Systematic Benchmarking of Macrosegregation: The Performance of a Modified Hybrid Model

TL;DR

This paper develops an analytical solution for alloy solidification and uses it to verify and improve a macrosegregation solver, demonstrating that a hybrid model better predicts segregation features in benchmark experiments.

Contribution

It introduces a new analytical solution for the Stefan problem with microsegregation and applies it to enhance macrosegregation modeling accuracy in benchmark scenarios.

Findings

Hybrid model predicts segregation morphology more accurately.

Improved estimation of melt flow and re-melting effects.

Better agreement with experimental segregation maps.

Abstract

Recently, a new alloy solidification benchmark, called AFRODITE, with well-defined setups and state-of-the-art measurements has emerged, enabling a thorough assessment of MacroSegregation (MS) solvers, particularly in terms of their ability to predict different features of MS maps. In this research, we first develop an analytical solution for the alloy-solidification Stefan problem, which involves melt, solid, and mushy regions. This new analytical solution extends a previous solution (S. Cho and J. Sunderland, "Heat-conduction problems with melting or freezing", J. Heat Transfer, vol. 91, pp. 421-426, 1969) by incorporating a linear microsegregation law as a function of temperature in place of spatial coordinate. Then, we adopt this solution to verify an OpenFOAM MS solver in a limiting condition, where only heat diffusion is present. Subsequently, to capture the MS map of the Sn-3%Pb AFRODITE benchmark, the solver is incorporated using the standard Blake-Kozeny-Carman permeability law and one of its hybrid variants, slightly modified in this work to better align with physics by ensuring a continuous transition of characteristics from the slurry to the porous regions of the mush. It is demonstrated that the hybrid model predicts the main features of the MS map, including the channel segregates morphology and peak segregation degree due to the pile-up effect, in much finer agreement with the experimental observation. Careful analyses of the results reveal that these improved predictions stem from the hybrid model's more accurate estimation of the re-melting, melt flow advection parallel, and advection normal to the solidification front.