Surrogate model for convective flow inside electromagnetically levitated molten droplet using magnetohydrodynamic simulation and feature analysis

Containerless processing of molten metals using electromagnetic levitation techniques is an important approach for material research, and the induced convection inside an electromagnetically levitated molten metal droplet is both difficult to measure experimentally and a key control parameter. This work proposed a surrogate model based on parametric numerical experiments of magnetohydrodynamic simulations of the ISS-EML facility, characterizing and quantifying the convective flow due to electromagnetic positioning power as a function of melt thermophysical properties. Feature mapping methods integrated with feature selection is used to construct an optimized formula in explicit form for the data fitting which maintains both of physical interpretability and mathematical tractability. The trained surrogate model is then applied for selected industrial important metals and alloys as examples that show the generalization ability of the model and how it can be utilized to quantify the induced convection and characterize internal flow patterns.