Predicting the oxidation kinetic rate and near-surface microstructural evolution of alumina-forming Ni-based single crystal superalloy based on machine learning

The near-surface degradation of γ`-strengthening phase caused by oxidation has a significant effect on the mechanical properties of thin-walled Ni-based single crystal (SX) superalloys for turbine blade applications. It is essential yet still challenging to link alloy composition, oxidation behavior and near-surface microstructure to guide the design of high-performance SX superalloy. In the present work, two key issues are breached from a data-driven perspective, which are the prediction of the oxidation rate constant ( K <sub>p</sub>) and the near-surface γ` evolution features respectively. Two modeling strategies are proposed to predict K _p from compositions. One is the machine learning (ML) model based on detailed feature engineering, the other is a novel model that describes the ln K p - U (alumina formation energy) relationship by several linear functions with different intercepts according to ML classified sub-regions. The generalization ability for predicting K _p is significantly boosted by an ensemble model obtained from the two strategies. Insights related to oxidation control mechanisms and the contributions of key descriptors are obtained from interpretable information. Further, the ML models of the relationship between the near-surface composition and the fraction, shape factor, and size of γ` are constructed based on the near-surface microstructure datasets. Dynamic simulations using DICTRA are utilized to link the K _p with the near-surface composition and to refine the Composition- K _p-Degradation Microstructure (C K M) prediction flow. This flow allows for the direct acquisition of diverse near-surface microstructural evolution features from simple alloy compositions with a verified remarkable generalization performance, which can be further applied to the design of Ni-based SX superalloys for thin-walled components.