Composition, heat treatment, microstructure and loading condition based machine learning prediction of creep life of superalloys

Creep life is a key property of superalloys that are typically used in advanced engine turbine. The creep life of superalloys is mainly determined by factors including compositions, heat treatment processes, microstructures and loading conditions. Nevertheless, it still remains a big challenge to link these factors and creep life, due to the amount of variables and complex relations regarding the factors affecting creep life. In the present work, we solve this issue by a machine learning method. The dimension of the factors affecting creep life is reduced by principle component analysis, followed by clustering of the principle components. Then a proper regression method is chosen for each cluster such that an optimal model is formed for each cluster. The results show that the predicted creep lives agree with experimental creep lives well. New combinations of composition, heat treatment, microstructure and loading condition with better creep lives are proposed for the development of superalloys. Additionally, the present machine learning method is compared with existing machine learning methods for creep of superalloys. The comparison shows that the accuracy and efficiency of the present machine learning method are both considerably improved. Hence, the present method is useful for effective development of superalloys.