Machine Learning and Explainable AI-Guided Design and Optimization of High-Entropy Alloys as Binder Phases for WC-Based Cemented Carbides

Tungsten carbide-based (WC-based) cemented carbides are widely recognized as high-performance tool materials. Traditionally, single metals such as cobalt (Co) or nickel (Ni) serve as the binder phase, providing toughness and structural integrity. Replacing this phase with high-entropy alloys (HEAs) offers a promising approach to enhancing mechanical properties and addressing sustainability challenges. However, the complex multi-element composition of HEAs complicates conventional experimental design, making it difficult to explore the vast compositional space efficiently. Traditional trial-and-error methods are time-consuming, resource-intensive, and often ineffective in identifying optimal compositions. In contrast, artificial intelligence (AI)-driven approaches enable rapid screening and optimization of alloy compositions, significantly improving predictive accuracy and interpretability. Feature selection techniques were employed to identify key alloying elements influencing hardness, toughness, and wear resistance. To enhance model interpretability, explainable artificial intelligence (XAI) techniques—SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME)—were applied to quantify the contributions of individual elements and uncover complex elemental interactions. Furthermore, a high-throughput machine learning (ML)–driven screening approach was implemented to optimize the binder phase composition, facilitating the discovery of HEAs with superior mechanical properties. Experimental validation demonstrated strong agreement between model predictions and measured performance, confirming the reliability of the ML framework. This study underscores the potential of integrating ML and XAI for data-driven materials design, providing a novel strategy for optimizing high-entropy cemented carbides.