Coupling high-throughput experiment and machine learning to optimize elemental composition in nickel-based superalloys

Establishing relationship of elemental composition and mechanical property is a tremendous amount of work in superalloys. Here, machine learning coupled with high-throughput experiment is adopted to construct “composition-hardness” model in nickel-based superalloys. The hardness estimated from experiment agrees well with the predicted value. Furthermore, optimal composition of high-hardness superalloys is accurately predicted by simulated annealing algorithm. Subsequently, optimal composition is validated by Thermo-Calc software, further demonstrating the effectiveness of current approach. Here, a design strategy combined with high-throughput experiment and machine learning is proposed, which may be believed for accelerating the design of advanced materials with excellent performance.