Accelerating High-Entropy Alloy Design via Machine Learning: Predicting Yield Strength from Composition

High-entropy alloys (HEAs) have attracted significant attention due to their exceptional physical, chemical, and mechanical properties. The current development of HEAs primarily depends on time-consuming and costly trial-and-error approaches, which not only hinder the efficient exploration of new compositions but also result in unnecessary resource and energy consumption, thereby negatively affecting sustainable development and production. To address this challenge, this study introduces a machine learning-based methodology for predicting the yield strengths of various HEA compositions. The model was trained using 181 data points and achieved an R² performance score of 0.85. To further assess its reliability and generalization capability, the model was validated using external data not included in the collected dataset. The validation was performed across four categories: modified Cantor alloys, refractory HEAs, eutectic HEAs, and other HEAs. The predicted yield strength trends were found to align with the actual experimental trends, demonstrating the model's robust performance across various categories of HEAs. The proposed machine learning approach is expected to facilitate the combinatorial design of HEAs, thereby enabling efficient optimization of compositions and accelerating the development of novel alloys. Moreover, it has the potential to serve as a guideline for sustainable alloy design and environmentally conscious production in future HEA development.