UDK 621.715:621.74.047 ISSN 1580-2949 Izvirni znanstveni članek MTAEC9,38(5)249(2004) SIMULATION OF MACROSEGREGATION IN THE DC CASTING OF BINARY ALUMINUM ALLOYS SIMULACIJA MAKROIZCEJANJA PRI POLKONTINUIRNEM ULIVANJU BINARNIH ALUMINIJEVIH ZLITIN Miha Založnik1, Božidar Šarler2, Dominique Gobin3 1Impol,d. d.,Sektor razvoj,Partizanska 38,2310 Slovenska Bistrica,Slovenija 2Politehnika Nova Gorica,Laboratorij za večfazne procese,Vipavska 13,5000 Nova Gorica,Slovenija 3Laboratoire FAST,UMR CNRS 7608,Campus Universitaire,Bât. 502,91405 Orsay cedex,France miha.zaloznikŽp-ng.si Prejem rokopisa – received: 2004-04-22; sprejem za objavo - accepted for publication: 2004-09-13 A common defect that occurs in direct-chill (DC) casting is macrosegregation: an inhomogeneous distribution of alloy components across the cross-section of the solidified casting. This defect can lead to non-uniform mechanical properties,which affect the behavior of the metal during downstream forming and heat treatments. The main mechanism behind macrosegregation is the transport of segregated alloying elements at the scale of the casting by the relative movement of liquid and equiaxed solid grains in the mushy zone,induced as well by pouring as by thermal and solutal natural convection. These macroscopic phenomena are physically quite well understood,yet very difficult to quantitatively model. In this work,first results obtained from an axisymmetric computational model of the DC casting of aluminum alloy billets are presented. The physical model of the solute transport is based on the one-phase continuum mixture model of dendritic solidification,supplemented by constitutive models of microsegregation. It takes into account solute transport by diffusion and advection. A rigid solid phase is assumed. The coupled mass,momentum,energy and solute conservation equations and microsegregation models are solved using the finite-volume method (FVM). The results of simulations of macrosegregation in an Al-4,5 % mass fraction of Cu billet are presented and explained. The model’s deficiencies related to numerical discretization errors and physical modeling are recognized and identified as issues for further research. Keywords: solidification; direct-chill casting; macrosegregation; modeling Makroizcejanje,neenakomerna porazdelitev legirnih elementov po ulitku je pogost defekt pri polkontinuirnem ulivanju. Povzroči lahko neenakomerne mehanske lastnosti materiala,ki poslabšajo njegove karakteristike med nadaljnjimi postopki obdelave. Makroizcejanje je posledica transporta legirnih elementov na merilu ulitka zaradi relativnega gibanja kapljevinaste faze in enakoosnih kristalnih zrn v trdno kapljevinastem medfaznem območju. Do gibanja faz pride zaradi naravne konvekcije zaradi temperaturnih in koncentracijskih gradientov ter zaradi prisilnega toka pri vtoku taline. Poznanje posameznih makroskopskih pojavov je pri tem dokaj dobro,vendar jih je zelo težko kvantitativno modelirati. V tem delu so predstavljeni prvi rezultati numeričnega modela makroizcejanja pri polkontinuirnem ulivanju. Fizikalni model transporta sestavin temelji na kontinuumskem modelu enofazne mešanice za dendritsko strjevanje. Dopolnjujejo ga konstitutivni modeli za opis mikroizcejanja. Model upošteva transport sestavin z difuzijo in advekcijo zaradi toka kapljevine,pri čemer je v kašastem medfaznem območju predpostavljena toga trdna faza. Sklopljene enačbe ohranitve mase,gibalne količine,energije in sestavin ter mikroizcejni modeli se rešujejo z metodo kontrolnih prostornin. Predstavljeni in pojasnjeni so rezultati simulacij makroizcejanja v drogu iz binarne zlitine Al-4,5 % Cu. Ugotovljene so nekatere pomanjkljivosti modela zaradi numeričnih diskretizacijskih napak in nepopolnosti fizikalnega modela. Ti področji sta razpoznani kot relevantni področji za nadaljnje raziskave. Ključne besede: strjevanje; polkontinuirno ulivanje; makroizcejanje; modeliranje 1 INTRODUCTION Solidification is one of the key phenomena in the processing of metals. Most metals used for technical applications are alloys,i.e.,mixtures composed of several chemical components. The solidification of such mixtures differs in many respects from that of pure substances. Phase change takes place over a range of temperatures,thus solid and liquid can coexist in equilibrium at various temperatures. Moreover,due to different solubilities in the solid and liquid phases the components can be rejected from or preferentially included into the forming phase (segregation). The solid-liquid interface is usually not smooth and develops into a complex microscopic growth structure,macro-scopically regarded as a mixture zone of coexisting solid and liquid phases,called the mushy zone. The processes of heat,mass and species transfer in the mushy zone are closely interlinked. One of the most widely used industrial metal processing technologies is direct-chill (DC) casting. During casting many defects can occur that are a direct consequence of the transport phenomena taking place in the process. A common defect that occurs in the direct-chill casting of aluminum alloys is macrosegregation,an inhomogeneous distribution of alloy components in the solidified casting. It is caused by solute transport,primarily due to the flow of segregated liquid in the mushy zone,which is a result of • buoyancy forces due to thermal gradients (thermal natural convection), • buoyancy forces due to concentration gradients (solutal natural convection), MATERIALI IN TEHNOLOGIJE 38 (2004) 5 249 M. ZALOŽNIK ET AL.: SIMULATION OF MACROSEGREGATION IN THE DC CASTING • density differences between the two phases (solidification shrinkage), • inlet flow (bulk convection). Macrosegregation can lead to non-uniform mechanical properties that affect the behavior of the metal during subsequent treatments. It is therefore desirable to be able to simulate the casting process in order to predict the influence of casting parameters on the resulting macrosegregation pattern. Besides prediction,modeling is aimed at an improved understanding of the basic mechanisms involved. A commonly observed solute distribution pattern in a DC cast billet shows a one-dimensional radial concentration distribution. A solute-depleted region is present in the billet center,adjoined by a positive segregation zone spreading into the outward radial direction,an adjacent thin negative segregation zone and another positive segregation layer at the surface. The enriched subsurface layer is attributed to solidification-shrinkage induced flow,while the other patterns are a consequence of an interplay of thermal convection,solutal convection and shrinkage flows. The exact mechanism is not yet completely understood. Recently,the effect of advective species transport by low-concentration free-floating crystal grains has been much discussed as an important factor 1234,especially in the formation of the negative centerline segregation zone. The first numerical-model study of macrosegregation in DC casting was made by Flood,Katgerman and Voller 5. Their model assumed a rigid mushy zone moving at the casting speed and failed to predict macrosegregation properly. Reddy and Beckermann 6 presented a computational investigation of the influence of grain refining on macrosegregation,where they simulated grain refining via an increased mushy-zone permeability. Good explanations for the computed concentration profiles could not be provided. Generally, models predict positive subsurface segregation,which is attributed to shrinkage flow. Some models also include the effect of exudation on surface macrosegregation, modeled successfully and in detail by Thevik,Mo et al. 7. A few years ago Vreeman and Incropera 8 revealed a problem pertaining to upwind discretization of the species-transport equation,especially in the simulation of DC casting. They showed that the previously used inconsistent discretization leads to severe errors in the concentration field solution. Subsequent studies thus use the corrected discretization method. Even though the accuracy of the numerical solutions of metal melt flow and species transport during solidification has been addressed in the literature it has mostly been clearly insufficient in simulations of large-scale castings (e.g.,in 9). It thus remains an important issue in further research 10. Recently,the problem of numerical diffusion in the simulation of macrosegregation in DC casting was addressed by Venneker and Katgerman 11,who performed computations using several finite-volume discretization schemes on different meshes. They did not explicitly consider the flow field; however,they did show that the use of poorly performing discretization schemes induces considerable numerical errors,affecting the predicted macrosegregation profile. 2 DESCRIPTION OF THE MODEL 2.1 Macroscale transport model The macroscopic conservation equations are transport equations for heat,mass,momentum and species. They are formulated as a one-phase continuum mixture model 12,where the mixture quantities in a two-phase (solid-liquid) mixture are defined in terms of phase quantities and phase fractions as follows. fi m g V 1m j giIV j Pm = gsPs + g 1P1 vm = fsvs + f1 v1 h m = fs h s + f1 h 1 Cm = fsCs + f 1 C1 Continuity equation for the mixture. dt ¦ + V-(pmvm) = 0 (1) (2) (3) Mixture momentum conservation equation. dt + V-(pmvmvm) = -Vp + V-|/a1 m-Vvm Ě A <"1 Pm ¦(vm -vs) (4) K a -p1 g rŠßT (T-T0) + ßC(Cm-C0) -V-Špm(f svsvs +f1v1v1 -vmvm The velocity of the solid is defined to be vs =vcast everywhere,since all the solid mush is assumed to be coalesced in a porous matrix and connected with the bulk solid. The permeability is modeled by the Kozeny-Carman relation K= K0 g 13/(1–g1)2. The mixture energy conservation equation is: d(pm h m ) dt -+V-(pmvmhm) = V-(kVT) (5) -V-Špm(f svshs+f1v1h1 -vmhm The diffusive term (Fourier law) is reformulated in terms of the mixture enthalpy using the supplementary thermodynamic state equations. T hs(T) = href + jcpsdT (6) T h1(T) = hs(T) + Leut + J(cp1 -cps)dT (7) T« The final form of the equation is obtained. 250 MATERIALI IN TEHNOLOGIJE 38 (2004) 5 M. ZALOŽNIK ET AL.: SIMULATION OF MACROSEGREGATION IN THE DC CASTING d(Pm hm ) dt + V-(pmvmhm) vi k Vh + V k V(hs-hm) (8) c ps (9) -V-lPm(fsvsh s+f 1 v r1h1 -vmhm)] Mixture species conservation equation. ČPmCm) + V-(pmvmCm) = V-(pmDmVCm) +V-Špm(f sDsVCs+f1D1VC1 -DmWCm)] The phase quantities (fs, f l, h s, h l, Cs, Cl) that still appear in the mixture transport equations (3),(4),(8), and (9) are modeled by supplementary thermodynamic models,defining phase enthalpies (Equations (6),(7)) and models of microscale transport phenomena, describing the relations between the local phase concentrations. These models are unified in a so-called microsegregation model,which is described in the following section. 2.2 Microsegregation model Modeling of the transport of species at the microscale is usually based on the following simplifying assumptions 13. • One-dimensional model geometry,defined by a control volume between the periodic secondary dendrite arms. • Perfect mixing in the liquid phase (C l = C l*). • Thermal equilibrium (uniform temperature) at the microscale. • Thermodynamic equilibrium at the phase interface (T = TS( C l), Cs* = kpC l). The microsegregation models are defined by a system of three equations. 1. Definition of mixture enthalpy following from Equations (2),(6),and (7). r hm(T,f1)=jcpsdT + f1 Leut + J( T ref T + (c T cps )dT 2. Linearized liquidus line equation. T(C1) = Tf + mLC1 (10) (11) 3. Microscale species transport model,which defines the link between the mixture concentration Cm,the liquid concentration Cl and the liquid fraction fl. In a general form it can be written as C1(Cm,f1) (12) Examples of microscale transport models are the inverse lever rule C1 = C kp+f 1(1-kp) (13) or the Scheil model (in the differential,open-system formulation) f 1dC1+( k-1)C1df1 = dCm (14) By the substitution of variables in Equations (10)-(12) it is possible to analytically obtain a function F(f1) = 0 (15) The root of the function is found using Newton's method,giving fl. Then the temperature T can be calculated from Equation (10) and the liquid concentration C l from Equation (11). The solution of the microsegregation model thus gives the temperature T, the liquid fractionf l and the liquid concentration C l from the mixture enthalpy hm and the concentration Cm,which are obtained from the solution of the macroscopic transport model. The remaining phase quantities needed in the macroscopic model (fs, hs, h, C s) are then calculated using Equations (1),(2),(6),and (7). 2.3 Solution procedure The set of macroscopic transport equations (Equations (3),(4),(8),and (9)) was solved with the finite-volume method (FVM) on an axisymmetric geometrical model. The computer code by Mencinger 14 for the solution of thermal and flow fields in DC casting, which was extended by the macroscopic species transport and microsegregation models,was used. The top boundary at the inflow was assumed to be isothermal at a prescribed casting temperature,the inflow velocity was fixed at vin = (ps/pl)vcast. The casting velocity was vcast = 10-3 m/s. The bottom boundary was assumed to be adiabatic. Although the latter assumption is not completely valid because of the short domain length,it should not significantly affect the basic mechanisms of macrosegregation. The left boundary (centerline) is adiabatic because of axisymmetry. The slip boundary conditions for velocity and chill boundary conditions describing the heat transfer to the cooling water 1115 are applied at the right boundary. They are summarized in Table 1. Table 2 summarizes the material properties as already used in 1115. The SIMPLE algorithm 16 was used for the pressure-velocity coupling. The upwind-difference scheme was used for the discretization of the advective fluxes in all equations. Because of the high thermal diffusivity of aluminum alloys this is believed to cause only minor numerical diffusion in heat transfer. For momentum and species transport the upwind scheme was used to prevent stability problems with higher order schemes,which would be unstable for the mesh Péclet numbers (PeČ) and mesh Reynolds numbers (ReAx) encountered in the advection-diffusion transport of species and momentum respectively. Pe vAx D 4-103 (16) ps MATERIALI IN TEHNOLOGIJE 38 (2004) 5 251 M. ZALOŽNIK ET AL.: SIMULATION OF MACROSEGREGATION IN THE DC CASTING Re pvAx /u 4101 (17) The axisymmetric geometrical model was discretized with a nonuniform mesh. The mesh distance was kept uniform in the upper part of the computational domain, containing the liquid and mushy regions,and was gradually sparsened in the z direction in the lower solid part of the domain. The domain has to be long enough in the z direction to ensure proper solution of the temperature field,since heat is conducted through the solid part of the billet. The domain length was 1.0 m in the case without species transport (Figure 1) and 0.8 m in the full macrosegregation case (Figures 2-5). Table 1: Boundary conditions for mold and direct-chill zones 11,15 Preglednica 1: Robni pogoji za področji kokile in neposrednega hlajenja s filmom hladila 11,15 Position /m hchill / (W/m2K) -0.00 >z> -0.12 0 -0.12 >z> -0.18 150 gs+ 1500 g1 -0.18 >z> -0.19 150 + zŠm19 00..18 (20000 150) -0.19 >z> -0.23 20000 + z0m]+00 1199 (10000 20000) -0.23 > z 10000 3 RESULTS AND DISCUSSION Before proceeding to the full macrosegregation problem,the partial problem of coupled heat transfer and fluid flow was solved to approximately characterize the quality of the velocity-field solution. Since advection by liquid flow is the main mechanism of species transfer,a correct flow solution is important for a correct macrosegregation simulation. Because of the awareness of the presence of numerical errors and slow mesh convergence inherent of the first order scheme,which was used to ensure stability,a mesh convergence study was made. Some of the results are shown in Figure 1, where profiles of the vertical velocity component along the billet centerline are shown,as obtained in calculations on four different computational meshes, with the mesh spacing gradually decreasing by a factor of 2. It was found that even the solution on the densest mesh was not sufficiently close to the asymptotic convergence range,where convergence could be well characterized. However,due to a prohibitive increase in the computation time with mesh refinement,the mesh was not refined further. The simulation of macrosegregation was made on a domain with reduced length L = 0.80 m,using a 136 × 252 finite-volume mesh. The mixture concentration field in the billet is shown in Figure 4. Figure 3 shows the corresponding velocity and liquid-fraction fields. In the case without species transport only thermal natural convection is present,and it causes a clockwise Table 2: Thermo-physical properties of Al-4.5%Cu 11,15 Preglednica 2: Snovne lastnosti zlitine Al-4.5%Cu 11,15 Solid density ps kg/m3 2750 Liquid density P1 kg/m3 2460 Solid specific heat cps J/kgK 958 Liquid specific heat cps J/kgK 1054 Solid thermal cond. k s W/mK 180 Liquid thermal cond. k 1 W/mK 95 Solid diffusivity D s m2/s 5 · 10–12 Liquid diffusivity D s1 m2/s 5 · 10–9 Latent heat at eutectic Leut J/kg 3.90·105 Viscosity µ Pas 1.3 10-3 Thermal expansion c. ßT K–1 1.17·10–4 Solutal expansion c. ßC -0.73 Reference temperature T0 K 950 Reference concentration C0 - 0.045 Reference density P0 kg/m3 2460 Permeability constant K0 - 6.67·1011 circulation in the liquid sump. The flow velocity is large enough to cause a net upward flow (in terms of absolute velocity) in the billet center,while part of the flow deflects downwards in the center due to inertia. In the full problem (Figures 2-5) the flow is modified due to the influence of solutal buoyancy. As can be seen in Figure 2,the liquid is relatively quiescent at the bottom and in the center of the liquid sump,which appears to be caused by the counteracting effect of thermal buoyancy and the layering tendency of the heavy high-concentration liquid. Next to the liquidus front a complex flow structure consisting of small vortices forms. Five -0.0020J- ¦ vŠm/«] mashl 27x77 _ _ vŠm/s] mesh2 52x152 ---------------vŠm/s] mesh4 102x302 ---------------vŠm/s] meihB 202x602 -0.2 z/m Figure 1: Vertical velocity component (v) profiles along the billet centerline obtained in the computations without species transport Slika 1: Profili navpične komponente hitrosti (v) vzdolž središčnice droga,dobljeni pri izračunih brez upoštevanja transporta sestavin 252 MATERIALI IN TEHNOLOGIJE 38 (2004) 5 M. ZALOŽNIK ET AL.: SIMULATION OF MACROSEGREGATION IN THE DC CASTING . r/m Figure 2: Liquid density (?l) and absolute velocity (v) fields in the billet obtained in the full macrosegregation computation. The density field is shown only in the liquid and mushy zones (fl > 0). Slika 2: Polji gostote kapljevine (?l) in absolutne hitrosti (v) v drogu, dobljeni pri izračunu makroizcejanja. Polje gostote je prikazano le v kapljevinastem in kašastem področju (fl > 0). vertically arranged flow cells can be identified,delimited by large density gradients. Each cell contains one or two vortices of thermosolutal origin. The whole flow -0.1 "°-2 -0.3 -0.4 W=mm ¦il J llllül 'Wi ' * Hiß 'JMĘĘĘrf:- ¦¦'¦¦¦'¦ f, u 1.0 10 0.9 9 0.8 8 0.7 7 0.6 6 0.5 5 0.4 4 0.3 3 0.2 2 0.1 1 0.0 0.00 0.05 0.10 r/m 0.15 0.20 Figure 3: Relative velocity field (v-vcast) and liquid fraction isopleths in the billet obtained in the full macrosegregation computation Slika 3: Polje relativne hitrosti (v-vcast) in linije konstantnega deleža kapljevinaste faze v drogu,dobljeno pri izračunu makroizcejanja r/m Figure 4: Mixture Cu concentration (Cm) field in the billet Slika 4: Polje koncentracije Cu mešanice (Cm) v drogu structure close to the liquidus front seems to be very complex and further studies will be necessary to determine and explain it more accurately. In the solidification zone the liquid is enriched due to segregation of the solute (Cl > Cm; Cs < Cm). Since copper has a larger density than aluminum,a solutal downward flow of enriched liquid can be observed in the high-fl portion of the mushy zone (along the liquidus front). This flow carries enriched liquid away from the top-subsurface part (top right in Figure 4) of the mushy zone,replacing it with lower-concentration liquid,which 0.10 r/m Figure 5: Radial Cu concentration profile in the solidified billet Slika 5: Radialni profil koncentracije Cu v strjenem drogu 0.20 MATERIALI IN TEHNOLOGIJE 38 (2004) 5 253 M. ZALOŽNIK ET AL.: SIMULATION OF MACROSEGREGATION IN THE DC CASTING penetrates into the mush from the bulk liquid region. The heavy,enriched liquid flows down the slope of the mushy zone front,accumulating at the bottom of the sump. As the circulation,driven by thermal convection, carries some high concentration liquid out of the mushy zone into the bulk-liquid sump,smaller circulations, driven by strong solutal buoyancy can be observed. They are characterized by a rapid deflection back downward due to the strong effect of increased density. In the low-fl regions of the mushy zone the permeability of the porous mush rapidly decreases. Therefore,the buoyancy and inertia do not influence the flow,and so the drag forces dominate. The flow is driven mostly by solidification shrinkage,as the liquid fills up the space left by the shrunken solidified metal. The liquid in the low-fl mushy region is highly segregated (phase diagram),thus even small relative velocities can result in considerable net solute transport. This can be clearly seen along the whole mushy zone. The direction of the shrinkage-induced flow is parallel to the liquid fraction gradient. As the enriched liquid flows towards the solidus front it is replaced by the lower concentration liquid from regions of the mushy zone with higher fl. The mushy zone thus appears as solute-depleted in terms of the mixture concentration. Flowing towards the solidus front,the enriched liquid accumulates at the front and the concentration rapidly increases in the direction normal to the fl gradient. This is the mechanism causing typical positive subsurface segregation. Also,due to the predominance of shrinkage flow in the densely packed mushy zone,the segregation regions can already be identified before solidification is complete. There is no solute transport along the solidification front anymore. The resulting macrosegregation profile,shown in Figure 5,does not properly predict the commonly observed distribution,characterized by a solute-depleted region in the billet center,adjoined by a positive segregation zone spreading in the outward radial direction,an adjacent thin negative segregation zone and another positive segregation layer at the surface. The reasons are believed to be deficiencies with the present model. They can be divided into numerical and physical modeling. Both significantly influence the macrosegre-gation prediction. The discrepancy in the centerline segregation is clearly a consequence of the simplified physical model,which lacks a description of the transport of low-concentration free-floating crystal grains,which are thought to sediment at the bottom of the sump and cause the common negative centerline segregation 1,4,15. Instead,the present model logically predicts a settling of heavy copper-rich liquid at the sump bottom,which results in positive centerline segregation. The numerical deficiencies are related to use of upwinding in cases with large Pe?x and Re?x (Equations (16),(17)),where it causes considerable diffusion-like numerical discretization errors. While the presented macrosegregation model is at the beginning of 254 its development,and was not expected to definitely explain the mechanism of macrosegregation in DC casting,the two separate problems that were identified, point out the relevant issues for further work. First,the numerical difficulties have to be elucidated and resolved. The verification can proceed through systematic convergence studies 17 and a comparison with solutions that use alternative methods 18. Only after a satisfactory solution of the model equations is achieved,the physical model has to be extended. Presently,the effect of solute transport by free-floating crystal grains seems to be the most important physical mechanism for further study. 4 CONCLUSIONS Simulations of macrosegregation in DC casting of a binary aluminum alloy were performed using a classical solidification model. An analysis of the computational results showed a rather complex interplay of predominantly advective species transport modes (double-diffusive convection and shrinkage flows) and has provided explanations for the predicted macrosegre-gation pattern. The predicted concentration profile is similar to those previously obtained in refs. 11,15. While numerical problems regarding the solution of species transport are known from the literature and were recognized in macrosegregation results presented here, difficulties in obtaining a high-quality flow-field solution were shown on a reduced model (disregarding transport of species). Numerical and physical modeling issues for further research were identified,which are to bring us closer to understanding and a quantitative prediction of the macrosegregation formation in DC casting. ACKNOWLEDGEMENTS Helpful discussions with Dr. Jure Mencinger, formerly with the Laboratory for Multiphase Processes, are gratefully acknowledged. The work was supported by Impol d.d. and additionally by the Slovenian Ministry of Education,Science and Sport (MŠZŠ) and the Ministry of the Economy through the Young Researchers program as well as by MŠZŠ through grant L2-5387-1540-03. The exchange between the French and Slovenian research groups was funded through the Proteus program (project SLO-FR 5/2003). NOMENCLATURE C mass concentration of copper C0 reference concentration in buoyancy term cp specific heat D mass diffusivity f phase mass fraction g phase volume fraction g gravitational acceleration MATERIALI IN TEHNOLOGIJE 38 (2004) 5 M. ZALOŽNIK ET AL.: SIMULATION OF MACROSEGREGATION IN THE DC CASTING h specific enthalpy hchill heat transfer coefficient at chill k thermal conductivity K permeability kp binary equilibrium partition ratio L computational domain length Leut latent heat of fusion at eutectic m mass mL liquidus line slope p pressure PeM mesh Peclet number r, z cylindrical coordinates ReČ mesh Reynolds number t time T temperature T0 reference temperature in buoyancy term Teut eutectic temperature v vertical velocity component v velocity V volume Ax mesh distance ßT volumetric thermal expansion coefficient ßC volumetric solutal expansion coefficient µ dynamic viscosity p density p0 reference density in buoyancy term l liquid mixture solid S solidus equilibrium conditions 5 REFERENCES 1 H. Yu,D. A. Granger,Macrosegregation in aluminum alloy ingot cast by the semicontinuous direct chill (DC) method,Aluminum Alloys: Their Physical and Mechanical Properties,EMAS,UK, 1986,17–29 2 T. L. Finn,M. G. Chu,W. D. Bennon: The influence of mushy region microstructure on macrosegregation in direct chill cast aluminum-copper round ingots,Micro/Macro Scale Phenomena in Solidification,ASME,New York,1992,17–24 3 C. J. Vreeman,M. J. M. Krane,F. P. Incropera,The effect of free-floating dendrites and convection on macrosegregation in direct chill cast aluminum alloys,part I: Model development,International Journal of Heat and Mass Transfer,43(2000)5,677–686 4 C. J. Vreeman,M. J. M. Krane,J. D. Schloz,Direct chill casting of aluminum alloys: Modeling and experiments on industrial scale ingots,Journal of Heat Transfer ASME,124(2002)10,947–953 5 S. C. Flood,L. Katgerman,V. R. Voller: The calculation of macro-segregation and heat and fluid flows in the DC casting of aluminum alloys,Modeling of Casting,Welding and Advanced Solidification Processes V,TMS,1991,683-690 6 A. V. Reddy,C. Beckermann,Modeling of macrosegregation due to thermosolutal convection and contraction driven flow in direct chill continuous casting of an Al-Cu round ingot,Metallurgical and Materials Transactions B,28B(1997)3,479–489 7 H. J. Thevik,A. Mo,T. Rusten,A mathematical model for surface segregation in aluminum direct chill casting,Metallurgical and Materials Transactions B,30B(1999)1,135–142 8 C. J. Vreeman,F. P. Incropera,Numerical discretization of species equation source terms in binary mixture models of solidification and their impact on macrosegregation in semicontinuous,direct chill casting systems,Numerical Heat Transfer B,36(1999)1,1–14 9 N. Ahmad,H. Combeau,J.-L. Desbiolles,T. Jalanti,G. Lesoult,J. Rappaz,M. Rappaz,C. Stomp,Numerical simulation of segregation: a comparison between finite volume method and finite element method predictions and a confrontation with experiments, Metallurgical and Materials Transactions A,29A(1998)2,617–630 10 C. Beckermann,Modeling of macrosegregation: Applications and future needs,International Materials Reviews,47(2002)5,243–261 11 B. C. H. Venneker,L. Katgerman,Modelling issues in macro-segregation predictions in direct chill castings,Journal of Light Metals,2(2002)3,149–159 12 W. D. Bennon,F. P. Incropera,A continuum model for momentum, heat and species transport in binary solid-liquid phase change systems. I. model formulation,International Journal of Heat and Mass Transfer,30(1987)10,2161–2170 13 H. Combeau,J.-M. Drezet,A. Mo,M. Rappaz,Modeling of micro-segregation in macrosegregation computations,Metallurgical and Materials Transactions A,27A(1996)8,2314–2327 14 J. Mencinger,Polkontinuirno ulivanje aluminijevih zlitin: izračun časovnega razvoja temperaturnega in hitrostnega polja (in slovenian; Direct chill casting of aluminum alloys: Calculation of time evolution of temperature and velocity fields),Technical report, Politehnika Nova Gorica,2002 15 C. J. Vreeman,F. P. Incropera,The effect of free-floating dendrites and convection on macrosegregation in direct chill cast aluminum alloys,part II: Predictions for Al-Cu and Al-Mg alloys,International Journal of Heat and Mass Transfer,43(2000)5,687–704 16 J. H. Ferziger,M. Peri},Computational Methods for Fluid Dyna-mics,Springer Verlag,2nd edition,1996 17 P. J. Roache,Verification and Validation in Computational Science and Engineering,Hermosa publishers,Albuquerque,1998 18 B. Šarler,J. Mencinger,Solution of temperature field in DC cast aluminium alloy billet by the dual reciprocity boundary element method,International Journal of Numerical Methods for Heat and Fluid Flow 9(1999)3,269–297 MATERIALI IN TEHNOLOGIJE 38 (2004) 5 255