical dimension of the transition zone. The effects of castingspeed and melt temperature on the distribution and amount of floating grains are illustrated inFig. 3. In the range of low casting speeds (Fig. 3a), „floating” grains spread across the entirecross-section of the billet at a low melt temperature (700 °C) and are confined to the central partof the billet at higher melt temperatures. Their amount decreases with increasing melt tempera-ture, which agrees well with earlier predictions [9]. In this casting regime, the increase in melttemperature results in diminishing of the region with stagnant flow in the central part of the bil-let where the „floating” grains may grow (Fig 3c, e), with a larger fraction of „floating” grainsgoing upwards in the slurry zone and, apparently, remelting. In the range of high casting speeds,the distribution of „floating” grains is similar at any given melt temperature, with the maximumfraction in the central part of the billet, Fig. 3b. However, the total amount of these coarse grainsincreases with melt temperature. We can suggest that deepening of the sump and more severecurrents in the vicinity of the mushy zone with increasing the melt temperature and the castingspeed as shown in Fig. 3d, f create more possibilities for „floating” grains to form, grow andsettle.

3.2 Hot Tearing

Our experimental observations show that susceptibility to hot tearing during direct-chill castingdepends on the composition, increases with increasing the casting speed, and decreases withincreasing the melt temperature [4, 5].

Hot tearing susceptibility of an alloy depends on the dimensions and properties of the so-called vulnerable solidification range that is confined between the isotherms of rigidity and thesolidus [14], or in the case of the billet – on the dimensions of the mushy zone.

The increasing casting speed results in widening of the transition region, especially in thecentre of the billet, where hot cracks are typically observed (see Fig. 2b, c), and in higher tensilestrains concentrated in the central part of the billet [5]. Computer simulations show that on in-creasing the melt temperature, however, the mushy zone either retains its dimensions in the cen-tre of the billet or becomes thinner, meaning that under the same casting conditions the alloyspends the same or less time in the vulnerable range [4]. At the same time, the liquid bath (dis-tance from the melt surface to the liquidus) and the sump as a whole become deeper with in-creasing melt superheat, resulting in a larger metallostatic pressure on the mushy zone and,therefore better feeding of the melt to potential cracks, Fig. 3e, f. This is revealed in the struc-ture as „healed” cracks – relatively long paths spread along several grain boundaries and filledwith eutectic, the amount of the latter also increasing with the melt temperature [4].

4 Conclusions

Experimental and numerical studies of DC casting show the direct correlation between the de-gree of macrosegregation, hot tearing susceptibility, and amount of floating grains, on one hand,and the vertical distance between the liquidus and solidus isotherms and melt flow patterns inthe sump of the billet, on the other hand. Macrosegregation and hot tearing susceptibility increa-

276

se with the casting speed. The increased casting temperature, however only affects the subsur-face segregation and causes less hot tears.

5 Acknowledgements

This work is done within the framework of the research program of the Netherlands Institute forMetals Research (www.nimr.nl), Projects MP 97014 and MC 02134. Contribution of J. Zuide-ma, Jr., V.I. Savran, and Q. Du to computer simulations is highly appreciated.

6 References

[1] D.G. Eskin, J. Zuidema, Jr., V.I. Savran, L. Katgerman, Mater. Sci. Eng. A 2004, 384, 232–244

[2] E.F. Emley, Intern. Met. Rev. 1976, Review 206, 75–115[3] V.A. Livanov, R.M. Gabidullin, V.S. Shepilov, Nepreryvnoe lit’e alyuminievykh splavov

(DC Casting of Aluminium Alloys), Metallurgiya, Moscow, 1977, p. 168[4] D.G. Eskin, V.I. Savran, L. Katgerman, Metall. Mater. Trans. A 2005, 36A (in press)[5] Suyitno, D.G. Eskin, V.I. Savran, L. Katgerman, Metall. Mater. Trans. A 2004, 35A,

3551–3561[6] H. Nagaumi, Sci. Technol. Adv. Mater. 2001, 2, 49–57[7] J.F. Grandfield, P.T. McGlade, Mater. Forum 1996, 20, 29–51[8] M.C. Flemings, Solidification Processing, McGraw-Hill, New York, 1974, p. 364[9] E.D. Tarapore in Light Metals 1989 (Ed.: P.G. Campbell), The Minerals, Metals and

Materials Society, Warrendale, USA, 1989, p. 875[10] J.M. Reese, Metall. Mater. Trans. B 1997, 28B, 491–499[11] A.D. Shestakov, Izv. Ross. Akad. Nauk, Metally 1996, 6, pp. 130–138[12] M. M’Hamdi, A. Mo, C.L. Martin, Metall. Mater. Trans. A 2002, 33A, 2081–2093[13] I.I. Novikov, Goryachelomkost tsvetnykh metallov i splavov (Hot Shortness of Non-Fer-

rous Metals and Alloys), Nauka, Moscow, 1966, p. 299[14] D.G. Eskin, Suyitno, L. Katgerman, Progr. Mater. Sci. 2004, 49, 629–711

277

Effect of Casting Speed and Grain Refining on Macrosegregation of a DC Cast 6061 Aluminum Alloy

R. Kumar Nadella1, D. Eskin1 and L. Katgerman2

1 Netherlands Institute for Metals Research, Delft, The Netherlands2 Delft University of Technology, Department of Materials Science and Technology, Delft, The Netherlands

1 Introduction

Direct Chill (DC) casting remains a major processing route to produce large aluminum ingots,which are used for downstream processing. During this process, grain refining is normally em-ployed to achieve finer equiaxed grains. However, the occurrence of defects such as hotcracking, macrosegregation etc. needs to be properly understood from the quality point of view.Process parameters such as casting speed, water flow rate and melt temperature can be tailoredto get a sound product. Macrosegregation, which is the non-homogeneous distribution of al-loying elements over a large length scale, needs to be minimized as it affects the properties ofthe finished product. This is because during subsequent homogenization, while microsegregati-on is reduced, macrosegregation remains essentially unaffected. The nature of solidificationprofile during DC casting together with convective flows in the ingot sump is responsible forthis phenomenon [1]. Further, the degree of macrosegregation can be such that the compositionin certain regions across the thickness of the ingot may be outside the registered limits esta-blished for the alloy.

In addition to the casting process parameters, grain refining has an important effect on mac-rosegregation. Despite the large amount of literature and excellent reviews on DC casting [1,2],there has been little systematic work concerning the effect of grain refining on macrosegrega-tion. This is particularly evident in case of commercial Al alloys, which contain more than onealloying element.

2 Experimental Methods

Commercial-scale DC casting experiments were conducted in a pilot DC casting installation atthe Delft University of Technology, the details of which can be found elsewhere [3]. Experi-ments were carried using an AA 6061 with the composition (in wt.%): 0.51 Si, 0.23 Cu, 0.95Mg, 0.12 Fe, 0.007 Mn, 0.076 Cr, 0.016 Zn with and without grain refining. Different castingspeeds were used in these experiments while maintaining a constant water flow rate (170 l/min)and casting temperature (715 °C). The sump depth is measured by a digital length meter. DCcast billets of 192 mm in diameter obtained from a 200 mm round hot top mould are longitudi-nally sectioned in the center and samples of approximately 20 mm wide and 20 mm high werecut at suitable locations in the horizontal cross section of the billet. Care was taken to ensurethat the sampling represents the steady state conditions during DC casting. These rectangularbars were analyzed by a spark spectrum analyzer across the billet diameter for the compositionvariation. Measurements were taken in all 4 sides at regular intervals of approx. 10 mm, and the

278

average values are reported. The absolute error in these measurements is ± 0.05 wt. %. Concur-rent microstructural observations were carried out close to the surface and in the central portionof the billet. Samples anodized in a 3% HBF4 water solution were observed under cross-polari-zed light to check the grain size.

3 Results and Discussion

The process chart for a typical DC cast experiment along with an illustration of an AA 6061 bil-let is shown in Figure 1. To obtain a grain-refined ingot, a known quantity (2 kg/tonne) of grain-refiner in the form of Al-5Ti-1B was added to the liquid metal in the furnace just prior to com-mencing the DC casting. The average Ti concentrations were about 0.01 wt.%.

To study the macrosegregation patterns in this alloy, the composition profiles of the majoralloying elements (Mg and Si) are plotted across the whole diameter (192 mm). The generaltrend for all the observed patterns is a negative segregation in the center and minor positive seg-regation at the mid-radius. Further, strong segregation zones are noticed close to the surface.

The effect of grain refinement on the macrosegregation profiles is shown in Figure 2 for acasting speed of 8 cm/min. It can be seen that grain refining does not seem to have a considera-ble influence on the composition profiles. However, at higher casting speeds, it is observed(Figure 3) that both grain refined and non-grain refined samples exhibited higher segregationlevels, particularly in the central portion of the billet. For the grain-refined ingot, significant Tienrichment is observed in the central portion, which is again found to be higher at increasingcasting speed. Figure 4 illustrates this at a casting speed of 12 cm/min. From the above resultsso far, it appears that the casting speed has major influence on macrosegregation compared tothe grain refining

Microstructural observations (Figure 5) near the surface and at the center indicate significantgrain refining with Al-5Ti-1B addition. The grain size is refined from 400 m to around 75 m.Irrespective of the grain refinement, grains with larger dendrite arm spacing are seen in the cent-

Figure 1: Typical process chart for the DC casting experiment of AA 6061

279

er. Further, for the grain-refined ingot, the grain size is coarser in the center compared to thesurface.

In general, the movement of inter-dendritic liquid (solute-rich) in the mushy zone and thetransport of solid grains (solute-lean) account for the segregation patterns observed in DC cast-ing [1]. In addition to the natural convection of solute-rich liquid, the transport of solute-leansolid phase from the periphery of the billet to the center leads to the negative centerline segrega-tion. Isothermal dendrites formed early in the solidification process are detached and carried bythe strong natural convection currents into the molten metal pool [4]. They grow isothermally ata temperature close to the alloy liquidus. It is generally accepted these isothermal dendrites withcoarse cells observed in the center of the billet increase the severity of negative segregation [5-7]. The central part of the billet thus shows duplex structure with a mixture of grains exhibitingfiner and coarser internal structure. Our observation of grains with larger dendrite arm spacingin the center of the billet present case (Figure 5b,d) is in qualitative agreement with the abovestatement.

The effect of casting speed on macrosegregation of various Al alloys is well known [3, 4].The main parameters that influence structure formation and macrosegregation during DC cast-ing are the sizes of the transition region along with the flow pattern in the slurry region of thebillet [3]. Increase in casting speed leads to the widening of the transition region, especially inthe central portion of the billet. In the present work, the sump depth measurements indeedshowed a great variation ( from 25 mm to 68 mm) as the casting speed increased from8 cm/min to 12 cm/min. Quite the same depths were recorded with grain-refining ingots. This

Figure 2: Macrosegregation profiles for AA 6061 at a casting speed of 8 cm/min (a) Non-grain refined and (b) grain refined [ Mg and Si]

280

means that the increase in the severity of segregation with higher casting speed can be directlyrelated to the depth of the sump in both cases.

As expected, the surface region of the DC cast material is characterized by a strong segrega-tion zones (in the present case either positive or negative depending on how close the measure-ments are made in relation to the surface). This is related to the shrinkage driven flow of solute-rich liquid [4]

With respect to the effect of grain refining on macrosegregation, however, there are conflict-ing reports in the literature. Finn et al [8] who showed that grain refining produced positive cen-terline segregation due to the improved permeability of the mushy zone. Opposite trends areobserved by Lesoult et al [9] who showed that grain refining causes more severe centerline seg-

Figure 3: Compositional variation of Ti across the grain refined billet cast at 12 cm/min

Figure 4: Compositional variation of Ti across the grain refined billet cast at 12 cm/min

281

regation. Greater centerline depletion of Mg is observed in AA5182 with grain refining andconnected to the formation and transport of isothermal dendrites (or “floating grains”) in thegrain-refined ingot [10,11]. Gariepy and Caron [5] studied the effect of various methods andgrain refiner on macrosegregation. They found that liquid flow, which worked against naturalconvection, and smaller amounts of grain refiner, reduced macrosegregation. In the presentcase, the balancing effect of the increased permeability of the mushy zone along the increasedamounts of isothermal dendrites in the central zone may lead to a nearly unchanged macroseg-regation patterns in the grain refined and non-grain refined samples. Further work, however,needs to be directed towards the quantification of these floating grains. Also it would be inter-esting to examine the effect of Ti concentrations (i.e. amount of grain refiner) on structure andmacrosegregation.

4 Conclusions

Direct chill casting experiments with and without grain refining at different casting speeds wereconducted on an AA 6061. Significant structural refinement is observed. The concentration pro-files for major alloying elements (Mg and Si) showed a negative segregation in the center andclose to the surface. The severity of segregation increases with higher casting speed both in non-

Figure 5: Optical microstructures of non-grain refined and grain refined billets in the center (b, d) and close to the surface (a, c). [Arrows in (b, d) show “floating grains”]

282

grain refined and in grain-refining billet. On the other hand, grain refining does not seem tohave any considerable effect with respect to macrosegregation.

5 Acknowledgements

This work is done within the framework of the research program of the Netherlands Institute forMetals Research, Project MC4.02134.

6 References

[1] J. F. Grandfield and P.T.McGlade, Materials Forum, 1996, 20, 29–51[2] E. F. Emley, Int. Met. Rev., 1976, 206, 75–115[3] D. G. Eskin, J. Zuidema, Jr., V. I. Savran and L. Katgerman, Mat. Sci. Engg, 2004, A384,

232–244[4] H. Yu and D .A. Granger, Fundamentals of alloy solidification applied to industrial pro-

cesses, Proc. NASA symposium, 1984, p. 157[5] B. Gariepy and Y. Caron, Light Metals 1991, The Minerals, Metals and Materials

Society, Warrendale, PA, p. 961[6] R. C. Dorward and D. J. Beerntsen in Light Metals (Ed.: C. M. Bickert), The Minerals,

Metals and Materials Society, Warrendale, 1990, p.919[7] Men G. Chu and John E. Jacoby in Light Metals (Ed.: C. M. Bickert), The Minerals,

Metals and Materials Society, Warrendale, 1990, p.925[8] T. L. Finn, M .G. Chu, W. D. Bennon, Micro/Macro Scale Phenomena in solidification,

ASME, New York, 1992, p. 17[9] G. Lesoult, V. Albert, B. Appolaire, H. Combeau, D. Dalouz, A. Joly, C. Stomp, G. U.

Grün and P. Jarry, Sci. Tech. Adv. Mat, 2001, 2, 285–291[10] A. Joly, G. U. Grün, D. Daloz, H. Combeau and G. Lesoult, Mat. Sci. Forum, 2000,

329-330, 111–120[11] A. M. Glenn, S. P. Russo and P. J. K. Paterson, Met. Mater. Trans A., 2003, 34A

1513–1523

283

Effect of Melt Flow on Macrostructure and Macrosegregation of an Al–4.5% Cu Alloy

A. N. Turchin1, D. G. Eskin1, L. Katgerman2

1 Netherlands Institute for Metals Research, Delft, the Netherlands2 Delft University of Technology, Department of Material Science and Engineering, Delft, the Netherlands

1 Abstract

One of the major problems in production of large ingots and billets is macrosegregation. Theoccurrence of this defect is associated with melt flow in the liquid sump of billets or ingots. Inthis paper the effect of melt flow on macrostructure and macrosegregation of an aluminum alloyis investigated using the electromagnetic pump, which allows one to organize the controlledmelt flow along the solidification front. The experiments are performed on an Al–4.5% Cu alloyin a wide range of melt flow velocities and temperatures. Computer simulations of the experi-ment with a CFD software are used in the interpretation of the formation of macrosegregation inthe presence of melt flow. The results show strong influence of melt flow on the macrostructureand macrosegregation. The concentration profile in the sample solidified without flow showsnegative macrosegregation in the center part. With increasing melt flow the positive macrose-gregation can be observed. The interpretation of the structure formation in the presence of meltflow is discussed.

2 Introduction

Melt flow involved in all casting processes has significant consequences for structure andmacrosegregation evolution during solidification. The published studies show that the melt flowinfluences the size and distribution of grains, affects grain morphology [1–6], e. g., causes thecolumnar to equiaxed transition (CET) [4, 5] and the appearance of feathery grains in aluminumalloys [6]. Macrosegregation is a major problem during the production of large ingots and bil-lets where it appears due to compositional differences in liquid and solid. It is believed to becontrolled by melt flow in the liquid sump of a casting. Many experimental and numerical stu-dies have been done in order to understand and to predict the macrosegregation development[7–10]. The detailed description of causes of solute movement due to fluid flows in casting pro-cesses can be found elsewhere [11]. It was shown that the segregation during casting is a resultof several types of flow in the transition zone: shrinkage flow during the cooling [12], buoyancyinduced flows due to thermal and solutal convection [13], and forced convection leading to mo-vement of grains [14]. However, it was found in some works that flow can not only promote thesegregation but also almost eliminate it [15].

This study is aimed to investigate the effects of controlled linear melt flow on macrostructureand macrosegregation formation in an Al–4.5% Cu alloy using specially dedicated electromag-netic pump.

284

3 Experimental Technique and Computer Simulations

An experimental set-up consists of an electromagnetic pump, control, melt-guiding and data ac-quisition systems [16]. A „flow-through” melt-guiding system was designed in this work in or-der to obtain a constant unidirectional bulk flow along solidification front with the control of theflow rate. Working on magneto-hydrodynamic principle, the pump creates a linear melt flowthat goes through a rectangular launder with dimensions of 800 60 70 mm into the crucibleplaced on digital scales. Solidification occurs under conditions of constant melt flow along thesolidification front on the bottom side of a water-cooled bronze chill built in the bottom surfaceof the launder. The chill has a rectangular shape 34 mm 110 mm in plane section with the in-ner cavity about 10 mm 100 mm. The water-cooled chill is designed to reproduce the linearmelt flow motion along the solidification front. The objectives of experiments were to study theeffect of flow velocity and melt temperature on macrostructure and macrosegregation of an Al–4.5% Cu alloy1. An experimental Al–Cu alloy with a chemical composition of 4.40% Cu,0.20% Si, and 0.12% Fe was prepared using 99.95% pure aluminium and an Al–47.7% Cu ma-ster alloy.

Temperatures in the liquid bath of the pump and at the entrance to the launder were control-led and measured during experiments. Melt velocity is controlled by a stepwise change of volt-age at the electromagnet. The weight per unit time is initially measured by the digital scalesMettler Toledo. The linear melt flow velocity was then recalculated from the values of weightper time. The samples were obtained at melt temperatures in the liquid bath between695–745 °C and in a velocity range of ~0.05–0.60 m/s.

After experiments, samples were sliced in the middle section in the longitudinal direction forthe examination of macrostructure. After cutting in the longitudinal direction, samples werepolished, etched with 45 ml HCl, 15 ml HNO3, 15 ml HF and 25 ml H2O solution (Tucker's rea-gent) for approximately 25 seconds in order to investigate the macrostructure of the whole sec-tion. The second half of each sample was used for measurements of chemical composition alonglongitudinal direction (flow direction) close to the chill surface. Measurements were taken each20 mm using a spark spectrum analyzer SpectroMax. The absolute error for measured copperconcentration was ±0.01 wt %.

Computer simulations of the effects of process parameters, such as melt flow velocity andtemperature, under conditions of constant melt flow including solidification were performed us-ing the Flow-3D commercial software (Version 9.0). A two-dimensional flow of a molten alu-minum alloy Al–4.5% Cu was assumed in a launder. A computational block consisted of twoobstacles: one plate and a built-in chill with the cavity reproducing the geometry used in exper-iment. Thermophysical parameters for an Al–4.5% Cu alloy used in the simulations are de-scribed elsewhere [ ]. The computational time was 30 s. The grid consisted of structureduniform 2D mesh: 35840 cells in 640 56 square. The melt flows from the left side of the com-putational block with the initial temperature 700 °C at various velocities specified as the left-hand boundary conditions. The interface between obstacles and flowing melt is characterized bya heat transfer coefficient of 150 W/m2K with the temperature of obstacles 373 K. The initialcondition of void state is specified as 298.15 K. When liquid reaches the cavity the solidifica-

1 All alloy compositions are in wt%

285

tion occurs. The heat transfer coefficient between the bottom of block and flowing melt is1150 W/m2K [19]. The right-hand boundary condition is specified as outflow.

4 Results

Macrostructure examination of samples obtained under different conditions shows that meltflow dramatically changes the morphology and distribution of grains. While the macrostructureobtained under „no-flow” condition generally consists of equiaxed grains and small fraction ofcolumnar grains at the bottom of the sample (Fig 1 a), the columnar and feathery grains orientedtowards incoming flow appear in the whole volume of samples by applying relatively slow flowvelocities (up to 0.10 m/s) (Fig. 1 b, c). The zone of equiaxed grains develops in the lower partof a sample obtained at 0.10 m/s. This zone expands and the grains become finer with furtherincreasing of flow velocity (Fig. 1 d, e). The increasing of melt temperature up to 745 °C pro-motes the formation of columnar and feathery grains within the given velocity range.

The effects of flow velocity and melt temperature on macrosegregation pattern are demon-strated in Fig. 2. Figure 2 a shows the strong influence of flow velocity on the extent of mac-rosegregation. The profile of sample obtained under „no-flow” conditions exhibits so-callednegative macrosegregation with minimum in the center of the sample. The negative profile re-mains after applying slow melt flow (~ 0.03 m/s). On further increasing of flow velocity thepositive profile of macrosegregation with the shift of maximum copper concentration towardsthe incoming flow can be observed for velocities of 0.05 m/s and 0.10 m/s. Finally, at a velocityof 0.30 m/s the pronounced positive segregation profile in the center of the sample is obtained.

Figure 1: Macrostructure of the Al–4.5% Cu alloy obtained at 700 oC and at various melt velocities of (a) without flow, (b) 0.03 m/s, (c) 0.05 m/s, (d) 0.10 m/s, and (e) 0.30 m/s; bottom side–chill surface, melt flow direction–from left to right

286

The effect of melt temperature at a constant velocity of 0.03 m/s on macrosegregation pat-terns is shown in Fig. 2 b. Increase in melt temperature has a little effect on the segregation pro-file with a tendency to produce larger segregation.

5 Discussion

The experimental results show that the parameters of unidirectional flow such as velocity andtemperature have significant effects on macrostructure, namely: (1) solidification in flowingmelt results in the formation of columnar and feathery grains in comparison with „no-flow”conditions where macrostructure exhibits equiaxed grains; (2) columnar grains are oriented to-wards incoming flow; and (3) increasing of melt velocity results in CET and grain refinement.Numerous works have been performed on effects of bulk [4, 5, 10, 15] and unidirectional flows[17, 18] on macrostructure development during solidification. Data produced by computer si-mulations (Fig. 3 a) show that the macrostructure composed of columnar grains is formed underconditions of high thermal gradient and low undercooling that agrees well with earlier observa-tions [17, 18]; and development of equiaxed grains in the lower part of samples (Fig. 1 d, e) isassociated with strong melt recirculation and, consequently, specific thermal and concentrationfields in this region [4, 5, 10] (Fig. 3 b).

To the best of our knowledge, there are no papers where the macrosegregation pattern is an-alyzed in samples obtained under conditions of constant unidirectional melt flow. The forma-tion of negative segregation profile in the sample obtained under „no-flow” conditions can beexplained by the motion of liquid melt (Fig. 4 a). The transport of solute-enriched liquid fromthe center of the sample to the periphery and stagnant region in the center of the sample with a

Figure 2: Macrosegregation profiles for copper concentration (Xi – Xnominal)/ Xnominal in the longitudinal section of the samples obtained at 700 °C and at different flow velocities (a) and at velocity of 0.03 m/s (b) and different melt temperatures; melt flow direction–from left to right

287

potential for settling down solute-poor grains contribute to the development of negative segre-gation in the center of the sample. This trend remains throughout solidification with increasingof solid fraction. In spite of the fact that the macrostructure entirely changes to columnar at arelatively slow velocity of 0.03 m/s, the profile does not alter and exhibits again the negativesegregation. It can be a result of almost identical velocity pattern as during „no-flow” conditionswhich tend to enrich the periphery regions.

It is known that applying of melt stirring in the liquid sump of the billet during DC castingprocess leads to the formation of normal, namely, positive segregation [20]. The same tendencyof profile changing is also observed in the present work by applying the strong fluid flow (from0.20 m/s). This fact may be a result of appearance of strong flows in the opposite direction to

Figure 3: Temperature field in the chill region at an initial melt temperature of 700 °C and a melt flow of 0.05 m / s after 30 s (a) and velocity pattern at 700 °C and 0.30 m/s after 20 s (b); bulk melt flow direction–from left to right

Figure 4: Velocity pattern at 700 °C under „no-flow” conditions after 3 s, arrows in the central part are artificiallyenlarged (a) and in the central part of the chill region at the flow velocity of 0.30 m/s after 15 s (b); bulk melt flowdirection–from left to right

288

the initial flow and, thus, may increase melt stirring in the vertical section of the sample, as canbe seen in Fig. 4 b. In that way the flow influences the transfer of the solute to the center fromperiphery of the sample. The formation of zone of equiaxed grains apparently formed by frag-mentation in the right bottom corner of samples further contributes to the solute depletion ofthis region (Fig. 2 a).

A transient situation is observed in a velocity range between 0.03 m/s and 0.10 m/s. Increas-ing of melt temperatures enlarges the positive segregation at a flow velocity of 0.03 m/s (Fig. 2b). The growth of coarse columnar grains and, consequently, the formation of large interdendrit-ic channels with increasing of melt temperature can promote stronger solute transfer within theinterdendritic region.

6 Conclusions

Effects of parameters of unidirectional flow such as melt temperature and flow velocity were in-vestigated. It is found that the solidification in flowing melt results in the formation of columnargrains deflected towards incoming flow and promotes CET and grain refinement in recirculati-on zone. Moreover, the melt flow changes macrosegregation pattern from negative to pro-nounced positive, when strong flows are applied.

7 Acknowledgements

This work is done within the framework of the research program of the Netherlands Institute forMetals Research (www.nimr.nl), Project MC 4.02134.

8 References

[1] D. Daloz, H. Combeau, S. Sterdjella, B. Commet, Ph. Jarry, Aluminium 2004, 6, 603–608

[2] R. S. Rerko, H. C. de Groh, III, C. Beckermann, Mater. Sci. Eng. A 2003, 347, 186ñ197[3] K. Kubota, K. Murakami, T. Okamoto, Mater. Sci. Eng. A 1986, 79, 67–77[4] S. R. Chang, J. M. Kim, C. P. Hong, ISIJ International 2001, 41, 738–747[5] W. D. Griffits, D. G. McCartney, Mater. Sci. Eng. A 1996, 216, 47–60[6] S. Henry, G.-U. Gruen, M. Rappaz, Metall. Mater. Trans. A 2004, 35A, 2495–2501[7] D. G. Eskin, J. Zuidema, V. I. Savran, L. Katgerman, Mater. Sci. Eng. A 2004, 384,

232–244[8] J. P. Gu, C. Beckermann, Metall. Mater. Trans. A 1999, 30A, 1357–1366[9] Y. Yang, Q. Zhang, Y. He, Z. Hu, Sci. Techol. Adv. Mater. 2001, 2, 271–275[10] W. D. Griffits, D. G. McCartney, Mater. Sci. Eng. A 1997, 222, 140–148[11] C. Beckermann, Int. Mater. Rev. 2002, 47, 243–261[12] M. J. M. Krane, F. P. Incopera, Metall. Mater. Trans. A 1995, 26A, 2329–2339[13] A. V. Reddy, C. Beckermann, Metall. Mater. Trans. B 1997, 28B, 479–489[14] C. J. Vreeman, F. P. Incropera, Int. J. Heat Mass Transfer 2003, 43, 687–704

289

[15] B. Zhang, J. Cui, G. Lu, Mater. Sci. Eng. A 2003, 355, 325–330[16] A.N. Turchin, D.G. Eskin, L. Katgerman in Light Metals 2005 (Ed.: H. Kvande), The

Metals, Minerals and Materials Society, Warrendale, USA, 2005, p.1025[17] K. Murakami, T. Fujiyama, A. Koike, T. Okamoto, Acta Mater. 1983, 31, 1425–1432[18] L. L. Rishel, PhD Thesis, University of Pittsburgh, USA, 1993[19] D. R. Poirier, G. H. Geiger, Transport Phenomena in Materials Processing, The Mineral,

Metals and Materials Society, Warrendale, USA, 1994[20] V. I. Dobatkin, N. A. Anoshkin, Mater. Sci. Eng. A 1999, 263, 224–229

290

Quenching Study on the Solidification of Aluminum Alloys

D. Ruvalcaba1, D. Eskin1, L. Katgerman2, J. Kiersch2

1Netherlands Institute of Metals Research, Delft, The Netherlands2Delft University of Technology, Delft, The Netherlands

1 Abstract

Real-time observation of the microstructure during solidification may show in-situ evolution ofthe morphology in aluminum alloys. However, existing techniques are expensive, time consu-ming and not readily available. Therefore, the metallographic technique coupled with quen-ching of the microstructure during solidification is still considered as an alternative method. Themajor problem in employing this procedure is overestimation of solid fraction as compared withexisting solidification models, e.g. lever rule and Gulliver–Scheil models. The present researchis focused on studying the evolution of the microstructure during solidification employing thequenching technique in order to understand the overestimation of solid fraction and its behaviorduring quenching. Binary aluminum alloys (Al–3 wt% Si and Al–7 wt% Cu) were melted, soli-dified at cooling rates between ~0.03 and ~0.3 K/sec and quenched at different temperatureswithin the semi-solid region. Two quenching rates were achieved, i.e. ~50 and ~100 K/sec. Thesolid fraction evolution was measured by image analysis. Solute profiles over cross-sectioneddendrites were obtained by EPMA. Finally, the results were compared with lever rule and Gul-liver–Scheil calculations.

2 Introduction

It is important to understand the development of the microstructure during solidification of alu-minium alloys since the final properties and defects depend on this development. Commonlyfound defects such as porosity, hot tearing and microsegregation depend on the solid fraction;while permeability and dendrite coherency, which determine the properties of the macroscopicmushy zone, depend on the development of the microscopic mushy zone (e.g. on microsegrega-tion) [1, 2].

Nowadays, methods such as: calorimetric techniques [3], optical-intensity measurements [4]and X-ray microtomography [5] may provide some information about the microstructure devel-opment. Nevertheless, some of these methods do not give information about the morphologicalchanges while others such as X-ray microtomography are not readily available and are still un-der development. On the other hand, the quenching technique has been widely used as an alter-native method to reveal how the microstructure develops during solidification of alloys. Thistechnique needs, however, to be completely understood in order to produce reliable results,since the quenched interface develops instabilities that cause overestimation of solid fraction inaluminium alloys and modifies the original shape of the interface [1, 3, 6–9].

Quenching increases the rate of heat extraction and in turn undercooling at the interface. It isassumed that undercooling promotes the formation of equiaxed grains and a refined structure

291

ahead of the interface, as a result, there is a distinction between the microstructure before (i.e.columnar/coarse dendrites) and after (i.e. equiaxed/fine dendrites) quenching [1]. If quenchingrate is high enough, the microstructure can be frozen and the distinction between solid and liq-uid is only determined by dendrites surrounded by the existing liquid at the moment of quench-ing. It has been found that overestimation of solid fraction is still present, even if interfaceinstabilities formed by quenching are considered as part of the liquid phase. It has been shownthat the overestimation becomes less if the two phase sample is held isothermally and if quench-ing takes place close to the eutectic reaction [6, 8, 9]. Accordingly, the microstructure is notproperly characterized with regard to the original solidification conditions.

Notably, it is important to understand the limitations of the quenching technique in order toemploy it adequately when studying the microstructure development during solidification.

3 Experimental Procedure

Two hypoeutectic aluminium alloys Al3 wt% Si and Al7 wt% Cu were prepared from commer-cially pure ingots (i.e. 99.99 wt% Al, 99.99 wt% Si and Al47.7 wt% Cu). Then, the alloys werecast in small rods which were then cut as 15 mm Ø × 30 mm high and 11 mm Ø × 14 mm highsamples. Two different crucibles were employed, an alumina crucible for the 15 mm Ø × 30 mmhigh samples and a graphite crucible for the 11 mm Ø × 14 mm high samples. The cruciblewith the sample was placed in a holder inside a vertical furnace. Then, a thermocouple was ca-refully placed in a drilled hole in the middle of the sample. After this, the sample was kept in thefurnace until reaching a temperature of ~680 °C and then carefully stirred by tapping the cruci-ble for less than one minute. Then, the sample was again heated up until its temperature reaches~700 °C. After ~2 minutes the furnace was turned off and the sample was cooled down insidethe furnace. Finally, the sample was quenched when a selected temperature was reached. Tem-peratures in the semi-solid interval (i.e. between the temperature of liquidus TL and the tempera-ture of eutectic TE) were selected from the equilibrium diagram. The accuracy of temperaturemeasurements is ~ ±1 °C.

Once the samples were quenched, they were cut, grinded and polished. Then the sampleswere analysed by optical microscopy, where the solid fraction was calculated by image analysiswith an accuracy of ~ ±0.02 of solid fraction. The -Al phase was considered as solid and the eu-tectic and instabilities were considered as the liquid phase at the moment of quenching, assum-ing the formation of instabilities happening after quenching. Finally, line scan measurementswere performed in order to analyse the solute distribution in dendrites (i.e. microsegregation) byElectron Probe Microanalysis (EPMA).

The cooling rates were calculated from the cooling curves recorded during solidification.First, the cooling rate before quenching for each sample is calculated from:

, (1)

where is the cooling rate, TL is the temperature of liquidus, Tq is the temperature of quen-ching, tL is the time at which TL was reached and tq is the time at which Tq was reached.

L q

L q

T TT

t t

T

292

Then the quenching rate was calculated from:

, (2)

where is the quenching rate, T500 is the temperature at which the sample is assumed to becompletely solidified which is 500 °C and t500 is the time at which this temperature was reached.

4 Results and Discussion

The quenching rates achieved were: ~50 K/s for the 15 mm Ø × 30 mm high samples and~100 K/s for the 11 mm Ø × 14 mm high samples. Table 1 shows the selected quenching tem-peratures and cooling rates for the different samples. It can be noticed that most of the samplesshow close initial cooling rates prior to quenching, therefore the analysis can allow comparisonbetween the samples. Also, the cooling rates prior to quenching are very low, therefore the re-sults can be compared with the two solidification models which consider the solidification ex-tremes under normal solidification conditions. These models are equilibrium (i.e. lever rule) andnon-equilibrium solidification (i.e. GuilliverScheil), which are calculated using ThermoCalcSoftware.

Figure 1 demonstrates microstructures of the two alloys quenched at 50 K/s. This Figureshows that the formation of interface instabilities is less pronounced when quenching occurredat temperatures close to the eutectic reaction. The same behaviour was found at 100 K/s, how-ever, at this rate finer instabilities are found at higher temperatures. Figure 2 shows that in thecase of the Al3 wt% Si alloy, the overestimation of solid fraction compared to the lever rule andScheil is less at 100 K/s at temperatures above 615 °C. The sample quenched at 50 K/s and at

500

500

q

q

T TQ

t t

Q

Figure 1: Micrographs of samples quenched at 50 K/s for: a) Al3wt% Si at Tq = 634 °C; b) Al 3wt% Si at Tq = 578 °C; c) Al7wt% Cu at Tq = 637 °C; d) Al7wt% Cu at Tq = 551 °C

293

634 °C did not show a homogeneous distribution of phases which may be due to the size of thesample, therefore, this measurement of solid fraction may not be the one that was present closeto the thermocouple tip. At temperatures below 615 °C, there is no overestimation of solid frac-tion, which may mean that quenching was successful in freezing the microstructure. Also, no in-stabilities were found in these samples quenched at low temperatures. Only the samplequenched at 50 K/s and at 578 °C showed almost complete solidification exhibiting some coarsesilicon particles, attesting for the beginning of the eutectic reaction (Figure 1b).

Table 1: Selected temperatures (Tq) and cooling rates ( ) prior to quenching for Al3 wt% Si and Al7 wt% Cu alloys

In the case of the Al7 wt% Cu alloy, the samples showed to be more susceptible to quench-ing (Figure 1c). Overestimation was only found in samples quenched at 631 °C and 620 °C. Allthe samples showed no difference in behaviour with the quenching rate (Figure 3). The lower-temperature samples showed good agreement with Scheil approximation, which might meanthat back diffusion did not take place.

Line-scan measurements were performed over several dendrites in Al3 wt% Si samples byEPMA analysis. Figure 4 shows the concentration profile of silicon in a dendrite arm for theAl3 wt% Si alloy quenched at 100 K/s at 627 °C. It is noticed a change in the inclination of theslope when it reaches CS = 0.87 wt% Si. After this, the slope changes again at the solid contentat the quenched interface i.e. CSq =1.95 wt% Si. This change of slope takes place at the instabili-ty neck where the instability initiates at the moment of quenching. Several dendrites showed thesame behaviour with a change of slope at the onset of the instability. The solute content CS isclose to that calculated from Scheil approximation (i.e. ~0.87 wt% Si) for a solid fraction of

Al-3 wt% Si Samples Quenched at: Al-7 wt% Cu Samples Quenched at:Quenching temper-ature Tq [°C]

~50 K/s [K/s]

~100 K/s [K/s]

Quenching temper-ature Tq [°C]

~50 K/s [K/s]

~100 K/s [K/s]

634 0.05 0.03 637 0.04 0.06627 0.06 0.08 631 0.06 0.06616 0.08 0.10 620 0.06 0.09595 0.13 0.19 595 0.13 0.19582 0.13 0.19 583 0.13 0.19578 0.13 0.19 551 0.17 0.30

T

QT

QT

QT

QT

Figure 2: Calculated and experimental solidification paths for the Al3 wt% Si alloy

294

0.48 and a temperature Tq = 627° C which is the temperature of quenching of the sample. Howe-ver, the experimentally determined solid fraction for this temperature was ~ 0.75. The overesti-mation is quite high as compared with Scheil. This overestimation may be due to the existenceof large instabilities which formed at the very beginning of quenching and which may have hadsome time to grow and develop without leaving trace. The observation of microsegregation pat-terns and the corresponding changes in the slope of concentration profiles may provide a linkbetween the experimental observed structures and the morphology of the solid phase prior toquenching.

Finally, the samples quenched at 595 and 582 °C showed a solid fraction and CS falling be-tween equilibrium and Scheil, which may mean that the microstructure was frozen successfullywithout the formation of instabilities.

Quenching efficiency may depend on the characteristics of the solid-liquid interface. It is as-sumed that at the beginning of solidification (i.e. low solid fractions), the solid and liquid diffu-sion boundary layers have a large solute pile-up which provides a suitable place for nucleation.The formation of instabilities in the liquid pool may be due to detachment of dendrite branchesthat nucleate at the interface and that are displaced away from the interface to allow the nuclea-tion of more instabilities at the same active interface. The formation of instabilities may beovercome by mixing the liquid, breaking the boundary layers, or by giving enough time forpiled-up solute atoms to diffuse away and distribute homogeneously in the liquid (i.e. asachieved by isothermal holding).

Figure 3: Calculated and experimental solidification paths for the Al7 wt% Cu alloy

Figure 4: Concentration profile of Si for the Al3 wt% Si alloy quenched at 100 K/s and at Tq = 627 °C showing the solute content at the interface before quenching CS and the solute content at the quenched interface CSq

295

5 Conclusions

The study of the development of solid phase at high temperatures in the solidification range im-plies some difficulties when employing the quenching technique. Line-scan measurements thatshow a noticeable change in the solute profile may help in reconstructing the interface at themoment of quenching. By understanding the formation and kinetics of instabilities produced byquenching, it may be possible to uncover and describe the microstructure that forms beforequenching.

6 Acknowledgements

This work is performed within the framework of the scientific research program of the Nether-lands Institute of Metals Research (www.nimr.nl), Project MC4.02134b. The authors would liketo thank W. G. Sloof and K. Kwakernaak from the Department of Materials Science and Engi-neering of TUDelft for their assistance on EPMA analysis.

7 References

[1] Ø Nielsen, S. O. Olsen, Transactions of the American Foundryman’s Society 2002, 110, paper 02-096

[2] J. Thevik, A. Mo, J. Heat Mass Transfer 1996, 40, 2055–2065[3] D. Larouche, C. Larouche, M. Bouchard, Acta Mater. 2003, 51, 2161–2170[4] S. Steinback, L. Ratke. Scr. Mater. 2004, 50, 1135–1138[5] L. Salvo, M. Pana, M. Suery, M. DiMichiel, Ø. Nielsen, D. Bernard, in Proceedings of the

2nd International Light Metals Technology Conference (Ed.: H. Kaufmann) 2005 (in press)

[6] O. Pompe, M, Rettermayr, J. Crystal Growth 1998, 192, 300–306[7] S. W. Chen, C. C. Huang, Acta Metall. 1996, 44, 1955–1965[8] M. Rettenmayr, O. Pompe, J. Crystal Growth 1997, 173, 182–188[9] E. Tzimas and A. Zavaliangos, J. Mater. Sci. 2000, 35, 5319–5329

296

Numerical Study of the Influence of an Applied Electrical Potential on the Solidification of a Binary Metal Alloy

P.A. Nikrityuk, K. Eckert, R. Grundmann Institute for Aerospace EngineeringDresden University of Technology, D-01062 Dresden, Germany

1 Abstract

In this work we study numerically the influence of a homogeneous electrical field on the fluidand heat transfer phenomena at macroscale and mesoscale during unidirectional solidificationof a binary metal alloy. The numerical results showed that a pulse electric discharging appliedperpendicularly to the solidification front leads to a much stronger Joule heating of the liquidphase in comparison to the solid phase. It was found that on the mesoscopic scale the electriccurrent density is not homogeneous due to the complex shape of the dendrite and the differencebetween electrical conductivities of the solid and liquid phases. This inhomogeneity of the elec-trical current density in the mushy zone leads to the increase of the Joule heating of the dendritein comparison to the interdendritic liquid and induces a pinch force (electromagnetic Lorentzforce). The main features of the resulting convection in the interdendritic liquid are discussed.

2 Introduction

Control of solidification of metal alloys is one of the most demanding problems in the electro-magnetic processing of materials. One of the innovative methods of such a control is the pulseelectric discharging (PED). This method allows the modification of the microstructure duringsolidification [1–4]. The main feature of PED consists in a series of electric impulses passingthrough the solidifying melt. Due to the Joule heating caused by passing of an electric current,the temperature of the melt can increase. In the case of an inhomogeneous electrical current theinteraction of the current with its own magnetic field produces a Lorentz force. This phenome-non, the so called pinch effect, received recently considerable attention in magnetohydrodyna-mics [5–8].

The pioneering work of the study of the influence of the direct electric current passingthrough the solidified melt were performed by Mirsa [1]. It was shown experimentally that thedirect electric potential changes the nucleation and growth processes of the solid. But the mech-anism of modification of the grains size were not understood. Nakada and coworkers [2] studiedexperimentally the influence of PED on the solidification structure of Sn15wt%Pb alloy. Theelectric discharging was carried out parallel to the solidification front by means of two cylindri-cal electrodes located along side wall of the cavity. The electric current was non-homogeneous.It was shown that solidification structures were modified from large grains with dendrites to fin-er grains with globular dendrites by means of pulse electric discharging with a capacitor bank. Itwas proposed that the Lorentz force (pinch force) induced at the moment of discharge is respon-

297

sible for the break of dendrites into globular fragments due to high shear stress. But no numeri-cal simulations were carried out to support this hypothesis.

To sum up, our understanding of the complex interaction between electrical current and so-lidified melt is far from being complete. On of the reasons for the still rather empirical applica-tions of PED in unidirectional solidification of metals and alloys is the lack of detailedknowledge of the main mechanisms which are responsible for the grain size modification underthe influence of electric current pulse. Motivated by this fact, this paper presents the first nu-merical study of the influence of PED on the heat and momentum transfer during directional so-lidification of Sn15wt%Pb alloy. In this paper we show the details about the development of theelectro-vortex flow on mesoscale produced by interaction between the electric current passingthrough the melt and its own magnetic field. Furthermore we analyze the influence of the dura-tion of the electric current pulse on the cooling curves.

3 Problem Formulation

To study the influence of the direct current applied during unidirectional solidification of a met-al alloy on the macroscale heat transfer we consider a cylinder with non-conducting side wallsof the height H0 = 0.075 m and the radius R0 = 0.025 m filled with the superheated alloySn15wt%Pb, see Fig. 1a. Between the bottom and the top of the cylinder an electric potential isapplied. Thus a homogeneous electric current flows through the liquid and solid phases perpen-dicular to the solid front. Due to the homogeneity of the electric current there is no Lorentzforce produced by the interaction of the current and its own magnetic field inside of the cavityon the macroscale [4]. The top and side walls of the cavity are thermally insulated, while thebottom is cooled at a rate governed by the instantaneous wall temperature, TW(t), and a uniformand constant overall heat transfer coefficient, :

, (1)W W cq t T t T

Figure 1: Schematic description of the geometry: axisymmetric cylindrical cavity on the macroscale (a) and columnar dendrite of parabolic shape on the mesoscale (b)

298

where TC is the temperature of the cooling media. In this work the values for and TC were setto 10 W m–2 K–1 and 300 K, respectively to represent an intermediate velocity of solidificationof about VS 3 · 10–4 m s–1.

To study the physical processes in the mushy zone on the mesoscale we consider a columnardendrite of a paraboloid shape without resolving the complex morphology of the its boundary,see Fig. 1b. The paraboloid shape is an accepted approximation of the dendrite, c.f. the famouswork of Ivantsov [9]. The size of the domain considered is Rd = 10–4 m which approximates typ-ical dendrite arm spacing for VS 3 · 10–4 m s–1 [10]. The center of cylindrical coordinate sys-tem lies on the axis of the symmetry of the dendrite, and is moving with the velocitycorresponding to the solidification velocity VS. Thus we have the dendrite which is flowedaround the liquid phase which velocity equals to VS, see Fig. 1b.

3.1 Electromagnetic Field Calculation

To calculate the electric current density we use Ohm’s law:

, (2)

where is the electrical conductivity of the mixture of solid and liquid phases, is the velocityvector, is the magnetic induction vector. By mesoscopic consideration of solidification varies stepwise between solid and liquid phases. In the case of macroscopic consideration of the

variation a linear interpolation can be used:

, (3)

where is the volume fraction of liquid. In the liquid phase, corresponding to = 1, the electri-cal conductivity equals to l. In the solid phase corresponding to = 0, the electrical conductiv-ity equals to s. The electric field intensity is

. (4)

Here is the electric potential. To derive the electric potential we use the continuity condi-tion of the electric current:

. (5)

Inserting eqs. (2) and (4) into eq. (5) written in cylindrical coordinates (r, , z), we have:

. (6)

In this work we consider the axisymmetric case, thus .

For the calculation of azimuthal magnetic field, B , we use Biot-Savart’s law

j E u B

uB

1l s

E

E

0j

1 1z rr r u B u B

r r r z z r r z

0

299

(7)

where 0 = 4 · 10–7 H/m is the vacuum magnetic permeability. If there are no external mag-netic fields we have only an azimuthal component of the magnetic induction, B , given by:

(8)

For the better understanding of the dynamics of the electrical filed parameters during solidi-fication we simplify eq. (6) under the following conditions:

1. The electrical current density is homogeneous.2. During solidification there are only solid and liquid phases, i.e. no mushy zone exists.3. The side walls of the cavity are isolated.

In this case eq. (6) has an analytical solution:

(9)

where s–l is the electric potential on the boundary solid-liquid, 0 and 1 are the electric poten-tials on bottom and top, respectively, AS = H0 / HS and As = s / l. Thus eq. (9) allows us to cal-culate electric field intensities, ES, El and Joule heating terms , in solid and liquidphases, respectively. In this work we use eq. (9) for the validation of a solution of the eq. (6),see Section 3.

3.2 Macro-Energy Transport

In this study we restrict ourselves to the hypereutectic alloy Sn15wt%Pb which has the advan-tage of an initially stable stratification with respect to both the thermal and the solutal densitychange during unidirectional solidification. Thus without forced convection, the UDS ofSn15wt%Pb is not affected by thermosolutal convection. Furthermore shrinkage-driven flow isnegligible. Since the homogeneous electric field does not induced convection the energy trans-port equation has the following form [10]:

(10)

where = l + S (1 – ) and cp = cpl + cps (1 – ). The volume fraction of liquid is calcu-lated from the relation [11]:

0

1j B

0

0

R

zB r j drr

j

1 0 11 1

ss l

s

A AA A

2s sE 2

l lE

2pc T T H E

t t

300

(11)

where TS = 183 °C is the solidus temperature, Tl = 218.5 °C is the liquidus temperature. Thelast term in the eq. (10) is the Joule heat.

To further simplify the problem we evaluate the characteristic time scales. Namely the typi-cal solidification time for O(VS) 10–4 ms–1 and O(H0) 10–2 m is O (102 s) . The characteristictime for solute diffusion on the macroscale is given by O ( z2 / Dl) 103 s, where z is the sizeof the control volume (CV) ( z = H0 / 100) and Dl is the diffusion coefficient(Dl = 1.5 · 10–9 m2 s). These different orders of magnitudes justify the neglect of solute masstransport on the macroscale.

Table 1: Physical properties of Sn15wt%Pb alloy

Material properties of Sn15wt%Pb alloy were calculated from a linear dependence on themass concentration of its component and are given in the Tab 1. The material properties of purePb and Sn were taken from [12–14]. In this work we assume that the densities of the solid andliquid phases are identical and equal to 7889 kg m–3.

3.3 Fluid Flow on the Mesoscale

On the microscale the electric current density is inhomogeneous due to the difference between

s and l, and complex form of the dendrites. The interaction between the electric current andits own magnetic field produces the pinch force. This force induces a forced convection in themushy zone. To be able to capture the main flow structure taking place in the mushy zone wepresent a simplified model of the mesoscopic fluid flow. This model includes the Navier-Stokesequations decoupled from heat and mass transport in the liquid and solid phases. Here we as-sume that the dendrite is imbedded inside the heterogeneous fictious domain in which we glo-bally solve the fluid dynamics problem. The corresponding N-S equations are based on theporous media theory, introducing the permeability relative to each phase [15]:

(12)

(13)

where is permeability constant which prescribe immersed boundary conditions. This valuerelated to each phase is defined by

Solid Liquid

Thermal conductivity , W m–1 K–1 57.99 26.2Specific heat cp, J kg–1 K–1 210.85 233.8Molecular viscosity , N s m–2 – 1,873 10–3

Latent heat H, J kg–1 – 54140Electric conductivity , A V–1 m–1 7.48 106 1.8 106

s

l s

T TT T

0u

2L

u uu u p u Ft K

301

(14)

The Lorentz force have the following form:

(15)

Using eq. (2) the radial and axial projections of this force have the form:

(16)

(17)

To justify the neglect of the heat and mass transport we compare the characteristic timescales for heat, mass and momentum transports with respect to the size of the control volume( z = Rd / 100). In particular the characteristic time for the Joule heat transfer in the liquid phaseis O (cp z2 / ) 10–7 s. The characteristic time for the solute diffusion is given byO ( z2 / Dl) 10–3 s and the characteristic time for the viscous diffusion is O ( z2 / ) 10–5 s.This evaluation shows that the mass transport is the slowest process, while temperature diffu-sion is the fastest one. Thus, a temperature perturbation produced by Joule heating is dissipatedfaster than fluid flow appears. Thus we assume that the liquid and the solid phases have thesame temperature. In order to obtain first insights into the fluid flow induced by the pinch forcewe neglect the mass transport of the solute. However, to get the real insights into the transportprocess taking place on the mesoscopic scale during the PED it is necessary to consider momen-tum, energy and solute transports coupled with each other. This task is computationally de-manding and will be done in future work.

4 Numerical Scheme and Code Validation

The set of eqs. (6), (10), (13), (14) has been discretized by a finite-volume finite-difference ba-sed method. The time derivatives are discretized by a three-time-level scheme. The convectionterms are discretized by a central difference second order scheme with deferred acorrection[15]. The system of linear equations is solved by using Stone’s strongly implicit procedure(SIP). SIMPLE algorithm with collocated-variables arrangement was used to calculate the pres-sure and the velocities. Rhie and Chow stabilization scheme was used for the stabilization ofpressure-velocity coupling. More details about the coupling algorithm can be found in [16].

Time marching with fixed time step was used. For every time step the outer iterations werestopped if residual of energy equation is less than 10–4 and less than 10–13 for pressure and mo-mentum equations. Several grid-convergence and time-step-convergence tests were preformedto define proper grids and time steps leading to grid and time-step independent solutions. Forthe macro-scale energy transport simulations we used 20 70 grid, where first and second num-bers correspond to the numbers of CV in the radial and axial directions, respectively, and a timestep of 1 sec for diffusion controlled solidification with a PED duration of 30 sec. For the oscil-

, 10, 0

ifK

if

LF

LF j B

2Lr z rF E B u B

2Lz r zF E B u B

302

lating PED with a period toff = 1 s between switch on and off, we used 20 200 grid with atime step of 0.2 s.

For the mesoscale simulations we used a structured non-uniform 350 350 grid to calculatethe electric filed parameters and fluid flow for the case VS « 10–4 ms–1. The time step used wasset to 10–4 sec. For the calculation of fluid flow for the case VS = 10–4 ms–1 we used a structureduniform grid 250 250 and the time step of 10–5 s.

To validate the code we model the electro-vortex flow induced by an inhomogeneous directelectric current flowing between the sidewalls in a rectangular cavity. We consider the 2D casewhen the thickness of the cavity is much more less than both height and width. This geometry isa simplified variant of the liquid metal current limiter (LMCL) investigated experimentally byCramer et al. [7]. The 2D scheme of the device is shown in Fig. 2a. For the numerical simula-tions we used Cartesian coordinates (x, y). Fig 2b shows the spatial distribution of the velocityvectors induced by the Lorentz force. It can be seen that two large quasi symmetric vortices aregenerated. This is in good qualitatively agreement with the experiment [7]. The vortices in-duced are the product of the interaction between the electric current densities jx, jy and its ownmagnetic field Bz shown in Fig. 2c.

The validation of the solution of eq. (6) is done in the Section 4 by the comparison of the analyt-ical solution (9) with the numerical one.

5 Results

The first series of numerical simulations is devoted to the study of the macro-energy transportduring the solidification of the alloy under the influence of a pulsed electric current. Three caseswere simulated. The first is the diffusion controlled UDS without PED, the second and third oneconcern the UDS with PED and voltages = 0.05 V and = 0.1 V, which was initialized af-ter 50 sec of solidification and stopped after 80 sec of solidification. Fig. 3a shows the coolingcurves obtained at the positions z = 0.02 m and z = 0.065 m from the bottom. It can be seen thatduring application of PED the cooling rate in the solid and liquid phases decreases. For = 0.1the Joule heat is so large that the liquid phase is heated during the PED. To understand the im-

Figure 2: Code validations: scheme of the domain (a), vector plot of velocity scaled with 5 · 10–4 m/s (b) and contour plot of magnetic induction Bz scaled with 0 ( 1 – 0) · 0.5 (c). Here we used 0 = 0 V, 1 = 5 · 10–4 V,

= 106, = 6000, = 2 · 10–3, L0 = 35 · 10–3 m, rc = 5 · 10–3 m, 70 140 grid.

303

pact of the PED on the spatial behaviour of the temperature we plot in the Fig. 4 the axial pro-files of the temperature obtained at 70 sec. In comparison to the first case the temperaturegradient on the boundary between the liquid and the mushy zone is increased. To investigate theinfluence of PED duration on the cooling curves we depict in the Fig. 3b the comparison of T(t)gained for steady current and periodically switched on and off with the period toff = 1 s. It isclearly seen that in the case of an oscillating PED the Joule heating effect decreases.

To understand the increase of the temperature in the liquid phase we plot in the Fig. 5 the ax-ial profiles of the nondimensional electric potential calculated by means of eq. (6) at t = 70 s.For comparison we depict the analytical solution for in the case of a lacking mushy zone, i.e.the boundary between solid and liquid phases lies at Tl. For a better understanding of the Figurewe plot additionally the profile of the liquid fraction at that time. It can be seen that due to thehigher electrical conductivity of the solid phase in comparison to the liquid we have a leapchange in gradient of electric potential, in other words in Ez. Thus the electric field intensity inthe solid phase is less than that in the liquid phase. As a result the Joule heating, , in theliquid is higher than in the solid phase, see Fig. 6. This figure shows the dependency of theJoule heating on the ratio HS / H0, calculated by means of eq. (9).

2zE

Figure 3: Predicted cooling curves: comparison of cooling curves gained for different voltage (a) and com-parison of cooling curves for z = 0.02 V gained steady current and periodically switched on–off current (b). The period toff was set to 1 sec. Here z is the distance from the chill.

Figure 4: Predicted axial profiles of the temperature obtained for different voltages at t = 70 sec

304

The primary interest of the next series of calculations is the prediction of the evolution of theflow field induced by PED in the mushy zone and to study the distribution of electric potential,electric current and Joule heat on the mesoscopic scale, see Fig. 1b. For the calculation of elec-tric parameters we used the following boundary conditions: Between the bottom and the top ofdomain we set the potential difference , taken from the macro-simulations, equal to

= 5·10–4 V corresponding to a electric field intensity in the liquid phase of Ez = –5 Vm–1.Furthermore no current flows through the side walls in radial direction. Fig. 7 shows the pre-dicted spatial distribution of the electric potential and electric current density vectors. It can beseen that due to the difference in electrical conductivities of the solid and the liquid phases thecurrent density in the dendrite is higher than the current in the liquid. As a result the Joule heatincreases in the dendrite in comparison to the liquid, see Fig. 8a. The Joule heat has the maxi-mum value on the tip of the dendrite which is explained by maximum curvature of the surface inthat place.

Figure 5: Comparison of the numerically and analytically calculated electric potential scaled with l at t = 70 sec. Here is the volume fraction of the liquid s / l = 4.16

Figure 6: Analytically predicted dependence of the Joule heating term on the height of the solid phase HS scaled with the height of the cavity H0. Here ’ = l,s / l, E’ = EzH0 / , s / l = 4.16.

305

To study the fluid flow pattern induced by the pinch force we consider two cases. The firstcase corresponds to a very small solidification velocity VS = 10–4 ms–1. On referring to Sec. 2this allows us to use the no-slip boundary conditions on the top, bottom and side wall of the do-main, see Fig. 1b. The velocity vector plot, see Fig 8b, displays a toroidal vortex rotating inclockwise direction. Thus the interdendritic fluid flow washes the dendrite from the bottom tothe tip, and in that way it will modify the solute boundary of the dendrite. As a result the shapeof the dendrite will be changed. The second case is devoted to the growth of dendrite with anintermediate velocity VS = 1 · 10–4 ms–1. Thus, a considerable relative flow around the dendriteoccurs. In this case, the free-slip condition was used on the side wall, on the bottom: the veloci-ty was set to uz = VS and on the top outflow boundary condition was used. The velocity vectorsplots calculated for PED with Ez = –1 Vm–1 and Ez = –5 Vm–1 are displayed in Fig. 9. It can beseen for the induced velocities smaller than the solidification velocity there is no change in theflow around the dendrite, see Fig. 9a. But if a toroidal vortex rotating in clockwise di-rection appears near the tip of the dendrite, see Fig 9b. This vortex produces the upward flowwhich will transport the solute rejected by the dendrite to the upper part of the mushy zone. Wesuppose that this may lead to a larger constitutional undercooling.

maxz su V

Figure 7: Predicted distribution of electric parameters: nondimensional electric potential / (a) and vector plot of electric current density scaled with l / Rd (b). Here Rd = 10–4 m, = 5 · 10–4 V.

Figure 8: Predicted spatial distribution of the Joule heating term scaled with l ( / Rd)2 (a) and meridional

velocity scaled with 2.5 · 10–4 ms–1 (b). Here = 5 · 10–4 V.

306

To calculate the time required to establish a fully developed electro-vortical flow we intro-duce in analogy to [17] the volume-averaged meridional flow velocity as follows:

(18)

Its time history is presented in Fig. 10. The calculations were performed for two values of Ez.We found that the time of the flow establishment for all two Ez is of order O(10–3 s). This is avery promising result since the application of PED with a comparable pulse frequency wouldboth avoid the Joule heating and reduce the consumptive power of the PED devise.

We argue that the fluid flow induced by the pinch force in the mushy zone can cause a me-chanical fragmentation of dendrite. Probably the change of the solute concentration on the den-drite surface caused by the convection causes the constitutional fragmentation. Thus wesuppose that the main mechanism of the grain refinement by the application of PED is related tothe hydrodynamics of the turbulent regime which responsible for the refining of the grains byanalog with works [17, 18,19].

Summing up the results of simulations on macroscopic and mesoscopic scales, we are facedwith a paradoxical situation. On the macroscale there is no convection due to the homogeneouselectric current but on the mesoscale due the complex shape of the dendrite and s / l 1,toroidal vortices appear in the interdendritic liquid which may induce macroscale convection.This assumption needs detailed consideration in the future work. We note that in the case of in-homogeneous electric current on the macroscale a pinch force appears on both macroscale andmesoscale.

6 Summary

The results of the numerical simulations of the heat transport on the macroscale showed that apulse electric discharging applied perpendicularly to the solidification front leads to a muchstronger heating of the liquid phase in comparison to the solid phase (the heating is caused by

2 22

0 0

2 d dH R

rz r zd d

U r u u drdzR H

Figure 9: Predicted spatial distribution of meridional velocity scaled with VS = 10–4 m s–1 for Ez = –1 Vm–1 (a) and meridional velocity scaled with 2.5 · 10–4 ms–1 for Ez = –5 Vm–1 (b)

307

the Joule heating effect). We could show that a shorter duration of the PED pulses decrease theJoule heating of the melt. The numerical studies on the mesoscopic scale revealed that both dueto the complex shape of dendrite and difference in electric conductivities between solid and liq-uid phases the Joule heating in the dendrite is increased in comparison to the heating in the in-terdendritic liquid. The Joule heating reaches its maximal value on the dendrite tip. Theinhomogeneity of the electrical current density in the mushy zone induces a electromagneticLorentz force (pinch force). This force induces a toroidal vortex near the dendrite tip. It wasshown that for the domain with side length of 10–4 m the time required for the flow to be estab-lished has the order of magnitude 10–3 s for Ez of order of O(10) Vm–1.

7 Acknowledgements

The authors are grateful to Dr. M. Peric for the source code of the Navier-Stokes solver. Wethank Armin Heinze for stimulating discussions. Financial support by the Deutsche Forschungs-gemeinschaft (SFB609, B2) is gratefully acknowledged.

8 References

[1] A.K. Mirsa. Metallurgical Transactions 17A, 1986, 358–360[2] M. Nakada, Yu, Shiohara, M.C. Flemings. ISIJ International 30, 1990, 27–33[3] A. Prodhan, C.S. Sivaramakrishnan, A.K. Chakrabarti. Metallurgical and Materials Tran-

sactions 32B, 2001, 372–378[4] M. Gao, G.H. He, F. Yang, J.D. Guo, Z.X. Yuan, B.L. Zhou. Materials Science and Engi-

neering A337, 2002, 110–114[5] V. Bojarevics, T. Freibergs, E.V. Shilova, E.V. Shcherbinin. In Electrically Induces Vor-

tical, Kluwer Academic Piblishers, Dordrecht, 1988, p. 400

Figure 10: Time history of the volume-averaged meridional velocity for different axial electric field intensities Ez in the liquid phase for VS = 10–4 m s–1

308

[6] P.A. Davidson. An Introduction to Magnetohydrodynamics. Cambridge University Press, Cambridge, 2001, p. 431

[7] A. Cramer, G. Gerbeth, P. Terhoeven, A. Krätzschmar. Materials and Manufacturing Pro-cesses 19, 2004, 665–678

[8] I. Kolesnichenko, S. Khripchenko, D. Buchenau, G. Gerbeth. Magnetohydrodynamics 41, 2005, 39–51

[9] G.P. Ivantsov, Growth of Crystals, Consultants Bureau, NY, 1958, 1 , p. 76[10] D.M. Stefanescu. Science and Engineering of Casting Solidification. Kluwer Academic/

Plenum Publishers, New York, 2002, p.342[11] S. Chang and D.M Stefanescu. Metall. Mater. Trans. A, 27A, 1996, 2708–2721[12] The Goodfellow website: www.goodfellow.com. Metals, polymers, ceramics and other

materials[13] Verein Deutscher Ingeniere (VDI) – Wärmeatlas Berechnungsblätter für den Wärmeüber-

tragung, 9 Auflage, Springer-Verlag, Berlin, 2002[14] T. Iida, R.I.L. Guthrie. The Physical Properties of Liquid Metals. Clarendon Press,

Oxford, 1988, p. 288[15] K. Khadra, P. Angot, S. Parneix, J. Caltagirone. Int. J. Numerical Methods in Fluids 34,

2000, 651–684[16] J.H. Ferziger, M. Peric. Computational Methods for Fluid Dynamics, 3nd ed., Springer,

Berlin, 2002, p. 423[17] P.A. Nikrityuk, M. Ungarish, K. Eckert, R. Grundmann. Phys. Fluids 17, 2005, 067101.[18] A. Vogel. Metal. Science 12, 1978, 576–578[19] B. Willers, S. Eckert, U. Michel, I. Haase, G. Zouhar. Mater. Sci. Eng. A., 2005, in press[20] S. Eckert, B. Willers, P.A. Nikrityuk, K. Eckert, U. Michel, G. Zouhar. Mater. Sci. Eng.

A., 2005, accepted for publication

309

Microstructure and Strain Distribution Influence on Failure Properties in Eutectic AlNi, AlFe Alloys

P. Olaru1; G. Gottstein2; A. Pineau3 1 INAV-S.A.-Bucharest Romania2 IMM-RWTH-Aachen-Germany3 ENSM-Paris-France

1 Abstract

Eutectic AlNi, AlFe, alloys exhibit plastic strains to failure (usually in the range of 1%-5%),that those of conventional structural alloys. We have developed a technique to measure strainsat the scale of the microstructure and have used this method to assess the variation in failureproperties with microstructure. This method is capable of using the grayscale information in theimage of a gridded sample to obtain sub-pixel marker displacement, and can therefore accurate-ly determine small strain values. Microstructures that exhibit large variation in local strain dis-tribution tend to have higher variability in tensile properties, particularly tensile ductility,compared to microstructures that accumulate strain more uniformly. Orientation and morpholo-gy of lamellar plates in lamellar colonies play, also, a role in influencing the distribution ofstrain.

Local grain orientation, phase distribution and segregation are factors influencing the straindistribution, and therefore the properties of these materials.

2 Introduction

The combination of specific stiffness and good oxidation resistance at intermediate tempera-tures can provide significant weight saving for certain components. One of the obstacles to theapplication of eutectic AlNi, AlFe alloys components is the relatively low ductility of these ma-terials in tension. Because, this property is not clear explicitly dealt with in component design,some degree of damage tolerance and ductility in generally required so that stress concentrati-ons can be blunted and minor levels of damage do not produce immediate failure. Some scien-tist [1], have shown that a plastic strain to failure of 2.8%, is sufficient to blunt relatively largestress concentration in eutectic AlNi,AlFe alloys, and many of the currently eutectic alloysbeing considered have average strains to failure that meet this relatively low requirement [2].

Much investigated alloys samples, exhibit a large variations in failure stress and strain, butthe results of some samples do not reach the desired 2.8%, [3].

The aim of this work is to examine the influence of microstructure on variations in strengthand ductility in eutectic AlNi, AlFe alloys. This work describes the investigations in INAV-S.A.Bucharest and laboratories F.S.I.M.-U.P.B.

310

3 Experimental Procedure

Master ingots, with chemical composition given in Table1, were elaborated in vacuum induc-tion furnace by melting high-purity metals under argon atmosphere and were casting into gra-phite rods [4]. These ingots were machined into small- bars and placed in high purity aluminumcrucible of 57/47 10–3, outside/inner diameter [4]. Thin samples of eutectic alloys, were ela-borated by directional solidification in XTAL-VAR 97 installation [4], at a constant growth ofR l and a constant temperature gradient at solid/liquid interface of Gl.

Table 1: Chemical composition of eutectic alloys

In order, to eliminate the nucleation and growth of mis-orientated grains along the specimenlength, all thin samples were grown from seeds defined structure and crystallographic orienta-tion [4].

Sheet samples with an olmet geometry and a gage volume of 160 mm3 were prepared fromeach alloy. The surfaces of olmet samples were prepared by a low stress grinding procedure fol-lowed by hand treatment through 680 grit. The samples were electro polished at temperature(–30 C), to obtain a surface mirror finish and avoid hydride formation [5]. The polished sam-ples were gridded by evaporating gold through a 1400 line per inch titan mesh and then tested tofailure in Kammrath-Weiss in-situ tensile stage in Philips SEM 515-EDS. It was selected themicroscope magnification around 92 grains and 2300 to 3000 markers [Eut1-120x; Eut2-120x]were present in each acquired 706 by 468 pixel image. Tested was interrupted at regular load in-tervals to allow for image collection. The image resolution of the technique depends on anumber of experimental parameters including a higher resolution, signal to noise ratio, and thenumber of grayscales present, captured.

4 Results

The results obtained for samples with chemical compositions in Table 1, are shown in Table 2:

Table 2: Results of tensile testing

The value for mapped samples is indicated in parentheses. The surface displacement map-ping technique is an accurate technique for measuring in plane displacements of less than a pix-el, which allows the local strains in a sample to be measured with a high level of accuracy.Stress-strain curves obtained by averaging strains at each load interval for both eutectic samplesare shown in Figure 1.

No Alloy Ni [at%] Fe [at%] Al [at%]

Eut 1 Al-5.7%Ni 5.7 – BalanceEut 2 Al-1.7%Fe – 1.7 Balance

Alloy Samples test Plastic strain to failure Peak strain before to failure

Eut-1 7 1.9-2.4 % (2.6%) ~5.5%Eut-2 9 1.2-1.8% (2.0%) ~3.2%

311

These curve appear in a good shape, despite being measured over only 0.67 mm2 of the sur-face area of the Eut-1, and 0.58 mm2 of the Eut-2. Like the first conclusion, the results becomemore consistent as the level of strain in sample increases, and that there is considerable variationat low strain levels. The error bars used in the graph represent the variation in the strains in thefour quadrants of the analyzed area. The large variation in strain at each load level indicates thatthese areas are too small to provide an accurate measurement of average strain for the entiresample. This local variation in strain increased as the load and average strain increased for botheutectic alloys.

For measuring small strain during in-situ tensile tests in SEM -Microscopy, we introduced amethod, which use pattern recognition algorithms to locate markers on a gridded sample beforeand after deformation.

The distortions in the grid are used to calculate the strains at each marker. Gold grids with amesh size of 20 m were evaporated onto the samples though titanium grids affixed with adhe-sive tape. Electron backscatter images of the samples taken during testing were analyzed using aset of interactive routines written in the Operative Data Language (ODL). After all displace-ment has been determinate in this fashion, a polynomial fit maps the locations of a marker andits surrounding markers in the distorted image to the corresponding in the reference image.

The strains of the sample surface are then simply the coefficients of the polynomial expan-sion, and they are then contour plotted to show a map of the strains in the sample .The strainmaps can be overlaid on the original image or plotted separately.

Some strain maps obtained by this method can be seen in (Figure 2). This technique is to capture the development of strain in our sample, and the progression can

be followed as the load increase. His magnitudes of the strains are similar to those predicted bya continuum model, but the shape of the strain contours depends on the microstructure of thesample, in main manner.

Figure 1: Stress – strain curve from strain mapping, calculated for displacement mapped samples

312

5 Conclusions

This method of measurements useful for isolating the features associated with enhanced localstrain. The distribution of these local strain concentrated regions also apparently results in volu-me effects, where samples with smaller strained volumes higher strength in the absence of ex-trinsic flaws. The surface displacement mapping technique is essentially a two-dimensionaltechnique, so no information about out of plane displacements can be obtained.

As a finale conclusion, an apparent fracture origin may not be the actual fracture origin, sincethe real origin may be subsurface.

Figure 2: Strain maps by pattern recognition algorithm method to locate markers on a gridded sample before and after deformation

313

6 References

[1] G.Gottstein and L.S.Shvindlerman, Grain Boundary Migration in Metals, 1999, CRC Series in Materials Science and Technology, CRC Press Florida-U.S.A

[2] A.D.Rollet and D.Raabe , Comp. Mater. Sci., 2001, 21, 69[3] C.Maurice, “Proceeding of the first joint international Conference, ReX &GG1, Springer

Berlin, 2001, p.123 [4] P.C.Olaru, “Proceeding of the first joint international Conference, ReX &GG1, Springer

Berlin, 2001[5] M.J.Blackburn, J.C.Williams , Transactions of the Metallurgical Society of AIME,

239 (1967), 287

314

Dendrite Coarsening and Embrittlement in Continuously Cast Tin Bronzes

T. VirtanenTampere University of Technology, Tampere

1 Introduction

Tin bronzes (CuSnP) are widely used in different applications due to their good physical andchemical properties. They have excellent durability and resistance to plastic deformation i.e.high strength and toughness. Good mechanical properties are combined with competitive ther-mal and electrical conductivities. Long solidification range and low solute redistribution coef-ficient makes tin bronzes prone to several problems during solidification in casting process. Themain problem in upcasting is a defect called ‘tin sweat’ in which tin and phosphorous segregatebetween dendrites and further onto strand surface forming brittle eutectoid ( + +Cu3P). Alsocoarsening of dendrites is closely related to the phenomenon.

Coarsening means a change in dendritic length scale by disappearance of smaller dendritearms in favor of larger arms. It determines the local chemical composition and therefore it great-ly influences the properties [1]. It can also significantly influence the time necessary for homog-enizing heat treatment after casting [5]. Coarsening rate depends strongly on the fraction ofsolid, time spent in the solid/liquid region, dendrite arm spacing and the temperature [3, 4, 5].Large solidification range results in coarser dendrite branches [6].

Exudations are believed to result from separation of the casting from the chill surface. Dur-ing air gap stage, the shell reheats thereby permitting solute enriched (lower melting point) alloyto exude from interdendritic locations and/or grain boundaries, and to cover the shell surfaceand contact the mold wall [7]. Remelting leads to a considerable increase in permeability [8],while metallostatic pressure, casting temperature and heat transfer coefficient have also pro-nounced effect on the amount of surface segregation [9, 10].

CAS2 software has been developed by Miettinen for simulating phase changes and solutedistribution during solidification of binary copper alloys containing several alloying elements[11].Very important feature of the CAS2 model is that it takes the cooling rate into account. De-pending on alloy composition, cooling rate and dendrite arm diameter, the package determinesthe stable phases (liquid, fcc, bcc, compounds) and their fractions and compositions as a func-tion of temperature.

The aim of this work is to investigate the conditions leading to formation of undesired struc-ture in continuous casting of tin bronzes.

2 Experiments

Coarsening of dendrites and surface segregation phenomenon were studied experimentally byupcasting samples having different tin and phosphorous contents and varying the cooling condi-tions. Cooling condition was varied by changing casting speed and by adding external distur-

315

bance in cooling system. Tensile tests were carried out and the samples were examined withoptical microscope. The amount of eutectoid and coarse dendrites were evaluated and classifiedvisually and compared to the elongation. The sample structures were divided into four differentcategories (0–3), highest number presenting the worst-case change in dendritic length scale orappearance of surface segregation. The method is explained elsewhere in more detail [12]. Dueto the visual nature of the used method, trends presenting coarsening and segregation tendencyare indicative.

3 Results and Discussion

CAS2 package was used to estimate several features related to solidification of tin bronzes.Also an extensive casting campaign was carried out. In this paper the typical features of the re-sults are presented and the findings discussed.

3.1 CAS2 Model – Interdendritic Sn Composition

Due to segregation and formation of non-equilibrium phases, interdendritic tin composition isof great interest. It can be seen in fig. 1 that tin segregates strongly between dendrites during so-lidification. Segregation increases with increasing cooling rate. With the highest cooling rate of1000 C/s Sn composition of the last liquid drop is around 20 wt.% and yet the final Sn compo-sition in the interdendritic region remains low, at about 5 wt.%. It can be concluded that 1 wt.%tin concentration is not high enough for peritectic bcc phase and further brittle ( + ) eutectoidto form at any of the given cooling rates. Therefore the final cast structure is solid solution.

In figure 1b) it is shown that increase in nominal tin composition from 1 wt.% to 8 wt.% re-sults in great difference in the interdendritic composition during solidification. It can be seenthat cooling rate deviating from infinitely slow equilibrium cooling rate does not have majorimpact on the segregation behavior. Even if the cooling rate is very slow, compared e.g. to typi-

Figure 1: Interdendritic Sn composition versus temperature for different cooling rates. a) 1 wt.% and b) 8 wt.% Sn

316

cal upcasting process, only at 1 C/s, tin segregates as much as in the case of very fast coolingrate at 1000 C/s. Regardless of the cooling rate peritectic bcc phase, , is formed and furthertransformed to ( + ) eutectoid, which can be seen in sudden increase in curves at temperatureclose to 520 C. Curves show the maximum Sn concentration between dendrites, not the aver-age.

Cooling rate has more pronounced impact on secondary dendrite arm spacing (DAS) [13,14], i.e. dendritic length scale, than on segregation behavior.

3.2 Coarsening of Dendrites and Surface Segregation

Variation of DAS in cross section of a single sample is very small in the normal upcast struc-ture. In some samples containing 1 wt.% tin almost cellular growth morphology is found closeto the surface when cooling condition is disturbed. Cellular growth can be an indication of veryslow growth velocity [15], which in turn means that heat removal has either been slow or hasdropped suddenly. It can be seen in Figure 2 that in case of severe coarsening of dendrites andsurface segregation, dendrite morphology changes remarkably between the surface and the cen-ter of the sample. In Figure 2a coarse dendritic structure close to the surface is shown. Also con-tinuous films of brittle eutectoid between dendrites can be found. In Figure 1b, structure appearsto be much more homogeneous in the center, which is very typical for every sample havingcoarsened dendrites and surface segregation.

At the slowest casting speed, 0.2 m/min, coarsening close to the surface is found in all tincompositions (Figure 3a). Slow casting speed results in very rapid development of rigid solidshell which in turn results in formation of an air gap. Heat extraction is reduced significantlyand latent heat removal deteriorates. As a result first solidified dendrites get coarse. Faster cast-ing speed improves structure to a certain extent by making it more homogeneous i.e. less coars-ening and surface segregation.

The amount of coarsened dendrites also increases when the tin content is increased. Thismay be due to wider mushy zone and shrinkage. When the length of mushy zone is increasedthe time with both solid and liquid coexist is also increases. Small changes in air gap results insignificant variation in the length of the mushy zone liquid [16]. The increase of the phospho-

Figure 2: Sample (Sn 10 wt.% P 0.01 wt.%) in different locations. a) Dendritic structure close to the surface is very coarse and has continuous films of brittle eutectoid between dendrites and on the surface whereas dendrites in the centre b) are significantly more homogenous. Scale bar 0.1 mm

317

rous content does not increase the coarsening tendency in the normal cooling condition. Themore the alloy contains tin the more of it is in form of ( + ) eutectoid. Increasing length of themushy zone provides long open channels between dendrites in which solute rich liquid canflow. At the lowest tin content no surface segregation is observed (Figure 3b), not even at thelowest casting speed. Reason for this behavior is the fact that no brittle eutectoid is formed atany cooling rate as is indicated in Figure 1a.

It is clear that the disturbance and slow casting speed together strengthen uneven air gap for-mation ( uneven heat removal/-shell thickness –coarsening). Structure close to the surfacebecomes very inhomogeneous also with the smallest tin composition as seen in Figure 4a.

It can be concluded that when disturbance exists, coarsening occurs regardless of the tin compo-sition or casting speed. Coarsening depends strongly on the severity of external disturbance: themore cooling is uneven along the cast wire, the more coarsening occurs.

The difference in the appearance of surface segregation between normal and disturbed cool-ing condition is not big (Figure 3a and 4b), even though some increase can be seen in disturbedcooling condition. Cooling rate has both direct and indirect influence on formation of the sur-

Figure 3: a) Coarsening severity of dendrites in samples versus tin composition and casting speed. b) Appea-rance of surface segregation versus tin composition and casting speed

Figure 4: a) Coarsening severity of dendrites in samples with disturbed cooling condition and b) surface segrega-tion in samples with disturbed cooling condition

318

face segregation. If the cooling rate is slow, dendrites get coarser than in the normal cast struc-ture. In the area of coarse dendrites solute enriched liquid forms long continuous films andpools of brittle eutectoid.

3.3 Elongation

At lower tin contents the difference in elongation between different casting speeds and normal(5a) and disturbed cooling condition (5b) is not distinct. Visible surface segregation of brittleeutectoid is not formed in samples containing 1 wt.% tin, even if cooling condition is disturbed.In higher tin compositions the elongation is first improving with increasing casting speed andafter reaching maximum value it starts to reduce again. Pieche [17] has realized that 45 anglebetween the casting and crystal growth direction gives the best structure for cold-rolling defor-mation. Increasing tin content improves also the strength.

Disturbed cooling condition leads to distinct reduction in elongation in samples containing 8or 10 wt.% Tin. The phenomenon is clear for slower casting speeds. Faster casting speed resultsin better and longer contact between solidifying shell and the mould leading to even cooling ofwire. It can be seen as more homogeneous and smaller dendrite arm spacing along the wire.Also the oscillation of wire surface temperature ceases with increasing casting speed [18]. Ad-dition of phosphorous increases the amount of surface segregation: The higher the phosphorouscontent, the more of it reacts with copper forming copper phosphides. Eutectoid ( + +Cu3P)has very low melting point i.e. it remains quite long in liquid state.

4 Conclusions

Coarsening of dendrites occurs when heat removal drops suddenly after the formation of solidshell. This change in heat extraction can be caused either by slow casting speed or external di-

Figure 5: Measured elongation of as-cast samples versus casting speed and different tin compositions, a) normal cooling condition and b) disturbed cooling condition

a) b)

319

sturbance in cooling system. Even if the increase in alloying elements favor coarsening, coolingrate determines the final dendrite arm spacing. Formation of the surface segregation is influ-enced by all process variables - directly on tin and phosphorous content and indirectly on coo-ling rate. Change in the dendritic length scale is required to obtain severe surface segregation.Coarsening is not sufficient condition alone to significantly reduce the elongation. Presence ofthe brittle constituent is also required.

The cooling rate has more pronounced impact on dendrite arm spacing than on the segrega-tion behavior. It was shown that in spite of the cooling rate, tin segregates in samples containing8 or 10 wt.% Tin between the dendrites with the same final composition leading to formation of( + ) eutectoid. At low tin compositions brittle eutectoid does not form even with very highcooling rate

5 References

[1] L.M. Coluzzi-Mizenko, M.E. Glicksman, R.N. Smith, JOM, 1994, 51–55 [2] V. Rontó, A. Roósz, Int. J. Metals Res. 13(2001), 337–342 [3] T.Z. Kattamis et al., Transactions of AIME, 239(1967), 1504–1511[4] M. Chen, T.Z. Kattamis, Materials Science and Engineering, A247(1998), 239–247[5] K.P. Young, D.H. Kirkwood, Metallurgical Transactions A, 6A(1975), 197–205[6] Suyitno, D.G. Eskin V.I. Savran, L. Katgerman, Metallurgical and Materials Transaction

30A(2004), 3551–3561[7] M. Morishita et al., Light Metals 2000, 657–662[8] H.J. Thevik, Modelling of Casting, Welding and Advanced Solidification Processes VII

1995, 557–564 [9] P. Reisener, S. Engler, Metall 2(1995), 116–122[10] E. Haug , A. Mo, H.J. Thevik, Int. J. Heat Mass Transfer, 38(1995)9, 1553–1563 [11] J. Miettinen, CAS2 - Solidification analysis package for binary copper alloys, User

manual of DOS version 2.0.0. Report TKK-MK-162, Helsinki University of Technology Publications in Materials Science and Metallurgy, Espoo (2004), p.25

[12] T. Virtanen, Effect of process variables on embrittlement and formation of coarsened dendrites and surface segregation in upcast tin bronzes, Doctoral Thesis, to be published

[13] M.A. Martorano, J.D.T. Capocchi, Int. J. Cast Metals Res., 13(2000), 49–57[14] E.A. Kumoto, R.O. Alhadeff, M.A. Martorano, Materials Science and Technology,

18(2002), 1001–1006[15] W.F. Savage, 1980: Welding in the world No5/6 p.8[16] F. Kaempffer, F. Weinberg , Metallurgical Transactions, 2(1971), 2477–2483[17] Piesche R.G., Copper and its alloys, Proceedings of an International Conference, London,

1970, 85–91[18] Kotipelto A., Numerical and experimental analysis of heat transfer in the continuous

casting process of copper, Doctoral Thesis, Tampere University of Technology, 2002, p.138

320

Continuous Casting of Tin Containing Alloys and their Transformation

V. Lebreton, F. Sadi, Y. BienvenuCentre des matériaux P.M.Fourt, Ecole Nationale Supérieure des Mines de Paris, UMR/CNRS 7633, B.P.87, F-91003 Evry Cedex

1 Abstract

This paper deals with binary and ternary tin bronze processing. A first issue is the refinement ofthe as cast microstructure to facilitate homogeneization: vertical casting is usually associatedwith interdendritic spacings of about 100 micrometers while thin slab casting (horizontal usual-ly) reduces the length scale for microsegregation to below 20 micrometers. This is crucial sincethe homogeneization time at a given temperature in the solutioning range varies with the squareof that length scale. Microsegregation modelling is used to account for observations and micro-analyses and to guide process evolutions. The sequence of phase transformation in ternary Cu-Ni-Sn bronzes is presented to optimize the annealing and to reach a compromise strength / duc-tility / conductivity.

2 Introduction

Electrical and thermal conductivities together with corrosion resistance and mechanical strengthmake copper alloys attractive for a broad range of engineering applications (electric, telecom-munication and automotive industries). To reach such properties by conventional processes,each parameter of the processing route (temperatures, times...) must be optimized in function ofthe chemical composition of the copper based alloys. Generally, three principal stages can bedistinguished: first is the as-cast microstructure and the difficulties associated with the solidifi-cation heritage for the rest of the shaping process; the second stage consists in the possibility forthe alloys to homogenize and the last processing stage is the influence of hot or cold deformati-on and annealing treatments to obtain the best compromise between electrical and mechanicalproperties. To obtain this compromise, one must know the influence of the microstructure at allscale on these properties during aging after homogenization. Consequently, the sequence ofphase transformation in solid state is a paramount parameter, which can be studied by D.T.A,S.E.M, T.E.M, and resistivity measurements. The cases of Cu-Sn and Cu-Ni-Sn alloys are veryinteresting because they illustrate the solidification problems, put in light the importance of themicrostructure on electrical and mechanical properties and the requirement to establish in greatdetail the sequence of phase transformations. The first part of this paper deals with the refine-ment of the as cast microstructure to facilitate homogenization in Cu-Sn and Cu-Ni-Sn alloys.The sequence of phase transformation in ternary Cu-Ni-Sn is presented to optimize the an-nealing and to reach a compromise strength / ductility / conductivity.

321

3 The Cu-Sn and Cu-Ni-Sn Alloys in As-Cast Condition

The Cu-Sn and Cu-Ni-Sn alloys are generally processed by melting in an electric induction fur-nace. Some additions such as phosphorus, manganese or charcoal are introduced to minimizethe formation of oxides. The Cu-Sn binary phase diagram illustrates the difficulties associatedwith solidification processing for the rest of the shaping process: a succession of peritectic reac-tion involving brittle intermetallics, a broad solidification interval prone to macrosegregation orto incipient melting in reheating tin containing intermetallics in copper alloys are frequently re-sponsible for brittleness. In the as-cast state and beyond 4 wt% in tin, the + structure of tinbronze transforms during cooling in + and + (Figure 1). Both and are brittle.

The microstructure of a bronze containing 6 wt% tin produced with a horizontal continuouscasting process (section 40 cm x 1.9 cm) presents a columnar zones along the thermal gradient,and the inclination of the grains on the axis observed is associated with the speed of the process.The micrographs show two kinds of segregation :

• A macrosegregation of tin appears at the surface of the metal. This phenomenon is bound to the high rates cooling at the interface metal/mould and to stresses exercised on the metal by extraction rolls.

• A microsegregation at the dendritic scale with the phase forming in the majority the inter-dendritic zone (figure 2).

The measurements of the secondary dendrite arm spacing achieved on Cu-6 wt%Sn and Cu-9 wt%Sn as function of cooling rate permitted to identify the empiric law [3]:

= 320 (dT/dt) 0.58

: interdendritic spacings, dT/dt: cooling rate

Figure 1: Cu-Sn phase diagram [1] Figure 2 : Macrostructure and microstruc-ture in Cu-6wt%Sn alloy [2]

322

The interdendritic spacings are in the range 10–15 μm in the case of alloys produced with ahorizontal continuous casting, 100–150 μm with a vertical continuous casting and some mi-crometers with the Osprey Spray technology.

Similar observations have been achieved on the Cu-15 wt%Ni-8 wt%Sn produced with avertical continuous casting process. Although the macrosegregation has not been observed, theenrichment of the interdendritic spaces in tin element is present (Figure 3).

The segregation of Sn in interdendritic spaces formed during the solidification embrittles thematerial considerably. These phases are crack initiation sites as shown in figure 4. A similar re-sult is observed in the case of the Nordic Gold eurocoin monetary alloy in which the content intin doesn't exceed 1% in mass (Fig.5). The formation of Sn-rich phases is often considered to beresponsible for the hot cracking of copper based alloys. However the optimal mechanical fea-tures of these one are bound directly to the content of tin. Thus, the capacity of these alloys tohomogenize becomes important.

Isopleth sections of the Cu-Ni-Sn ternary equilibrium phase diagram present a single phasedomain with a c.f.c structure. Its extent is a function of the nickel and tin contents. The in-crease of the concentration of nickel has the effect of decreasing the solubility of tin in copperenlarging the two phase domain + to the detriment of the single phase domain.

Figure 3: X-ray microanalysis on Cu-15 wt%Ni-8 wt%Sn in the as-cast condition

Figure 4: Chrysocale Cu-Sn3 wt%-Zn9 wt% in as-cast condition : Sn-rich phase is a crack initia-tion site

Figure 5: Macrosegregation of Sn in Nordic Gold in as-cast condition

323

The difficulties associated with solidification processing for the rest of the shaping processcan be avoided if the alloys can be homogenized to presents a single phase at high temperatures.Another solution is to change radically the process (spray deposition process) but it is more ex-pensive and the productivity is low.

4 Phase Transformation in Solid State: Correlation Between Microstructure and Mechanical Properties

The optimization of the mechanical properties of the copper based alloys cannot be done wit-hout knowing the sequences of phase transformations in the solid state. That means to know theinfluence of each transformation on these properties. In the case of certain ternary alloys such asCu-Ni-Sn, the mechanism of strengthening is more complex and depends strongly on the phasetransformations sequence during aging which can be described as following:

According to this sequence, there are two stages of strengthening : the first is bound to theformation of a chemical modulated structure characterized by an alternance of tin poor and tinrich zones. This microstructure has for origin the spinodal decomposition which preserves thecrystalline structure of the matrix. The variation of stress during this stage follows the empiricalequation established by B. Ditchek et al. in 1978 and true later by Ph. Herman and D.G. Morisfor Cu-Ni15 wt%-Sn8 wt% alloy [6, 7]:

= 0.41 · A · · Y

A: amplitude of spinodal modulation, = 1/a · a/ c: differential size misfit, Y : elastic con-stante

Figure 6: An isopleth at 15wt% Ni for the Cu-Ni-Sn ternary phase diagram [4]

Figure 7: An isopleth at 7wt% Ni for the Cu-Ni-Sn ternary phase diagram [5]

324

With extended aging times this equation is no longer verified. Indeed, according to the phasetransformations sequence in the solid state, the ’-D022 metastable phase coherent with the ma-trix forms from the modulated microstructure. Thus, a second stage of strengthening can beobserved. Figure 8 shows the variation of the ultimate tensile strength with aging temperatureand aging time from a homogeneized state. The comparison between these curves and theT.T.T. diagram for Cu-15 wt%Ni-8 wt%Sn established by Zhao and Notis illustrates well theformation of the metastable phase ’-D022. This precipitation leads to an increment in yieldstress following the equation according to Labush [8] and Janson et al.[9]:

= 3.7 · ef · 4/3 f 2/3· ( /b)1/3

ef: shear modulus, misfit, f: final volume fraction of , particle diameter, b: modulus ofBurgers ve ctor

In the case of Cu-9 wt%Ni-2 wt%Sn and Cu-9 wt%Ni-6 wt%Sn this law has been verified.

For the Cu-15Ni-8Sn alloy, the precipitation kinetics are fast since a treatment at 450 °C dur-ing 15 minutes is sufficient to produce the modulated structure and the phase ’-D022. After 15minutes, a decrease in mechanical resistance is observed due to the appearance of the equilibri-um phase -D03. The latter phase is born at grain boundaries and growths into the matrix withtime. For aging temperatures included in the range [250–400 °C] the L12 phase appears before-D03 with a discontinuous precipitation / morphology. The mechanical response to -forma-

tion is a loss of ultimate tensile strength first, then the ultimate tensile strength increases again.This phenomenon is due to the increasing fraction precipitates of the metastable phase L12.Nevertheless, the mechanical resistance doesn’t seem to reach those associated with a treatmentleading only to the ’-D022 phase.

Figure 8: Effects of ageing time and temperature on the ultimate tensile strength in Cu-15 wt%Ni-8 wt%Sn alloy

Figure 9: T.T.T. diagram for the Cu-15 wt%Ni-8 wt%Sn alloy [4]

325

5 Conclusion

Tin containing copper alloys often present macrosegregations and microsegregations due to thesegregation of tin during solidification. To eliminate these problems a homogenization treat-ment is required. This treatment is shorter for fine solidification structures (horizontal casting).Another solution is to change radically the process (spray deposition process) but it is more ex-pensive and the productivity is low.

The ternary alloys of the system Cu-Ni-Sn in the copper rich corner present complex se-quences of transformations of phases in the solid state complex resulting of a competition be-tween thermodynamic phase stability and kinetic of precipitation. If the ’-D022 phase is oftenmentioned like the hardening phase of the ternary alloys Cu-Ni-Sn by its degree of coherencywith the matrix, its action is accentuated by contribution of the internal stress generated by thespinodal decomposition that occurs before.

6 References

[1] T.B. Massalski, Binary phase diagrams, 1986, 2 [2] F. Sadi, Y. Bienvenu, F. Bacou, R. Bailly, Matériaux 2002[3] T.F. Bower, M.R. Randlett, “Solidification structure of copper alloys ingots”, Metals

Handbook, 9th, 9, ASM, 638–645[4] J.-C. Zhao, M.R. Notis, Acta Mater. 1998, 12, 4203–4218[5] J.-C. Zhao, M.R. Notis, Scripta Materiala. 1998, 11, 1509–1516[6] PH. Hermann, D.G.Morris, Metallurgical and Materials Transactions A. 1994, 25A,

1403–1412[7] M. Kato, T. Mori and L. H. Schwartz, Acta Metall., 1980, 28, 285 [8] R. Labusch, Physica Status Solidi, 1970, 41, 619 [9] B. Janson and A. Melander, Scripta Metall., 1978, 12, 497

Figure 10: -D03 is born to grain boundaries. The '-D022 is observed in the matrix (bright field, T.E.M)

Figure 11: Discontinuous precipitation in Cu-15 wt%Ni-8 wt%Sn (S.E.M)

39

Suplier Session

40

329

Horizontal Casting Technology for Copper Products

W. Müller, P. SchneiderSMS Meer GmbH, Mönchengladbach, Germany

1 Introduction

In companies manufacturing copper tubes new investments might become necessary in thefoundry due to various reasons:

1. Replacement of existing vertical continuous or semi continuous casting plants2. Increase of casting capacity3. Producing copper billets inhouse instead of purchasing

For the selection of a suitable casting machine different process types are available. The ver-tical semi continuous casting requires low investment and operator’s skill with the disadvantag-es of re-melting of head and tail, mostly manual operation and transportation, frequent startingrisk and high probability of unstable production conditions with reduced quality.

These disadvantages are overcome by a vertical fully continuous casting machine, whichhowever, requires a higher investment. As an alternative the horizontal casting process could beconsidered with the advantages of a continuous operation and a lower investment than the verti-cal fully continuous caster, especially regarding foundation, intermediate transportation of rawmaterial or cast billets and labor costs. Furthermore the space requirements are significantlylower compared to typical vertical casting plants.

A further alternative might be the horizontal casting of tube shells with the subsequentprocessing on a high reduction rolling mill. This is an attractive option in those cases where anexisting extrusion press is working already at the limit or where a complete new production lineis in consideration.

2 Horizontal Continuous Casting of Copper Billets

Although the horizontal casting process is well established for casting brass billets for morethan 30 years, its application for producing copper billets is still an exception compared to thevertical casting process. However one new horizontal casting plant for copper billets has beendelivered by SMS Meer Technica in 2005 and recently started the production.

The typical layout of such a horizontal billet caster for copper is shown in Picture 1 withchannel type induction melting furnace, holding furnace for superheating and buffering of theliquid metal, attached cooler / graphite mould, cooling water distribution unit, secondary cool-ing water station, withdrawal unit, movable saw and run out conveyer with storage table. A vid-eo clip may demonstrate the function of the plant.

The one-strand horizontal casting plant for copper billets of a diameter of 300 mm has a typ-ical casting speed of 110–140 mm/min resulting to an output of 4.2–5.3 t/h. At an operating

330

time of 5.400 h/year a total casting capacity of 22.700–28.600 tons per year is expected. Multi-ple strand casters of up to 3 strands are also available today.

Experience has shown that a pre-condition for a reliable and quality-wise good production isthe careful desoxydation of the copper melt and the avoidance of any oxygen pick-up in order toreduce the slag amount occurring with phosphor addition to a minimum. Residual oxygen con-tent will also reduce the lifetime of the graphite die, which can be extended to up to five dayswith a content of oxygen below 10 ppm.

The cast copper billets show a very smooth surface and an internal structure as typically ob-tained by horizontal casting (Picture 2).

Since nearly all copper billets are further processed by an extrusion press to copper tubes, thequestion always comes up about the behavior of horizontal cast billets compared to vertical castbillets in the extrusion press. Concern is the eccentricity of the extruded copper shells, especial-ly in the high ratio extrusion, where the eccentricity of the shell cannot significantly be im-proved during the subsequent drawing process. Extrusion trials have been carried out withvertically and horizontally cast billets of the same outside diameter (305 mm) in a new extru-sion press. The eccentricity of most extruded tubes of size 65 3.1 were in ranges below 5 %independent of the type of casting. It therefore appears that the horizontally cast billets show thesame quality as the vertically cast. However, it must be taken into account, that these trials wereonly based on a few billets and the results need to be confirmed in larger test series.

It should be mentioned that our sister company SMS Eumuco has developed a new press,where the billet is kept in the center by a new centering system, which reduces the eccentricity

Figure 1: Horizontal copper billet caster

331

in a reliable way. Levels of eccentricities mostly below 3 % were reached by extruding alumi-num alloy No. 5052 as shown in Picture 3.

Increased experience in horizontal continuous casting of copper billets combined with newdevelopments in the extrusion process open possibilities to again consider the horizontal castingas a real alternative to vertical casting. The total investment of a horizontal caster is significant-ly lower than for a vertical fully continuous caster and total operating costs are lower than for asemi continuous caster.

Figure 2: Internal structure of horizontally cast copper billet

Figure 3: Eccentricity of aluminum tubes in a 45 MN indirect extrusion press

332

3 Tube Shell Casting

In tube manufacturing an alternative to the conventional process route billet casting and extrusi-on is the directube® process, where tube shells are cast and further processed on a high reduc-tion planetary rolling mill before drawn to final sizes. Picture 4 shows the general layout of thecomplete directube® plant with charging equipment for cathodes and return scrap, two meltingfurnaces, horizontal continuous casting line, which basically consists of a pressure controlledholding furnace with coolers /graphite moulds, withdrawal unit, movable saw, transportationequipment and surface milling machine. The rolling line consists of a loading section, mandrelthrust block arrangement, shell feeding section, planetary rolling mill, movable shear and inlinecoiling unit.

The shown plant was realized in 2000. The horizontal continuous casting line is designed forfour-strand casting of shells with an outer diameter of 100 mm and a wall thickness of 25 mm.Based on the number of strands an annual capacity of copper tubes between 10.000–25.000 t/ycan be reached. This capacity corresponds to a casting speed of 375–450 mm/min.

One important factor for an excellent casting quality is the low content of oxygen. Specialmelting furnaces with a siphon spout in combination with a closed transfer, launder, installed atthe tilting axis of the melting furnace, avoid the pick-up of oxygen.

A further key component within the casting plant is the three-chamber holding furnace withthe pressure control as shown in Picture 5 with its various operation modes.

Section 1 shows the furnace filled with copper at the start of a casting campaign indicatingthe casting chamber, inductor, pressure chamber together with filling chamber. By increasingthe pressure in the pressure chamber during production the metal level in the casting chamberalways remains constant and the constant metallostatic pressure in the mould ensures stablecasting conditions and results in an uniform product quality.

Figure 4: directube® process

333

Section 2 represents the status just before refilling with metal from the melting furnace. Thefresh melt will be nearly completely absorbed in the pressure chamber and therefore turbulencesand splashes of slag into the mould with a negative effect on quality are avoided during the re-filling operation.

At the end of a production campaign before changing the graphite die of the cooler, the pres-sure controlled holding furnace is emptied in the pressure chamber to the minimum level andalso in the caster chamber to just above the cooler as represented in Section 3. Then the pressureis released and the melt level drops below the cooler (Section 4), thus avoiding any metal lossand enabling a rapid change of the cooler/die assembly of a multi-strand casting unit which areall together mounted on a common adapter.

The two plugs at the rear end of the holding furnace are placed for emergency and complete-ly emptying of the holding furnace. During all operation stages the metal content in the holdingfurnace can be monitored with a weighing device in the furnace base frame. This patented meth-od allows automation of the melt transfer from the upstream melting furnace. The advantages ofthis furnace for production and quality are:

1. Constant metallostatic pressure above the mould at all times, irrespective of the filling status of the furnace therefore stable casting conditions and product quality.

2. No slag formation of the metal by using nitrogen in the pressure chamber reduction of slag inclusions in the cast product.

3. Avoidance of bath turbulence during recharging reduction of the risk of break-outs, increase of production safety.

4. Quick mould change, as the metal level drops below the mould due to release of pressure reduction of down time and no metal loss.

Figure 5: Tube shell casting with pressure controlled holding furnace

334

Depending on the required production rate for tube shells the horizontal casting plant isequipped for one up to four strands, which are cut to approximately 18 m single length and fur-ther processed to tubes with coil weights of 1.000 kg. The grain size of the produced shells canbe influenced by the proper adjustment of cooling water flow and graphite design in order toachieve grain size as desired for the final product. Picture 6 shows the micro section of tubeshells with large and with small grains.

Table 1: Production Cost Overview for different copper tube production processes

General Plant Data High Ratio Low Ratio directube

Finish tube capacity T/year 60.000 60.000 60.000Total output (consusmption figures) Factor 1,36 1,36 1,23Net Production hours per year h/year 5760 5760 5760Yield factor % 80 80 80

PersonnelOperators (shift personnel) People 38 35 42Other shift personnel, foreman, technician People 0 0 0Adminastrative and laboratory personnel People 0 0 0Sum required personnel People 38 35 42Average income per year ( /year/person) 32.500 32.500 32.500

Investment costsMachinery and equipment Mio. Euro 26,10 27,00 26,50Tools, spare and wear parts Mio. Euro 0,75 0,65 0,77Foundations, erection and commissioning Mio. Euro 4,70 4,86 4,14Infrastructure, production halls, bay cranes, etc. Mio. Euro 0,00 0,00 0,00Properties and office buildings Mio. Euro 0,00 0,00 0,00Additional investments Mio. Euro 1,31 1,35 1,33Sum Investments Mio. Euro 32,85 33,86 32,73

Consumption figuresElectricity MWh/year 57.466 56.319 52.697Cooling water m³/year 2.039.040 2.240.640 2.027.520Natural gas nm³/year 5.899.034 5.880.012 4.423.357Protective gas nm³/year 338.498 336.838 586.363Compressed air m³/year 676.800 641.520 747.000Lubricants and other consumables Euro/year 65.250 67.500 66.250

Production costsMachinery, foundation, erection, tools, spare partsEuro/ton 68,43 70,55 68,19Other investments (long depreciation period) Euro/ton 0,00 0,00 0,00Additional costs (rental fees, maintenance, etc.) Euro/ton 17,47 17,96 17,69Electricity, cooling water, natural gas, etc. Euro/ton 109,19 108,47 95,05Tools Euro/ton 20,20 17,00 10,42Intermediate storage costs for copper materials Euro/ton 23,92 25,24 19,38Personnel Euro/ton 20,74 19,19 22,82

Sum production costs, without copper Euro/ton 259,96 258,40 233,54

335

By bypassing the extrusion press of the conventional process route the tube shell casting of-fers advantages regarding production costs as shown in table 1, where the production costs ofthe high ratio and low ratio extrusion are compared with the directube® process.

Basis for the sample calculation is the production of 60.000 t/y of sanitary tubes. In order tohave a comparable basis, three directube® lines have to be considered. Cost advantages for me-dia, tools and intermediate storage result to lowest production costs for the directube® process.

4 Conclusion

New concepts and developments in continuous casting of copper for the production of coppertubes contribute to the demand of the industry for alternative cost saving process routes for hig-her product quality and better productivity.

Horizontal continuous casting plants for copper billet and copper tube shell production mir-ror this most important requirement of copper tube makers for medium production capacities.

Figure 6: Macrosections of tube shells of DHP copper, 100/50 mm (small grain sizes and large grain sizes)

336

Horizontal Direct Chill (HDC) Casting of Aluminium – the HE Universal Caster

F. Niedermair, H. ZeillingerHertwich Engineering, Braunau, Austria

1 Introduction of “The Universal Caster”

HDC casting has well earned its place in modern Al-casthouses, and is still gaining momentum.Hertwich Engineering has successfully commissioned some 50 Horizontal Direct Chill CastingPlants (HDCs) world wide to date. Todays generation of HDC-casting machines is one of themost versatile pieces of equipment, which may be employed to produce any of the following:

• T-bar• Foundry ingot• Busbar and anode rod• Extrusion billet• Forging stock• SSM-feedstock• HDC for magnesium ingots etc.•This is why the HERTWICH horizontal caster really deserves the attribute “universal caster”.Let’s look at some of the most commonly produced shapes.

2 T-Bar and Foundry Ingot

Over the past few years especially the mass producers of remelt product have discovered theHorizontal Caster to fulfil their demanding needs in terms of product quality and process con-trol. Large scale production of high quality foundry ingot has been shifted from ingot belts toHDC. T-bar/Ingot casting on VDC casting machines have lost ground to the over the years de-veloped HDC casting process. Figure 1 shows T-bars produced on the HDC casting machine.

The VDC process has the following drawbacks compared to the HDC process:

• Higher costs of the VDC caster, especially due to greater building height required, necessity of 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 is rather labour intensive, whereas with the HDC, continuous production runs of 3 to 20 days are common. For T-bar production only one to two operators per shift are needed (two operators for cast start and stop).

• On VDC plants sawing is not integrated in the process, so that an additional sawing station plus operator is required. HDC casting employs an automatic flying saw, which cuts the T-bars to length without disturbing the casting process.

337

Pouring of metal into open moulds causes dross:Sows, pigs and ingots were traditionally produced employing the open mould technology.

Although this technology was improved over many years, dross formation and inclusions arestill unavoidable. Due to cascading, turbulence occurs when filling the mould. So a relativelybig unprotected surface area is offered to the atmosphere for oxidation. The dross formation ismainly ruled by the metal temperature, pouring height and pouring rate. Values achieved duringproduction of pure aluminium sows are shown in table 1.

Table 1: Dross formation during production of pure aluminium sows

The HDC process is almost free of dross formation. It results in savings due to avoided metallosses and in inclusion-free products. On the HDC the metal flows smoothly, protected by anundisturbed oxide layer via launder and tundish to the mould. Thus leaving no chance for ox-ides and other impurities to get into the product.

The HDC cast T-bar and foundry ingot are chilled at least ten times faster than sows andpigs. This ensures a fine and uniform grain structure as well as a uniform analysis throughout

Pouring height [m] Temperature [°C] Dross formation[kg per ton of poured metal]

approx. 0,2 to 0,3 700–770> 800

0,2–0,40,3–0,6

approx. 0,6 to 1,0 approx. 750approx. 850 to 900

2,5– 5–7

Figure 1: T-bars produced on Hertwich HDC

338

the cast product. A further step ahead in the production of remelt products in terms of quality isthe combination of the HDC process together with an Inline Degasser and Ceramic Foam Filter(CFF). Both items are needed to obtain T-bars and foundry ingot free from porosity and inclu-sions.Summary of advantages of HDC products vs. sow and open mould ingots

• low hydrogen, extremely low oxide inclusions• fine uniform grain structure• consistency and uniformity of alloying elements’ content and distribution• no gravity segregation• no cracks and shrink holes and no water inclusions• consistent dimension, straightness, weight• smooth surface, easy for stacking and strapping, compact bundles

2.1 Design Details

The most common HDC caster has a width of 3.000 mm. The 3 m wide caster may produce upto 13 tonnes per hour T-bar or 8 t/h foundry ingot.

At present HE supplies the largest HDC caster built to date. It is designed to cast 4 strandsT-bar 850 300 mm or 24 strands ingot 106 106 mm simultaneously. Continuous productionis 17 t/h for T-Bar and 12 t/h for ingot.

Figure 2: HDC Casting of foundry ingots

339

A wide range of alloys can be produced, for instance ranging from pure aluminium to 12%silicon and up to 5 % magnesium. (Ref. Figure 2) Each product has a dedicated exit route,downstream of the flying saw (Ref Figure 3).

Provided all exit systems are installed, a product change can be undertaken within one shift,by changing to a different tundish/mould set and loading the new applicable cast recipe on thePC. The fully continuous HDC process ideally lends itself for automation. This advantage hasbeen well exploited by Hertwich Engineering. All downstream equipment is fully on-line withthe casting machine and no additional personnel is required. Sawing, weighing, hard stamping,ink marking, labelling, stacking and strapping is carried out fully automatic. (Ref. Figure 4)

Figure 3: Typical layout of the Universal Caster

Figure 4: Foundry ingot automatic marking, stacking

340

During past years the plants were improved consistently and now feature automatic caststarts and stops as well as automatic tundish adjustment. The HDC plant is controlled by theHertwich PCPLC system, which offers an error-manager system and a menu-type casting reci-pes. Besides, all important plant parameters are monitored, controlled and stored and are availa-ble to a clients host PC for further processing or storage.

3 Busbar and Anode Rod

In the primary aluminium business, a HDC plant is often initially purchased for producing bus-bar for potline construction, but designed to allow later conversion to froundry products. In pha-se two the busbar caster is then typically turned into a T-bar or foundry ingot caster to producehigh quality remelt products for foundries, by adding the relevant handling equipment. Excel-lent electrical conductivity of busbar, good surface finish and straightness are achieved. The in-line flying saw cuts continuously-cast busbars to exact finished length. Busbar lengths of 0,6 mto 20 m can be cast, no additional cutting operation is required and the number of welding jointsis greatly reduced. Furthermore, HDC busbar production cost is significantly lower than VDC.Figure 5 shows the production of busbar on a HDC-Casting Machine.

4 Extrusion Billet, Forging Stock, SSM Feedstock

4.1 Extrusion Billet

To operate their own remelt plant poses a big challenge to most extrusion firms. Hertwich Engi-neering has developed a remelt plant to meet the limited capacity requirement of a typical extru-der. The “Compact Type Remelt Plant” represents an economically very interesting concept,particularly in the range 4.000 to 20.000 tons per year, which comprises an integrated, conti-

Figure 5: Production of busbar on Hertwich HDC casting machine

341

nuously operated, fully automated system starting with the handling of scrap and ending withdelivery of cut to length homogenized extrusion billets (logs). Within recent years Hertwich En-gineering has supplied some 20 such plants to extruders worldwide. Heart of the CTRP is a HEHDC billet caster. Figure 6 shows the schematic of a Two Chamber Furnace and HDC Caster.

The Horizontal Caster is the key machine in the effective HERTWICH Compact TypeRemelt Plant (Figure 7)In-house generated extrusion scrap can be charged by a dedicated charging machine into theTwo Chamber Melting and Casting Furnace. The stationary furnace consist of a melting and aholding chamber. Applying the submersion melting process allows remelting of profile scrap ata metal loss of less than 0,5 %. Primary metal and clean scrap from the market may be remeltedas well.

For contaminated scrap, like painted profiles, HE offers a 3 Chamber Furnace. This furnaceevaporates and combusts the hydrocarbons from the paint prior to melting. Thereby additionalmetal loss is avoided, increasing the thermal efficiency and destroying harmful compounds likedioxins etc. Through a tap hole in the holding chamber, the metal flows via an Inline Degasser

Figure 6: Schematic of a Two Chamber Furnace and HDC Caster

Figure 7: Layout of a CTRP for production of extrusion billets from clean and contaminated scrap

342

and CFF to the Horizontal DC Casting machine. Melt flow from tundish to mould is close to thebottom of the tundish, hence no oxides floating on the surface of the melt may ever get into themould. The mould is of short water cooled design with integrated lubrication for continuouscasting. Size of casting machine and number of moulds determine the production rate.

Depending on the size of machine and billet diameter typical casting rates range from 1000kgs/hour to 3000 kgs/hour. However, smaller or larger machines are available. A billet diameterchange is done fast and easy by changing the set of tundish and moulds. Figure 8 shows theHDC caster during billet production.

Casting campaigns last several days, for instance a casting cycle starts on Monday morningand stops on Saturday morning. Cut to length billets can automatically be pin stamped on thecut face. A flying saw integrated to the HDC casting machine automatically cuts the continu-ously cast strands into billets of the required length. The max. cut length of billets is ruled byexit conveyor length, usually 6 m or 7 m long billets are produced. After Sawing, billets are di-rectly fed into a HE Continuous Homogenizing Furnace for heat treatment.

4.2 Forging Stock (Spaghetti Production)

Traditionally forging stock is produced by VDC casting of regular billets followed by extrudingto the required diameter. On the HE HDC Caster forging stock can be directly produced in dia-

Figure 8: Horizontal Caster with Flying Saw, Billet Production

343

meters ranging from 25 mm to 125 mm, followed by scalping. The main advantages of HDCproduced forging stock are: no extrusion grain texture and substantially reduced conversioncost.

4.3 SSM Feedstock

SSM feedstock is readily produced on the HDC, by fitting an electromagnetic stirrer around themould. This concept has been developed and successfully tested some years ago.

5 Metal Cleanliness

It must be stated that proper upstream operation and equipment for adequate metal cleanliness isa prerequisite for producing a top quality product and achieving economical cast durations, 3 to20 days. Particular attention should be paid to the following:

• sufficient furnace capacity• proper furnace operation: incl. skimming, fluxing, alloying, sufficient settling time, furnace

change over (leaving a heel)• efficient degassing and –refining• adequate twin box type ceramic foam filter• rod feeding for modifying and grain refining.

As a service HE offers to consult on this matter individually.

6 Conclusion

The Universal Caster from Hertwich Engineering has become a familiar sight in cast houses andextrusion shops. Its versatility, the low investment costs involved, the high quality of the pro-duct and the lean operating labour required make this plant unique.The evident trend is away from ordinary sow – or open mould ingot casting and towards the su-perior HDC process - a clear step forward due to quality conscious customer demand.

344

Aluminium-Semi-Continuous Casting Technic, State of the Art

G. J. Brockmann Maerz-Gautschi Industrieofenanlagen GmbH., Tägerwilen, Switzerland

1 Introduction

The first patent concerning semi-continuous casting of aluminium billets, applicated by the for-mer VAW/Germany, took place at Germany in 1936 (Figure 1). Since then many different ideasfor designing casting devices followed. Finally the development focussed on the idea of lo-wering a casting table.

To lower a casting table, the plate may be moved with a thread spindle drive, a rope systemor with a hydraulic cylinder. Here the development focussed on using a hydraulic cylinder. Toguide the casting table with high accuracy, usually a guiding systems with rails at the pit wallswas installed. Due to problems caused by metal particles in the cooling water, sticking betweenbearing shoe and rail, the guiding system is now mostly implemented into the casting cylinder.

The size of casting machines varies from 2 t to about 120 t metal weight per cast. Thenumber of moulds depends on the billet or slab size and may be anywhere in the range from 1 to144 pieces. Typical Mould No.’s for billet moulds are 12, 24, 48, 72 and 96 to match a goodmould pattern. The cast length varies in the range of about 3’000 mm up to 7’000 mm, in singlecases even up to more than 10’000 mm. The billet sizes are between 3” and 14”. A typical billetdiameter is 7”, but also diameters up to 28” are of interest for forging ingots.

The geometry of the cast product may be round (extrusion billets) or rectangular (rollingslabs) or even show an elliptical or t-formed shape. Many modern casting machines allow thecast of billets and slabs with the same machine.

All wrought aluminium alloys are VDC-castable. About 70% of casted alloys is AA 6xxx,but also 1xxx, 3xxx and 7xxx are of interest.

New demands of the automotive industry request casting of tin-containing alloys (tin is ex-changing lead in the futures). Moulds for casting such special alloys are under development.

In the following description, typical design properties of state-of-the-art aluminium-semi-continuous casting machines – or better said: Vertical Direct Chilled (VDC) Casting Machineswill be discussed.

2 Casting Machine Layout

Main components of a state-of the-art casting machine (Figure 2) are:

• Casting pit (equipped with access ladder, tubing and pumps)• Base frame with pit shield (top of casting pit, mostly with integrated evaporation)• Pit covering platform• Safety devices• Mould table frame (mobile water frame, movable and/or tiltable)

345

• Casting cylinder (single action, hydraulic, mostly internally guided)• Platen (mostly with torque limiter)• Starting head base (roof or lattice design)• Starting heads (for billets) or blocks (for slabs)• Anti-tilt grid guard• Mould table (carrying billet or slab moulds)• Mould System (Hot Top, Airsol Veil©, Airglide© and similar)• Launder system (for side or central mould feeding)• Hydraulic power unit (separately for casting cylinder and actuators)

Figure 1: First VDC Patent

346

• Cooling water system (with cooling tower, tanks, filters and oil separator)• Mould oil supply system• Mould air supply and control system (if required)• Switch board and control desk (mostly with visualisation)

The components will be discussed in detail as follows:

3 VDC Components

3.1 Casting Pit

The casting pit is looking like a big tank, mostly with walls from concrete, sometimes from steelor partly with an internal steel cladding, submerged under the plants ground level.

The dimensions depend from the size of the casting table and the desired billet or slab length.A typical shape of a 25 t VDC may be about 4’000 x 2’500 mm square and 7’000 mm depth, theresulting volume is about 70 m3 then. Smaller pits have an extension at one side in the near-ground area to get access to the place under the casting table, bigger pits have a separate parallelpump pit with a door between both pits at the ground. The pit is equipped with cooling water

Figure 2: VDC Components

347

pumps with tubing and an access ladder. In the pit’s ground is usually a center bore for receiv-ing the casting cylinder. The bore’s length is another time about the same like the pit’s depth.The bore is inside cladded with a steel tube with an inner diameter like the cylinder’s outer di-ameter plus spacing.

3.2 Base Frame

The base frame is placed at the top of the casting pit, incorporated into the cast house foundati-on and made from mild steel profiles, corrosion-proof painted. The base frame includes the pitillumination with waterproof halogen lamps. Furthermore, the base frame has slots incorporatedfor the connected steam extraction device. An axial fan, installed into the steam collection sy-stem, guides the extracted steam through galvanized ductwork releasing the steam through astack into the atmosphere outside the building.

On the base frame a pit shield is mounted guiding splashing cooling water and liquid metalinto the casting pit.

3.3 Pit Covering Platform

A pit covering platform closes the open pit area during billet or slab removal. So the operatorcan handle the billets or slabs without danger. For small casting machines it can be moved byhand, bigger machines get a motor-driven platform, secured by proximity switched, connectedto an interlock system.

3.4 Safety Devices

To prevent operators from falling into the open pit, fences or chains according to internationalsafety rules are implemented. Such fences can be actuated manually or hydraulically.

3.5 Mould Table Frame

The mould table frame incorporates the cooling water system and is made from rectangular hol-low cross section from stainless steel. A painting is not necessary. The water connections to theexternal system are done by automatic couplings or bearing inlets. The connections to themoulds are realized by flexible hoses. The design may be done in three different ways to meetthe operators need:

• Movable: the table is rolling on rails and can move side wards• Tiltable: the table is fixed with bearings at one side and can be tilted-up by means of hydrau-

lic cylinders• Movable/tiltable: the table drives on rails into a tilt station beside the caster and can be tilted-

up there for easy maintenance work

348

Two double acting hydraulic cylinders tilt the frame from operation to upright position, max.85o to prevent internal pollution of the moulds.

3.6 Casting Cylinder

State of the art is an internal guide of the ram to assure a minimum rotation of the cylinder rodover the whole length of the stroke (Figure 3.). The maximum rod rotation is typically ± 4’ arc.

The casting cylinder is mostly a hydraulic single action ram type cylinder, secured to a heavysteel plate, which is incorporated into the casting pit foundation. The cylinder must be designedfor maximum rigidity with excentric loads on the platen. The ram may be fabricated from stain-less steel or from mild steel with a multi metal-oxide coating to guarantee a smooth, low-fric-tion lifting and lowering motion. The ram is solid up to a diameter of about 580 mm, biggerdiameters are designed as tubes with a 100 mm wall thickness, filled with steel balls after erec-tion.

3.7 Platen

The platen is the mounting platform for the starter head base and fabricated of mild steel rectan-gular hollow cross sections with a coating approved for molten aluminum service. Stainlesssteel flat bars, machined after welding onto the framework, offer accurate mounting surfaces forthe starter head base. The platen carries magnetic end switches for an accurate positioning of theplaten in the pit.

Between casting cylinder and platen, a torque limiter, is installed. This safety device protectsthe internal guiding of the casting cylinder in case that a piece of metal is jamming betweenplaten and casting pit wall, causing a high torque on the system. A shear pin is breaking then,releasing the platen for free rotation to prevent overload of the internal guiding system.

Figure 3: Casting Cylinder

349

3.8 Starting Head Base

There are two different designs for the starting head base:

• Roof top box design (mostly used for slab casting)• Lattice design (mostly used for billet casting)

The Base is made from mild steel with a special coating approved for molten aluminiumservice. The welded heavy-duty steel structure is precision machined as complete unit on boththe top and bottom surfaces to ensure proper alignment between the starting heads and themoulds.

The starting heads are mounted on supports with self aligning devices.

3.9 Starting Heads

The starting heads for billets are made from steel, aluminium or steel with a special shaped alu-minium top. They are movable and will be centred while beeing lifted up into the moulds bymeans of ribs or three-ball self centring devices (Figure 4.).For rolling slab casting the starting blocks are normally made from aluminium. The alignmentis done with centring pins. The centring is done by fixing the position of the moulds initially byhand.

3.10 Anti Tilt Grid Guard

A steel frame prevents the toppling of the billets once the mould table has been removed. It ismounted in the pit between the mould table and the starter head base.

Figure 4: Starting Head (right: inserted)

350

3.11 Mould Table

The mould table accepts the moulds and all piping for water, oil and air (if necessary). It may bepart of the moulds themselves (AirslipTM) or only support platform for separate moulds (AirsolVeil® and Airglide®). Mould table with mounted moulds and starting head base with installedstarting heads are to be handled as aligned units.

3.12 Mould System

• Hot Top: A feeder top ring from refractory material (Hot Top) reduces thetemperature drop during casting to improve the billet’s properties; approved technology for simple demands. Actual versions of this technology incorporate a continuous lubrication instead of grease coating of the bearing surface.

• Airsol Veil®: Hot Top mould with a modern air/oil mist supply forming a cushion of parting agent in front of the bearing area to reduce the cooling rate to form a thinner shell zone of the billet (Figure 5.).

• Airglide®: Design similar to Airsol Veil®; an additional solid graphite ring in the bearing area provides a non-wetting area, reduces therefore the oil consumption and minimizes the billet shell zone (Figure 6.).

• Others: F.e. AirslipTM incorporates a semi-permeable graphite ring for direct air and oil sup-ply. The effect on the billet properties is similar to the Airglide® technology.

3.13 Launder System

The launder distributes the molten metal to the connected moulds. It is a welded steel construc-tion with insulated pre-cast refractory sections; important is the minimizing of the temperature

Figure 5. Airsol Veil® Mould

351

drop between launder inlet and the final mould furthest away from the inlet. Launder systemsare equipped with preheating devices to keep the system dry between the casts.

3.14 Hydraulic Power Unit

The hydraulic power unit comprises oil reservoir, pumps, filters and valves. The system is ne-cessary for lifting-up the platen with mounted parts and for actuating f.e. the hydraulic cylindersfor tilting the mould table frame. Dropping the platen is done by its own weight and controlledby an electrical precision valve to enable a casting speed in the range of about 20 to300 mm/min. It is recommendable to operate the system with a flame retardant hydraulic fluid.

Additionally an emergency platen lowering valve for manual operation in case of an electri-cal failure is installed in reach of the operator.

3.15 Cooling Water System

The cooling water passing through the moulds is collected in the casting pit. The maximum wa-ter level is simply controlled by overflow, the minimum water level of 1’000 mm above pitground according to safety rules is controlled by a floating switch. Various water levels in thepit can be controlled by means of pressure sensors or several floating switches at the pit’s wall,controlling the drain pumps. An automatic back-flush filter is recommendable as well as an oilseparator to keep the cooling water clean. Separated hot and cold well cooling water tanks toge-ther with a cooling tower outside allow individually adapted cooling water temperatures by me-ans of a mixing valve. An additional pulsation valve, commonly used for casting rolling slabs,

Figure 6: Airglide® Mould

352

improves the cooling effect. Finally an insulated emergency over-head water tank is necessaryin case of an electrical failure to finalize a cast according to safety rules.

3.16 Mould Air and Oil Supply System

For moulds operated with parting agents a separated oil supply system must be installed, usingmostly biodegradable vegetable oils. Since semi-permeable graphite rings tends to plugging,mineral oil must be used then. The mould air must be clean and dry and is supplied by an aircompressor with air dryer.

3.17 Switch Board and Control Desk

Switch board and control desk allow to operate the casting machine in a convenient and safematter. The control desk is IPC and PLC supported and usually equipped with a visualization ofall caster functions. Connections to the plant’s Ethernet network allow improved functions(Scada). Operation, recipe administration, alarm logging and trend representation are to behandled in a user friendly way.

4 Product Demands

Geometry, dimensions and tolerances of the cast products are to be kept within narrow limitsaccording to agreed rules like DIN EN standards and others.

The surface quality depends on alloy, mould technology and casting parameters. The shellzone is depending from the mould design and has to be mechanically removed by milling/turn-ing afterwards – if necessary. The structure of billets and slabs may be improved by a followinghomogenization.

5 Economical/Ecological Aspects

Investments and caster operation costs have to be minimized. All parts should be robust with along lifetime and easily exchangeable in case of damage. With view to ecological demands noserious pollution should be caused by operating a VDC casting machine.

6 Development

The continuous improvement process in designing a VDC casting machine must be focussed at:

• Cost reduction of equipment and operation • Improvement of product properties• Extension of the technology to f.e. new alloys (i.e. tin alloys)

353

OCP Crucible Monitoring System in Long-Term Tests

W. Schmitz, F. Donsbach,Otto Junker GmbH, Simmerath / GermanyH. HoffLios Technology GmbH, Cologne / Germany

1 Introduction

One key design feature distinguishing induction furnaces from other heating equipment is thefairly thin ceramic lining between the live water-cooled copper conductor and the molten metalbath. Depending on the furnace size, the thickness of this ceramic lining varies between 10 and15 cm; it diminishes noticeably as a result of wear or crucible erosion. Inductor insulating mate-rials such as insulating varnish and bandages are heat resistant up to about 150–200 °C. If over-heating occurs at this point, the insulation may become damaged or even electricallyconductive, resulting in interturn short-circuiting of the coil. The coil repair effort required inthis case will render the furnace inoperative for several days, even if a spare coil is on hand. Inthe worst case, which has rarely been documented but is nevertheless a possibility, the melt maypenetrate all the way through to the water-cooled coil with all attendant risks of a furnace break-through and ultimately, a steam explosion.

These considerations, together with furnace users’ economically motivated demands for amaximum service life of the ceramic furnace lining, call for a technology which permits a “isu-al” inspection of the gap between the ceramic furnace lining and the induction coil. Addressingthis requirement, various technical solutions for monitoring the crucible have been proposedand implemented in the past.

2 Overview of Conventional Crucible Monitoring Systems

The most important of these is the classic earth leakage monitoring system. In this technology, ad.c. or a.c. voltage of a defined, fairly low frequency is applied to the induction coil and the sy-stem measures the current flow to earth. For this purpose the molten metal bath must be earthedvia an earthing rod in the bottom of the crucible. This earth fault monitoring system, althoughby now a standard feature on virtually all induction furnaces, has a number of disadvantages.For one, it is not selective, i.e., defined tests and disconnection steps must be carried out whene-ver an earth leakage is detected so as to determine whether the fault has occurred between thecoil and the molten metal or in any other part of the equipment, such as in the switchgear oreven in the water recooling system. Another disadvantage of this earth leakage monitoring me-thod is that in the event of infiltration or penetration of molten metal to the coil, evidence of thiscondition will be not be obtained until fairly late. As a result, the furnace must be emptiedquickly if a current flow between the melt and the coil is detected. In any event, minor damageto the coil may have occurred already. Despite these drawbacks, earth leakage protection sy-stems remain an essential safety tool in the operation of coreless induction furnaces and are un-

354

likely to be supplanted by more recent measuring technology. Earth leakage monitoring willalways remain in use as an ultimate safety feature since naturally, the entire furnace includingthe switchgear and water recooling system must be continuously checked for earth faults.

The use of thermocouples between the hot-face lining and the coil levelling mix, as well as inthe furnace bottom, is another technique employed. However, this method can yield only spotmeasurements (at least if the cost and effort involved is to be kept within reasonable limits) andis therefore not capable of monitoring the entire crucible.

In the past, wire netting in various geometrical configurations has been placed between thecoil levelling mix and the hot face lining. The idea is to detect an electrical continuity betweenan advancing tip of molten metal and the net. One particular disadvantage of this method is thatit provides no trend indication, i.e., no advance warning is given. Moreover, it is virtually im-possible to identify the fault location, and numerous spurious faults will be detected since thenet, as an electrical measuring device, is subject to many stray voltage and current effects in themagnetic field of the induction furnace.

A further process in industrial use1), 2) relies on the use of sensor grids comprising an array ofmetallic electrodes in a comb-type configuration. These electrodes are used to measure the elec-trical resistance of the ceramic lining. As this resistance is temperature-related, it is possible toinfer the temperature in specific crucible segments. Fault locations can thus be identified in rela-tion to the furnace circumference, and an advance warning functionality is obtained. However,given the resistance-based nature of the measurement, the system must be adapted to the specif-ic ceramic material used. The readings obtained will be affected by any change in the composi-tion of the ceramic lining, moisture effects, and furnace-induced magnetic interferencesinfluencing the electrical resistance measurement.

3 OCP Optical Coil Protection System

OCP (Optical Coil Protection System) stands for a latest-generation temperature measurementand monitoring technology which relies on fibre-optic sensors. Given their properties, such sen-sors are perfectly suited for interference-free monitoring of the crucible on induction meltingfurnaces. Figure 1 shows the typical crucible structure of a coreless induction furnace, with theOCP sensor cable firmly grouted into the furnace's permanent lining which is installed directlyon the coil.

Based on an optical fibre, the system utilizes a quantum-mechanical effect, the so-called RA-MAN effect, for temperature measurement. Laser light of a suitable wavelength and modulationfrequency is injected into the optical fibre. This laser light scatters on the bonding electrons ofthe solid state structure over the full fibre length and is detected as a backscatter spectrum. Thisspectrum contains the RAMAN lines, the intensity of which is a function of vibration levels inthe solid state fibre structure, which in turn depend on temperature. A new, patented ’optical ra-dar’ technique makes it possible to detect these lines locally and to measure an exact, high-reso-lution temperature profile through the optical fibre. Thus, OCP is a unique crucible monitoringsystem which enables us for the first time ever to determine the temperature field in the induc-tion furnace irrespective of refractory type and design. By selecting a radial resolution of 60measuring points, it is possible to represent the temperature curve in the manner of the familiaranalogue clockface (Figure 2).

355

By adopting an appropriate configuration of the sensor grids, the crucible can be verticallydivided into several regions, although only a single optical fibre is used in all cases. Points ofparticularly high temperature, e.g., due to infiltration, erosion or cracking in the crucible, canthus be accurately localized and checked for potential hazards to the coil insulation. The sensorcable is currently rated for a maximum continuous operating temperature of about 260 °C,

Figure 1: View of a typical crucible structure, showing the permanent furnace lining with the OCP sensor cable embedded (4)

Figure 2: OCP System monitor screen

356

which far exceeds the maximum temperature resistance of the coil insulation. The measuringsystem has a range of several kilometres, and a position resolution of 27 cm (related to thestraight length of the optical fibre). The temperature resolution of this system is better than 1 K.

4 Design and Installation of the OCP Sensor Cable

The core of the OCP sensor cable, first of all, consists of a commercially available high-temperature glass fibre of the type commonly used in telecommunications. For mechanicalprotection, this fibre is enclosed in a stainless steel tube measuring 1.2 mm in diameter. Thetube, in turn, is coated with a high-temperature insulating compound. The overall diameter ofthe sensor cable is 5 mm.

In order to provide the fullest possible crucible sensor coverage in the direct vicinity of thecoil, it is desirable to have a maximum length of sensor cable in the furnace. This is achieved byplacing the sensor cable on the inside of the coil in meandering curves, using ceramic deflectorelements, while taking into account the cable's minimum bending radius. Figure 3 illustratesthis layout on a 6-tonne steel melting furnace. Here we have two meandering cable layers, onefor the upper and one for the lower crucible region. As mentioned earlier, the cost and effortinvolved in providing extra layers is minimal. Once the cable is installed in this manner, theusual former is placed in the coil and a permanent lining made of high temperature resistantcorundum concrete is cast, with the sensor cable thus embedded therein. For new equipmentand coil overhauls, this can be done in the workshop. In the case of retrofits and under specialcircumstances, this work can even be carried out locally in the foundry.

Since the measurement is taken at the same end at which the laser light is introduced, it willsuffice, in the most basic case, to bring only one cable end out of the furnace. However, it hasbecome standard practice to bring out both ends of each meander layer (Figure 4), which are

Figure 3: Arrangement of the OCP sensor cable on the coil of a 6-tonne induction furnace or melting steel

357

then connected to a terminal box mounted on the furnace body. This has certain benefits at thecommissioning stage and also facilitates diagnosis. Moreover, in the unlikely event of a fibrebreaking inside the furnace due to inappropriate mechanical loads, it is thus possible to conductthe measurement in the reverse fibre direction, thus ensuring that the whole furnace can still bemonitored.

In a next step, the intended measuring end of the sensor cable is connected to an optical fibretransfer cable armoured to foundry standards. This cable is run to the location of the evaluatorand connected to one of its ports. The evaluator consists of the actual measuring device in a 19”configuration plus a PC for visualization and evaluation. Both are habitually mounted in an ap-propriate panel or cabinet. Alternatively, the display and evaluation functions can be assigned tothe visualization system of the melting furnace where this is technically feasible.

5 Display of Measured Temperature Data

The main screen of the OCP visualization system is shown in Figure 2. It displays a schematictop-down view of two furnaces, each comprising two meander layers, any one time. If morethan two furnaces are monitored, the user can freely select which of these should be displayedon the left and right-hand side, respectively. As a general rule, the temperature curves for all in-dividual meander layers are initially rendered in a screen window. For a less cluttered view, in-dividual layers can be suppressed. This has been done in Figure 2, where the left image showsthe current temperature distribution in the upper layer of Furnace 1 while the image on the rightgives the current temperature distribution in the bottom layer of Furnace 2. As shown, each ofthese temperature profiles can be rendered in a polar or linear view. It is also possible to displaya “relative” mode, i.e., a profile generated in relation to a given historical (reference) profile. In

Figure 4: Lead-in bushing of an OCP sensor cable on the coil

358

this case the user will see the current temperature deviation from the selected reference profile(offset view).

By selecting a playback function and entering a date in the respective window, past tempera-ture profiles can be viewed at any time. It is also possible to show temperature profiles in an an-imated or “video” mode between a user-defined start and end point. In this animated mode, thetemperature profiles can be rendered on the basis of absolute or offset values or in a “maxi-mum” mode. In the latter case the system examines where maximum average temperatures haveoccured during a selectable time period (e.g., a shift or crucible campaign). The temperatureprofiles associated with such events are then shown successively in animated form. The soft-ware can also display the temperature curve at individual measuring points, or the average tem-perature, over time.

6 Evaluation of Temperature Measurements

Figure 5 shows the screen for entering alarm parameters. For every parameter, the user can entera warning threshold and an alarm trigger threshold (possibly resulting in a furnace shutdown).The individual alarm criteria are “temperature”, “deviation from average”, “temperaturechange” and “uniformity”. At a more detailed level, these criteria are monitored as follows:

6.1 Temperature

The temperatures at one or more measuring points are monitored for overruns exceeding thesepreset thresholds.

Figure 5: OCP screen template for the definition of alarm thresholds

359

6.2 Deviation from Average

The measurements from which the temperature profile is plotted are initally processed into anaverage value representing the mean temperature in the respective zone. The system thenchecks whether the temperature at one or more measuring points deviates from this mean valueby more than the preset threshold.

6.3 Temperature Change

Here the system determines whether a time-related temperature gradient, defined as a thres-hold, is exceeded at one or more measuring points. The unit in which this threshold is set in thevarious input fields is °C/min.

6.4 Uniformity

The uniformity parameter is largely identical with the “deviation from average” criterion, ex-cept that the averaging step and check for threshold overruns is not carried out in a single stepover the entire circumference of the furnace. Instead, the system initially examines a radial sec-tor (“pie wedge”) whose thickness is defined by the user in angular degrees ( ) in the windowmarked “step”. This sector is then analysed in the same way as for the “deviation from average”criterion. In an interative process, this evaluation window is then advanced in a clockwise direc-tion one measuring point at a time. The analysis is continued until the evaluation window hascovered the entire circumference of the furnace. This is a valuable criterion when it comes to di-stinguishing local flaws from large erosion areas.

In the OCP visualization system, warning and alarm messages are indicated by appropriatesymbols in the temperature profiles (refer to Figure 2). The symbols can be suppressed to obtaina less cluttered view, although the underlying alarm functions remain active.

7 Practical Operating Examples

OCP systems are now successfully in use in coreless induction furnaces for melting copper al-loys, aluminium alloys, cast iron and steel. In the following part of this paper I will examinethree examplary cases in which crucible wear and premature crucible failures were detected in atimely and accurate manner.

7.1 2.5-tonne Vacuum-Type Coreless Induction Furnace for Melting Copper Pre-Alloys

This particular furnace is run in three-shifts to produce copper-iron pre-alloys. Such alloys poseexacting demands on the crucible material due to their aggressive chemical characteristics andfluidity. The tapping temperature is in the region of 1500 °C. The furnace is usually operatedwith a ready-made crucible consisting of refractory concrete. The space between this crucible

360

and the coil is backfilled with a dry ramming compound. A normal crucible campaign lastsabout 2 weeks, depending largely on the degree of sintering of the backfill mix. If sintering pro-

Figure 6: Temperature profiles on a 2.5-tonne vacuum-type induction furnace at the start (left) and end (right) of a trouble-free crucible campaign, shown in absolute temperature (a) and offset display mode

361

pagated too far towards the coil the crucible will be very difficult to break out; moreover, therewill be an increased likelihood of molten metal penetrating all the way to the coil in the event ofa crack formation in the crucible. The degree of backfill mix sintering grows over the cruciblecampaign, causing the thermal conductivity of the backfilling material to increase progressive-ly. As a result, the temperature in the permanent furnace lining and hence, the temperature localto the OCP sensor cable, will rise steadily. Figure 6a shows the temperature profile at the start(left) and near the end (right) of a normal crucible campaign in which no apparent local cruciblefailure occurred. Figure 6b gives the same data rendered in offset mode. Once a critical maxi-mum temperature had been identified over several crucible campaigns, the OCP system wasused as an indicator to identify the need for a scheduled re-lining.

Figure 7 shows the situation for a crucible that had been in use for a week, i.e., half the nor-mal crucible campaign. The graph on the left plots the absolute temperatures; its right-hand sidecounterpart gives an offset view of the same development. Over a span of a few charges, over-temperatures increasing from one charge to the next were identified in the 5 o’clock position,and the system eventually generated alarms of the “deviation from average” type. The cruciblewas broken out, and a crack was found at this point which had allowed the melt to infiltrate theramming compound.

7.2 6-tonne Coreless Induction Furnace for Melting Stainless Steel

Figure 8 shows the temperature profile measured in the lower regions of a 6-tonne inductionfurnace for steel near the end of a crucible campaign. The “deviation from average” alarm mes-

Figure 7: Temperature profiles obtained after crucible cracking with resultant infiltration, again shown in absolute temperature (left) and offset display mode

362

sages (left) indicate general erosion towards the furnace spout. The “uniformity alarm” messa-ges (right) point to the formation of caverns at four points. Measurements conducted on this

Figure 8: Temperature profile in the lower regions of a 6-tonne induction furnace for steel near the end of a crucible campaign. The alarm messages are of the “deviation from average” (left) and “uniformity” type.

Figure 9: Result of crucible measurements, showing an integral top-down view of the crucible. Uniform premature wear in the direction of the spout (12 o’clock) and local erosion in the 2, 5, 8 and 11 o’clock positions are readily apparent.

363

crucible prior to break-out confirmed the condition detected by the OCP system (Figure 9). Fi-gure 10 is a photo of one such cavern.

8 Detection of Cracks in the Crucible

When a crucible cools down, e.g., over a weekend, numerous cooling cracks will form naturallydue to volumetric contraction of the crucible material. These cracks will normally close again,due to thermal expansion of the crucible material, the next time the furnace is started up. Howe-ver, an appropriate heating curve must be used to ensure this. Otherwise, the progressively mel-ting metal may spontaneously penetrate still-open cracks and come dangerously close to thecoil. The situation becomes even more critical if the furnace is filled with liquid metal beforethe cracks have closed.

To simulate this situation, the following test was caried out: A thick-walled steel cylinder ofa diameter equivalent to the inside diameter of the crucibles normally used in this applicationwas placed in the middle of an unlined 1-tonne-furnace. This steel cylinder exhibited “artificialcracks” at a level about halfway up the furnace coil, these being in the form of 3 mm thick and100 mm wide pieces of steel plate welded to the cylinder in the 12, 3 and 9 o’clock positions.The plate in the 3 o’clock position was welded to the surface horizontally and ended about10 mm short of the furnace's permanent lining. An identical plate was welded on in the 12o’clock position, but in a vertical direction. In the 9 o’clock position there was anotherhorizontal plate which extended to the permanent lining. Finally, a 100 100 30 mm steelplate representing a cavern was welded to the cylinder in the 6 o'clock position.

Figure 11 shows a sketch of this arrangement.

Figure 10: Caverns in the lower regions of the crucible

364

The space between the steel cylinder and the furnace's permanent lining was filled with aquartzite dry ramming compound. The furnace was then switched on and operated at about350 kW. This procedure was intended to simulate a cold start with existing metal-filled cracks.(It should be noted, with regard to the following illustrations, that the steel cylinder was slightlyoffset in a counterclockwise direction, so that the individual messages appear not exactly at 12,3, 6 and 9 o’clock, respectively, but again, with some counterclockwise offset).

Figure 12 shows the temperature situation before the furnace was heated up, with the graphon the left giving absolute temperatures and the one on the right shown in offset mode.

Figure 13 illustrates the conditions measured 10 minutes after heating was started. A charac-teristic deformation of the temperature profile is already evident in the offset view. Figure 14shows the temperature profiles recorded 7 minutes later. The first “uniformity” alarms arepresent for all “cracks”. After 30 minutes, all “crucible defects” are clearly identifiable and re-ported by the corresponding alarm messages (Figure 15). It should be mentioned that the smalltest furnace allowed us to embed only a one-layer sensor cable of limited length, which gives aninferior position resolution. The position resolution will naturally be higher on a larger furnace.However, the unusually good temperature resolution of the measuring method is impressivelydemonstrated.

Figure 11: Thick-walled steel cylinder with welded-on steel plates simulating crucible cracks

365

Figure 12: Temperature profiles obtained before heating was commenced (9:30 a.m.)

Figure 13: Temperature profiles after 10 minutes (9:40 a.m.), showing first evidence of deformations

366

Figure 14: Temperature profiles after 17 minutes (9:47 a.m.), showing the first alarm messages

Figure 15: Temperature profiles after 30 minutes (10:00 a.m.), with all defects duly reported

367

9 Summary

The essential advantages of the Optical Coil Protection (OCP) system can thus be summarizedin the following key-words:

• Full protection against– Operational breakdown due to coil damage– Bodily injury and equipment damage due to molten metal breakthrough

• Recording and visualisation of temperature profile over the entire crucible campaign– Indication of developments and trends of refractory wear or metal penetration– Possibility to take action in good time to extend refractory life

• Direct temperature measurement, not resistance-based– Fully operational with a vast range of refractories and immediately after relining

• Optical (i.e., non-electrical) measuring method• Eliminates false signals or even sensor grid damage by the magnetic field of the induction

furnace• One single evaluator can monitor two furnaces, for example in tandem installations• Very high resolution, e.g. 60 spots over the circumference of an 8-tonne furnace crucible,

like the second marks on a clockface• Temperature measurement with a resolution better than 1 K• This distributed optical-fibre temperature measuring method has evolved into a mature

system which has been demonstrating its reliability as a central safety system for years in more than 300 installations worldwide.4), 5)

In closing, it should be noted that the use of an OCP system is by now means limited tocoreless induction furnaces as discussed herein. The system is also suitable for large-areatemperature measurements in other industrial furnace applications, e.g., to monitor theinductors on channel-type induction furnaces.

10 References

[1] Hopf, M.; Elektrowärme International - Edition B. Industrielle Elektrowärme 50 (1992) No. B2, B229-B23

[2] Hopf, M.; Gießerei 80 (1993) No. 22, pp. 746-751[3] „Verfahren zur Auswertung optisch rückgestreuter Signale zur Bestimmung eines strek-

kenabhängigen Meßprofils eines Rückstreumediums“ [Process for evaluating optical backscatter signals to determine a path-dependent measuring profile from a backscatter-producing medium]

[4] EP 0692705, 1995, Dr.-Ing. U. Glombitza[5] “Fire Protection Systems for Traffic Tunnels Under Test” Proceedings 12th International

Conference on Automatic Fire Detection, Rudolf Maegerle, Siemens Building Technolo-gies Ltd., Cerberus Division

[6] Tsakiridou, E.; “Mit Hightech gegen den Tod im Tunnel” [High-technology against tun-nel fatalities], VDI-Nachrichten, No.10, p. 3, VDI-Verlag 2001

368

A New Continuous Casting Process

H. Sommerhofer, P. SommerhoferSommerhofer Technologies, 8010 Graz, Austria

1 Abstract

Continuous casting using water as coolant is state of the art, but still has some disadvantages,one main problem is the building of vapor at low temperatures. In order to prevent these disad-vantages, we use liquid metal to cool the billet.

Laboratory scale experiments have been done to investigate the possibility to cast an alumi-num alloy, a magnesium alloy and copper using a low melting liquid metal as cooling material.

Now a pilot scale plant has been constructed by Sommerhofer Technologies. After several tests with the pilot scale plant, results on the castability of aluminum- and mag-

nesium alloys are existing now.Advantages of the new process: Much lower crack risk; constant high heat transfer coeffi-

cients; higher casting rates; larger process window for the coolant temperatures; nearly no sub-surface layer; no risk of explosions; no contaminated cooling water; exchanged heat at usabletemperature level.

2 Introduction

The cooling process in continuous casting is one main influence parameter in continuous ca-sting. Usually the cooling process consists of two or three steps. The first one is the cooling stepof the melt in the mould resulting in a thin shell containing melt in it, when the billet is leavingthe mould. The second cooling step is direct cooling of the billet with water, this is done in verydifferent ways. On the one hand casting defects should be prevent, on the other hand the castingrate should be as high as possible. One problem is that water as cooling medium is boiling atvery low temperatures – so the cooling effect is very strong at this low cooling water tempera-tures leading to a high temperature gradient between billet core and billet shell and hence tohigh stresses resulting in centre line cracks. Therefore an upper border for the casting rate forone casting alloy in one dimension is given by the building of centre line cracks.

It is possible to weak the direct cooling step by blowing away the cooling water at a definedheight under the mould exit or to withdraw the cooling water by additional units but it is notpossible to change the cooling character of water, when getting in contact with a hot surface.

Now, what happens, when water gets in contact with a hot surface? Looking at figure 1 wecan see, the way of the casting melt trough the mould getting solid and the curve of the heattransfer number for a conventional hot top mould, how it is used for continuous casting of alu-minum [1]. Looking at this graph we can see, that the heat transfer number may be diminishedin the region of the hot top, immediately below the hot top the heat transfer may increase toabout 10.000 W/m²K, decreasing a short distance below, where the new formed shell is strong

369

enough to shrink against the metallostatic pressure of the casting melt inside the shell. This re-sults in an air gap reducing the heat transfer number to values of about 100 W/m²K.

In the first region of direct cooling the billet surface temperatures are high, much higher thanthe boiling point of water. If the surface temperature is higher than the Leidenfrost temperature,a closed steam layer builds around the billet, blocking the heat transfer between coolant andcasting. When the surface temperature falls below the Leidenfrost temperature the steam layerbreaks partially and the heat transfer number increases to a maximum, where no steam is on thecasting surface. This maximum can be at 20.000 W/m²K or higher but also lower, depending onthe kind of secondary cooling (film cooling, spray water cooling, quenching or two phase cool-ing) and its operating parameters. Especially on the possibility to optimize the direct coolingprocess using water as coolant many investigations have been done and papers may be found inthe literature.

Figure 2 shows the possibility to change the heat transfer number for spray water cooling bycontrolling the cooling water impingement density. One can see that the curve of the heat trans-fer number may be moved upward applying higher cooling water impingement density, but thecharacteristically strong change of the heat transfer number when coming down from high sur-face temperatures (a few hundred W/m²K) to low surface temperatures (order of magnitude isabout 104 W/m²K) may not be changed. So the cooling rate of the casting is changing very

Figure 1: Heat transfer number for heat transfer from the casting to the cooling environment during the cooling process illustrated how it is mentioned in [1] with figures for the heat transfer number

370

strong when the outer part of the casting is solidifying leading to differences in grain size andquality of the billet.

So the tendency of water to boil at low temperatures is not good for the quality of the billet,furthermore too high amounts of water steam are influencing the casting process – steam mayrise up into the mould. In worst case the use of water as coolant may lead to a water steam ex-plosion or an oxyhydrogen explosion. In order to prevent the disadvantages of water as coolantone has to take another cooling material like a heat exchanger oil, gas, a molten metal or a mol-ten salt.

3 Experiments

Weighing advantages and disadvantages these different cooling materials we decided to take alow melting metal in liquid state as coolant for the continuous casting process. A more detaileddiscussion about the choice of the cooling material may be found in [3]. First experiments weredone with the experimental set up shown in figure 3. This is a very simple set up for investiga-ting the possibility of casting a billet in an insulating mould, cooled in a pool of low melting li-quid metal. In these tests three different classes of materials were cast each one representing aspecial feature of this material class. The materials cast in this first experimental set up wereMagnesium (AZ91) representing the class of reactive metals, aluminum AA6063 representingthe class of materials with high heat of fusion and pure copper representing the class of materialwith higher melting point.

Experiments using the apparatus in figure 3 have shown the principal feasibility to cool acasting during the continuous casting process in a pool of low melting metal in direct contactwith the billet solidifying in an insulating mould.

As a consequence of the low convection in the cooling bath the cooling rates were not veryhigh. A possibility to increase the cooling rate is to increase the convective motion in the cool-ant. Therefore a closed coolant cycle was applied and the coolant was brought into contact with

Figure 2: The Heat transfer coefficient as a function of surface temperature and cooling water impingement den-sity V for spray water cooling how it is shown in [2]

371

the billet at a defined velocity and in a defined angle to the billet. This resulted in doubling themaximum casting rate of tests with the first experimental set up.

The next step was to optimize the geometry of the coolant distribution unit and the wholemould design as well as testing different mould materials for casting of aluminum alloys. A dis-advantage of the first test apparatus was, that the coolant temperature could not be held constantbecause there was no place for a heat exchanger in the small apparatus we used. One additionaldisadvantage was the inaccuracy of the billet withdrawal unit.

The construction of a pilot scale continuous casting machine was the next step to get condi-tions near that of industrial work. In figure 4 a schematic depiction is given of the pilot scaleplant constructed and installed by Sommerhofer Technologies.

In figure 4 the coolant cycle is shown – it is a closed cooling system, which is held inert byan inert gas in order to prevent oxidation of the coolant. The coolant at its operating temperatureis pumped into the cooling box, where it takes up heat from the billet. After getting out of thecooling box the coolant runs through the heat exchanger, gives up the dissipation heat and re-turns to the coolant storage tank at operating temperature. The dissipation heat from the heat ex-changer can be used in the plant as process heat because the temperature level is higher than200°C. Depending on the size of the casting plant this amount of heat can save much money forenergy.

Figure 4 shows the coolant cycle and the necessary devices for vertical continuous casting,while Figure 5 shows the set up for horizontal continuous casting. It is possible to change thecasting direction at our pilot continuous casting machine by turning the roller bed with flyingsaw from vertical direction into horizontal direction and changing the tundish. So this continu-ous casting plant and this process in general is flexible concerning casting direction but alsoconcerning different casting materials.

Figure 3: Experimental set up for initial tests on continuous casting with liquid metal cooling device [3]

372

Furthermore the addition of lubricant or gas in order to reduce the friction between mouldand billet is necessary in most cases. Experiments have shown, that some aluminum alloys maybe cast in an insulating mould without lubrication but not all.

Anyway, casting with cooled mould allows highest casting rates, so the next step was to de-sign a cooled mould. First experiments with a cooled mould have shown that it is impossible tocast billets in cooled moulds without lubrication. In cases where the start up succeeds, big sur-face cracks are the consequence of no lubrication. So initial tests with manual lubrication shortbefore casting were done for a first determination of the best lubricant. Than the mould wasadopted with a continuous lubrication system in order to enable continuous casting.

Figure 4: Schematic depiction of the pilot scale plant for continuous casting on nonferrous metals using a low melting liquid metal as cooling material

Figure 5: Arrangement of tundish, cooling box, roller bed and flying saw for horizontal continuous casting using a low melting liquid metal as coolant (depiction without cooling cycle and lubrication system)

373

4 Results and Discussion

The results of the first test series may be found in Table 1, one can see that the maximum ca-sting rates for the first experimental set up were not very high, but it could be doubled by incre-asing the convective motion of the coolant when getting in contact with the billet surface.

It was the aim of this initial tests to get an overview about the important process parameters,the advantages and its borders. Answers to most of these questions were gained by doing the ex-periments with the set up in figure 3.

Table 1: Conditions for the first test series and gained maximum casting speed for working with insulating mould and direct cooling in a bath of liquid metal of about 160 °C

Table 2: Comparison of continuous casting on the pilot scale casting machine with insulating mould to casting with cooled mould

*) 40 cm/min is not the maximum casting rate for casting this alloy with cooled mould it is higher but not determi-ned until the this paper is written.

Table 2 shows the difference in casting rates when working with insulating mould to castingwith cooled mould for AlSi12Cu4NiMg. Although the maximum casting rate for a cooledmould is higher than 40 cm/min the difference is very impressive.

AA 6063 AZ 91 Pure Copper

Liquidus temperature [°C] 655 598 1083Solidus temperature [°C] 576 468 1083Casting temperature [°C] 690 630 1200Billet diameter [mm] 50 50 30Max. casting rate [cm/min] 11 7 11

AlSi12Cu4NiMgInsulating Mould

AlSi12Cu4NiMgCooled Mould

Liquidus temperature [°C] 586 586Solidus temperature [°C] 506 506Casting temperature [°C] 600 600Billet diameter [mm] 50 50Max. casting rate [cm/min] 13 > 40*)

Figure 6: Surface of a AlSi12Cu4NiMg billet cast with cooled mould at 40 cm/min and a billet diameter of 50 mm gained by vertical continuous casting, how it can be used as forging feedstock

374

In Figure 6 the surface quality of a AlSi12Cu4NiMg billet of a diameter of 50 mm is shown,one can see that the region of start up is very short and the surface is smooth after a short dis-tance. So the butt crop for this process is very low in percentage compared to the conventionalVDC process.

Experiments are showing that the maximum casting rate is higher than that of conventionalprocesses for aluminum alloys and will be much higher for copper and copper alloys due to thehigh and constant heat transfer number from the billet to the cooling metal.

In Figure 7 the uniform grain structure of a AA6082 billet cast with insulating mould withoutgrain refiner is shown.

Figure 8 shows the smooth surface of magnesium alloy AZ91, gained by casting with insu-lating mould without lubrication.

Figure 7: Macro structure of a AA6082 billet cast with insulating mould at 15 cm/min without grain refiner, bil-let diameter 50 mm gained by vertical continuous casting, how it can be used as forging feedstock

Figure 8: Smooth Surface of a Magnesium AZ91 billet of 50 mm diameter cast with insulating mould at a cast-ing rate of 7 cm/min

375

Figure 9 compares the macro structure of two copper billets cast in insulating mould andcooled in a pool of liquid metal with two different casting rates. It may be seen that the structuregets very fine at higher casting rate although the casting rate is small compared to that of castingwith cooled mould and convective contact with the coolant. So this process is able to improvethe grain structure of copper billets without grain refiner.

5 Conclusion

Much research and development has been done in order to gain a continuous casting process forindustrial application showing the following advantages:

• Constant high heat transfer number higher casting rate for aluminum alloys and much hig-her casting rates for copper and copper alloys

• Much lower crack risk (Temperature gradient between core and surface much lower)• Uniform grain structure from the core of the billet to the surface• Start up phase very short• Dissipation heat may be used as process heat (T > 200°C)• Cooling water treatment unnecessary• Explosions are impossible (Very important for reactive alloys of Mg or Al-Li)

This list of advantages is the reason, why this process is a revolution for continuous castingof non ferrous metals.

Figure 9: Comparison of macro structure of pure copper cast with apparatus shown in figure 3 with different casting rates, left picture 7,5 cm/min, right picture: 11 cm/min

376

6 References

[1] E. K. Jensen, „Mathematical model calculations in level pour DC-casting of aluminium extrusion ingots” in C. J. McMinn, Light Metals, AIME, 1980, p. 631–642

[2] R. U. Jeschar, u. Reiners et.al., „Wärmeübergang in der Sekundärkühlzone von Strang-gießanlagen“ in E. Lossack, Stranggießen, DGM-Informationsgesellschaft Verlag, 1986, 91–114

[3] H. Sommerhofer, Ph.D. Thesis, University of Leoben, 2003

377

Akshentsev, Yu.N. 265Anger, G. 87Anglada, E. 202

Bach, Fr.-W. 81, 95Bainbridge, I.F. 106Bast, J. 112Bender, W. 174Bienvenu, Y. 320Boender, W. 189Boller, K.H. 51Bombach, E. 112Brandt, R. 174Brockmann , G.J. 344Burghardt, A. 189

Chang Hung-Ju, 213Chao Long-sun, 213Commandeur, B. 101

Davey, K. 70Donsbach, F. 353Drezet, J.-M. 151Duck-young Hwang, 137Dundar, M. 23, 87Dürrschnabel, W. 51

Eberle, R. 226Eckert, K. 296Ellendt, N. 249Emmerich, H. 162Eskin, D. 271, 277, 283, 290

Frankenberg , R. 51Friedrich, B. 3

Garmendia, I. 202Geant a, V. 42Gottstein, G. 309Grandfield, J.F. 106Greenberg, B.A. 265Gremaud, M. 151

Gruber-Pretzler, M. 194, 219Grün, G.-U. 174Grundmann , R. 296

Haga, T. 70, 131, 143Hanada, K. 77Han-shin Choi, 137Hatsushikano, K. 77Hepke, M. 81Hoff, H. 353Honsel, C. 209Hoon Cho, 137Hwang Weng-seng, 213Hyung-ho Jo, 137

Jeong, H. 124Jurgk, M. 162

Katgerman, L. 271, 277, 283, 290Kayikci, R. 36Kazantseva, N.V. 265Keles, O. 23, 87Khoury, A. 51Kiersch, J. 290Kim, G. 124Kim, M. 124Koga, N. 70Kräutlein, C. 3Król, J. 118Krone, K. 3Krug, P. 101Kudashov, D.V. 256Kumai, S. 131, 143Kumar Nadella, R. 277

Lai Yi Lin, 213Landaberea, A. 202Lebreton, V. 320Louhenkilpi, S. 240Ludwig, A. 194, 219

Authors

378

Mäkinen, M. 234, 240Matsuzkai, K. 77Mayer, F. 194, 219Moiseev, J. 194Morishita, M. 29Müller, H.R. 51, 256Müller, W. 51, 329

Nagaumi, H. 182Neuer, G. 174Niedermair, F. 336Nikrityuk, P.A. 296

Oelmann, H. 51Olaru, P. 309

Pan Wensen, 213Park, J. 124Pedrós, P. 202Pineau, A. 309Porten, M. 51

Rappaz, M. 151Ricken, H. 209Rode, D. 51Rossberg, A. 81, 95Ruvalcaba, D. 290

Sadi, F. 320Savas, O. 36Schacht, S. 95Schliefer, H. 51Schmitz, W. 353Schneider, P. 329Schneider, St. 51Schwarze , M. 51

Shae K.Kim, 137Shimizu, T. 77Siquieri, R. 162Sommerhofer, H. 368, 368Specht, E. 118Stefanoiu, R. 42Streitenberger, P. 168Suvanchai , P. 182

Takeda, Y. 182Tockner, J. 249Tokuda, K. 29Tonn, B. 194Torisaka, Y. 77Turchin, A.N. 283

Uhlenwinkel, V. 249Umeda, T. 182Uoti, M. 234

van Klaveren, E.P. 189Vapalahti, S. 240Väyrynen, P. 240Virtanen, T. 314Voiculescu, I. 42Volkov, A.E. 265

Walter, M. 249Watari, H. 70, 131, 143Wolber, P. 51Wu, M. 194, 219

Zauter, R. 256Zeillinger, H. 336Zöllner, D. 168

379

3004, surface defects 29

AA6016, automotive applications 87AlBiZn alloys, numerical simulation 194Al-4.5% Cu alloy, melt flow effects 283alloys– 3004 29– 6061 277– aluminum 101, 131, 174, 271, 277, 283,

290– binary 296– casting speed 277– continuous casting 240, 320– crystallization conditions 265– eutectic 309– high strength 182– hypermonotectic 194– magnesium 70, 81– melt flow 283– molten 42– numerical simulation 202– quenching study 290– roll casting 143– strip casting 77– tenary 174– thermal conductivity 174aluminum – 6061 277– casting 151, 344– conductivity 174– DC Casting 271– electromagnetic casting 124– Fe alloys 309– high strength alloy 182– high strength alloys 101– horizontal direct chill casting 336– macrostructure 283– melt treatment 3– microporosity 36– nickel alloys 309– roll casting 143– solidification 290– strip casting 77, 131

automotive applications, AA6016 87

billet – liquid-metal coding 368– near net shape 182– steel 124– sump characteristics 271binary metal alloy, solidification 296binary tin bronze 320bronzes, dendrite coarsening 314bubbles, physical characteristics 42

calcium, strip casting 77caster, universal 336casting – aluminum 36, 131– continuous see continuous casting– continuous strip 70– direct chill see direct chill casting– electromagnetic 124– high speed 143– horizontal direct chill 336– liquid-metal-cooling 368– magnesium alloys 81– non ferrous metal 112– numerical simulation 151, 189– preworks 3– semi-continuous 344– simulation 209– strip 70, 77– tin bronzes 314– tin containing alloys 320– vertical 202casting processes, modelling 151casting speed 277casting table 344casting technology, copper 329chill casting, magnesium 95columnar solidification, macrosegre-

gations 219columnar to equiaxed transition 283conductivity, thermal 174

Subject index

380

continuous casting 51, 112, 137, 151, 194, 209, 213, 219, 226

– alloys 194, 202– copper 240– dhp-copper 234– magnesium alloy 70– metal wires 213– modelling 151– new process 368– simulation 226– sn-bronze 219– state of the art 344– technology 51– tin bronzes 314– tin containing alloys 320cooled mold 137cooling process, alloys 368copper– casting 151, 234, 240, 329– high purity 137– macrostructure 283– melt treatment 3crystallization conditions, alloys 265

DC casting, see direct chill castingdefects – casting 271– formation 271, 277– surface 23dendrites– coarsening, tin bronzes 314– growth 162dendritic length scale, changes 314deoxidized high phosphorus copper 234DHP-copper, see oxidized high phosphorus

copperdirect chill casting 189, 336– 3004 29– aluminum alloy 271, 277– magnesium 95

electrical potential, binary metal alloy 296

electromagnetic casting 124embrittlement, tin bronzes 314equiaxed dendrites 162

eutectic alloys, properties 309

failure properties, eutectic alloys 309FEM simulation, alloys 182fluid flow – effects 240– simulation 226

gas, inert 42gas bubbles, physical characteristics 42grain growth, monte carlo simulation 168grain refining 29, 277– surface defects 23grain structures, calculation 226

HDC casting, see horizontal direct chill ca-sting

HE universal caster 336heat transfer – copper 240– simulation 240high purity copper rod, fabrication 137high speed roll casting, alloys 143high strength alloys 101– Al-Mg-Si 182horizontal casting, copper 329horizontal direct chill casting 336hot top mould design 106hypermonotectic albizn alloys 194

induction furnaces 353inert gas, molten metal 42interacting dendrites 162

liquid metal cooling 368

macrosegregation 271– Al-Cu alloy 283– aluminum alloy 277– modeling 219macrostructure – Al-Cu alloy 283– tin-bronze 256magnesium– chill casting 95– high strength alloy 182

381

– roll casting 143– strip casting 77magnesium alloys 70– numerical simulation 202– strip casting 81melt flow, Al-Cu alloy 283melt treatment – aluminium 3– copper 3melting temperature , multi-component al-

loys 174metal wire 213metals – binary alloys 296– molten 42– non ferrous 112micro wrought shapes, non ferrous metal

112microporosity, aluminum casting 36microstructure – development in aluminum alloys 290– eutectic alloys 309– tin-bronze 256– titanium alloys 265modelling, continuous casting processes

151, 213mold – cooled 137– design 106– temperature fields 234monitoring system, optical coil protection

353monte carlo simulation, grain growth 168multi-component alloys, conductivity 174

non ferrous metal, continuous casting 112numerical simulation, alloys 194, 202,

296

OCP, see optical coil protectionoptical coil protection 353

plastic strains, eutectic alloys 309plasticity, titanium alloys 265powder metallurgy, superalloys 249process conditions, hot top mould 106

process parameters, DC casting 271pulse electric discharging 296

quality of water 118quenching studies, alloys 118, 290

refining, grains 277rings, superalloy 249rod fabrication, copper 137rods, modelling 213roll casting 131, – AA6016 87– high speed 143– horizontal 70– twin 77

segregation, minimization 256semi-continuous casting 344shapes, micro wrought 112silicone, high strength alloy 182simulations – alloys 182, 194, 296– continuous casting 151, 202, 209, 213,

226– dendrite growth 162– grain growth 168slit mold 124Sn-bronze, continuous casting 219solidification – aluminum alloys 290– simulation 226, 240– binary metal alloys 296– unidirectional structure 137spray forming – aluminum alloys 101– superalloy rings 249– tin-bronze 256steel, billet 124strain distribution, eutectic alloys 309strip casting – alloy 77– aluminium alloy 131– continuous 70– magnesium 81superalloy rings, post processing 249

382

surface – defects 23, 29– roughness 118

temperature fields, mould 234ternary alloys, conductivity 174ternary tin bronze 320thermo-mechanical model, casting 189tin bronzes– dendrite coarsening 314– structure 256tin containing alloys, casting 320

titanium alloys, crystallization conditions 265

twin roll caster, unequal diameter 131

UDC, see upward direct chillunidirectional solidification structure 137upward casting– continous 202– direct chill 95

water, quality 118wincast-conti 209