Discovery of a Ni-based superalloy with low thermal expansion via machine learning

Abstract

Traditional Ni-based superalloys are widely used due to their excellent high-temperature performance; however, their high thermal expansion at elevated temperatures limits their further application in the modern aerospace industry. To discover Ni-based superalloys with low thermal expansion for high-temperature environments, such as 900 °C. In this study, we developed a Support Vector Regression (SVR) model with a high coefficient of determination (R2 = 0.84) to predict the coefficient of thermal expansion (CTE) in Ni-based superalloys. Ten samples were selected and produced from a pool of candidates generated via genetic algorithm and the SVR model, revealing that seven of them exhibited superior CTE performance. Notably, the discovered alloy with the lowest CTE (11.0 e-6 °C-1 from RT to 900 °C) presented a remarkable 16.7 % reduction compared to the best performance in existing database. The finally selected alloy demonstrated both low thermal expansion and superior mechanical properties compared to traditional superalloys with low thermal expansion.