Characterization of the flow in the molten metal sump during direct chill aluminum casting
- PDF / 386,292 Bytes
- 9 Pages / 612 x 792 pts (letter) Page_size
- 14 Downloads / 268 Views
I.
INTRODUCTION
IN direct chill (DC) aluminum casting, convective fluid flow and heat transfer in the melt pool dominate conduction effects. It is clear that early, simple conduction models for the prediction of temperature profiles during the casting[1] will no longer suffice. Given this, it is surprising that the flow in the melt pool (or sump) is still poorly understood from an analytical, or scaling, point of view. The flow field would be expected to have a crucial impact on the growth and transport of dendrites, and hence on the level of macrosegregation in the final ingot. It has been recently proposed that macrosegregation occurs through two possible mechanisms. First, growing dendrite fragments are transported by convection from the edge to the center of the ingot, where they sediment out of the solution. This causes macrosegregation, because the dendrite fragments, having nucleated at, and been transported from, the ingot edge, will not be at the liquid bulk composition.[2] Second, floating grains may grow close to the liquidus for long times at temperatures greater than that of the solid mush. The surrounding fluid is replenished by convection and hence stays close to the bulk composition. The composition of these floating grains will therefore be further from the bulk composition than that of the mushyzone solid formed in situ.[3] Both macrosegregation models depend crucially on the temperature and flow fields in the sump, so knowledge of the variance of these properties throughout the sump, and for casting-parameter ranges of interest, could be an important determinant of the pattern and level of macrosegregation. In addition, an analytical model would be useful as a guide and check for computational simulations, since (a) a mathematical model will indicate areas of the flow field with large gradients, and hence where fine numerical meshing is appropriate; (b) the predicted structure of the flow field can be compared with output from the simulations to JASON M. REESE, Lecturer, is with the Department of Engineering, University of Aberdeen, Aberdeen AB24 3UE, United Kingdom. Manuscript submitted May 21, 1996. METALLURGICAL AND MATERIALS TRANSACTIONS B
check that macroflow features are being reproduced, at least qualitatively; and (c) an analytical model can predict characteristic magnitudes of flow features, which can then identify casting-parameter regions to be fruitfully explored by numerical simulation, thus reducing the time spent in serial calculations over the parameter space. Present examinations of the continuous casting of metals rely, almost exclusively, on computational fluid dynamics (CFD), simulations of the flow under predetermined operating conditions, combined with some form of solidification model for when the aluminum drops below its solidus temperature. However, an incomplete understanding of essential flow-field features may, in this case, lead to inaccurate simulations. For example, Li and Anyalebechi[4] developed a finite-element code to model continuous aluminum casting and inc
