Prediction and customized design of Curie temperature of Fe-based amorphous alloys based on interpretable machine learning

Chengcheng Liu,Yongchao Lu,Jianfa Feng,Weidong Cai,Hang Su
DOI: https://doi.org/10.1016/j.mtcomm.2023.107667
IF: 3.8
2023-12-03
Materials Today Communications
Abstract:Predicting the Curie temperature (T c ) is a crucial problem in the field of amorphous alloys. In this study, Fe-based amorphous alloys are taken as an example, and the composition and corresponding T c are collected through a literature review. Three feature construction strategies are employed to establish the relationship between the composition and T c using machine learning. The research findings demonstrate that the combination of the Meredig rule and the GBT algorithm yields the highest prediction accuracy. The features are constructed using recursive elimination and enumeration methods, ultimately resulting in an optimal 8-dimensional feature subset. Furthermore, the Shapley Additive exPlanations (SHAP) values are introduced to analyze the interpretability of the prediction model, providing a feature importance ranking and their critical values. Finally, by replacing the Fe and P atoms with Mn and Si atoms, respectively, in the Fe80P13C7 alloy, Fe62Mn18P(13-x)C7Six (x = 4,5,6) alloys are successfully designed to exhibit a Tc close to room temperature (335 K), enabling customized Tc design for Fe-based amorphous alloys.
materials science, multidisciplinary
What problem does this paper attempt to address?