Gaussian process regression-based Bayesian optimization of the insulation-coating process for Fe-Si alloy sheets

High-efficiency Fe-Si alloy sheets have recently gained increasing attention in the automobile industry, and these sheets must be coated with insulation to reduce energy loss. However, it is difficult to maintain the coating without peeling and to realize high electrical insulation in the high-temperature heat treatment process during coating. In this study, using an artificial intelligence algorithm - Gaussian process regression (GPR)-assisted Bayesian optimization (BO) - we successfully developed a zirconia-based coating material for Fe-Si alloy sheets, yielding high heat resistance and high-quality surface properties. The coating material developed through the optimized process exhibits a high-quality silvery-white surface, the absence of coating damage even after heat treatment at temperatures exceeding 1100 K, and a surface current value of 600 mA or less, which is a measure of insulation. Notably, compared to the existing trial-and-error method, the number of experiments required to simultaneously achieve the target characteristics was reduced to less than 0.1% using the GPR-assisted BO, demonstrating the feasibility of the approach. This result also validates the efficiency and effectiveness of the proposed method in achieving multidimensional nonlinear optimization in the actual mass production of steel.