Improving directional solidification process using Bayesian optimization method

Ronghui Ma, Department of Mechanical Engineering

In a directional solidification process, a planar solidification front can reduce radial segregation and suppress multiplication of dislocation, and thus help to achieve solidified materials with uniform properties and low defect density. The shape of the solidification front is largely determined by the temperature distribution in the melt, which is dictated by cooling at the bottom and heating on the lateral wall of the container. The lateral heating decays with time to sustain a continuous solidification, and the profile of the lateral heating power as a function of time needs to be optimized to achieve a planar interface throughout the directional solidification process. In this study, we use Bayesian optimization to identify the optimal heating power variation with time.