Knowledge-enabled data-driven smart design advanced high-entropy alloys with attractive dynamic mechanical properties

High-entropy alloys (HEAs) demonstrate significant potential for applications in extreme environments, such as high temperatures, irradiation, and high-speed impacts. The huge compositional space and rigorous experimental conditions present substantial challenges for the rational design of HEAs. In this work, a knowledge-enabled data-driven machine learning (ML) material design strategy is utilized to construct feature spaces and to propose key physical parameters (KPPs) based on the underlying mechanisms and the parameters of empirical and phenomenological constitutive models for the prediction of yield stress and fracture strain of HEAs designed at high strain rates. These KPPs are determined by Pearson correlation, SHAP values, feature importance analysis, and exhaustive screening and predicted based on the dynamic properties of HEAs employing diverse ML models, among which the XGBoost model performing optimally. Combined with experimental validation, it is revealed that shear modulus (G), thermal conductivity, and strain rate are treated as the KPPs in determining the dynamic yield strength of HEAs, while Young's modulus, ratio of bulk modulus (B) to G, and strain rate are the KPPs in estimating the dynamic ultimate strain of HEAs. In the smart design of targeted dynamic properties of alloy at different strain rates, W-Ni-Fe-Co-Cr, Ti-Zr-Nb-V, Ti-V-Nb-Al, Ti-Zr-Nb-Ta-Al, W-Mo-Fe-Ni-B, and Ti-Zr-Ta systems have been investigated systematically. It is worth mentioning that the predicted strength and plasticity of the best compositions within W-Ni-Fe-Co-Cr system agree well with those experimental ones at strain rates of 3000 s-1 and 4000 s-1. Interpretability of models in dynamic domains can be achieved through the development of ML strategies in the routine of data, model, and knowledges in terms of KPPs, paving a path for the subsequent multi-objective optimizations to address the classical strength-plasticity trade-off, thus to enable the design and discovery of advanced HEAs with attractive dynamic mechanical properties.