Mapping the creep life of nickel-based SX superalloys in a large compositional space by a two-model linkage machine learning method

Abstract

Accurate prediction of the creep life is important during the alloy design and optimization of nickel-based single crystal superalloys, especially for those with expensive alloying elements such as Re and Ru. In this study, a two-model linkage method was developed to predict the creep life of nickel-based single crystal superalloys through a data-driven machine learning approach, based on a small dataset collected from literatures. Owing to the small data amount, the precision of the model which was constructed with the alloy composition, creep temperature, and creep stress as inputs and used to predict the creep life is low, but the precision of the model which was constructed with the alloy composition, creep temperature, and creep life as inputs and used to predict the creep stress is relatively high. The high-precision model was applied to judge and adjust the prediction results of the low-precision model, thus effectively improving the accuracy of the overall creep life prediction. Moreover, the well-trained model-linkage was used to analyze, as a showcase, the interactive effects of alloying elements on the creep life of nickel-based single crystal superalloys. The prediction results provided a clear mapping of synergistic effects in the investigated compositional space, which was very consistent with previous, experimental results. This consistency thus confirms the effectiveness of the two-model linkage method in studying synergistic effects of multiple alloying elements. This study will benefit the compositional design and optimization of next-generation nickel-based single crystal superalloys.