Design of Ni-based turbine disc superalloys with improved yield strength using machine learning.

A machine learning (ML) process on composition optimization was performed to design Ni-based turbine disc superalloys with improved yield strength. Based on published data of polycrystalline Ni-based superalloys, the design process is finished through regression algorithm, feature rank method, and genetic algorithm, which is simple and high-efficient to optimize composition. The two designed alloys are assessed using the Calculation of Phase Diagram (CALPHAD), finding the microstructure of both alloys according with superalloys. Comparing with commercial Ni-based turbine disc superalloys, the designed alloys have trade-offs on mechanical and physical properties. The ML architecture and the assessment results are then discussed, indicating that the design ability of ML to automatically optimize is promising. This work is of much practical significance in reducing trial-and-error test to improve design efficiency for material design, providing an effective way to development novel Ni-based turbine disc superalloys. [ABSTRACT FROM AUTHOR]