Machine learning-based prediction and optimization of vickers hardness and ultimate tensile strength in eutectic Al–Si/kyanite/graphite hybrid metal matrix composites

Hybrid metal matrix composites (MMCs) based on eutectic Al–Si alloys are promising candidates for lightweight structural applications. This study combines experimental synthesis via stir casting with machine learning-based modeling to develop eutectic Al–Si alloy reinforced with kyanite (Al₂SiO₅) at varying content (0, 2, 4, and 6 wt%) and constant 2.5 wt% graphite. Microstructural analysis by optical microscopy and energy-dispersive spectroscopy confirmed uniform reinforcement distribution. Mechanical testing revealed the 6 wt% kyanite composite achieved ∼107 HV hardness (+14.23% improvement) and ∼194 MPa tensile strength (+17.47% improvement) over the base alloy. Five supervised machine learning models (linear regression, decision tree, random forest, support vector regression and artificial neural networks) were trained using four-fold leave-one-out cross-validation to predict mechanical properties from composition and processing parameters. Random forest regression achieved highest accuracy with R² = 0.82 (hardness, RMSE = 2.94 HV) and R² = 0.96 (ultimate tensile strength, RMSE = 1.95 MPa). Feature importance analysis showed that kyanite content is the primary governing factor, succeeded by stirring speed, consistent with metallurgical principles. The validated machine learning model facilitates swift virtual property prediction for untested compositions within the 0–6 wt% kyanite range under constant processing circumstances, illustrating the efficacy of data-driven design methodologies for expedited materials development.