Machine learning prediction of hardness in solid solution high entropy alloys

The mechanical properties of high entropy alloys (HEAs) are enhanced by solid solution strengthening (SSS) mechanism, which is of great importance for the design of HEAs. The compositional design space of HEAs is extensive, enabling the attainment of varying levels of hardness by combining different SSS elements and phases. Consequently, the selection of appropriate alloy compositions and their proportions to achieve the desired hardness has emerged as a present research challenge in the field. Against this backdrop, achieving accurate predictions of the hardness of HEAs has become a focal point of scholarly investigations. Machine learning (ML) adeptly captures the nonlinear relationships between alloy hardness and various characteristic features, and is widely regarded as a pivotal approach in the realm of hardness prediction for contemporary HEAs. A complete ML-based workflow for hardness prediction of HEAs is presented in this paper. Five ML models were trained using the selected feature descriptors. Eventually, a Random Forest model was obtained, which demonstrated a predictive correlation coefficient of 0.91 on the independent test dataset. Furthermore, the Shapley Additive exPlanations (SHAP) theory is employed to elucidate the influence of empirical parameters on the hardness of HEAs.