Interpretable Machine Learning and Genetic Algorithm-Based Optimization of Ultimate Tensile Strength And Melting Completion In Alloys

Abstract In this study, we present a comprehensive machine learning-based approach for optimizing alloy compositions with the goal of simultaneously maximizing Ultimate Tensile Strength (UTS) and approaching a target Melting Completion temperature. Using a dataset comprising elemental compositions of various alloys and their corresponding mechanical properties, we developed predictive models based on the Random Forest Regressor algorithm. SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) were employed to interpret the feature contributions and determine the most influential elements on both UTS and melting behavior. The analysis revealed that elements such as Fe, Mn, Co, and Mo significantly contribute to optimal alloy performance, while elements like C and V play a critical role in enhancing UTS. A multi-objective optimization was conducted using a Genetic Algorithm (GA), yielding an optimal composition that achieved a predicted UTS of 1520.7 psi and a Melting Completion temperature of 1407.4°C, closely aligned with the target of 1460°C. Our approach demonstrates the potential of combining interpretable machine learning with evolutionary optimization to accelerate intelligent alloy design and discovery.