Abstract
Abstract Although LM13-based hybrid metal matrix composites reinforced with Zircon and carbon have been widely studied for tribological applications, the combined influence of chill-induced microstructural refinement and sliding speed on wear behaviour and predictive reliability remains insufficiently understood. This study investigates the effect of varying chill levels on wear resistance, wear–speed sensitivity, and the accuracy of statistical and machine-learning models for LM13/Zircon/Carbon hybrid composites. LM13 specimens were fabricated by copper chill casting with chill levels of 12%, 9%, 6%, 3%, and as-cast condition, and were subjected to dry sliding wear tests at different rotational speeds. The experimental data were analysed using ANOVA to determine the significance of sliding speed and chill level, followed by linear regression, polynomial regression, Random Forest, and Support Vector Regression for predictive modelling. The results showed a consistent increase in wear rate with increasing RPM across all chill conditions. Among the models, polynomial regression exhibited the highest predictive accuracy (R2 = 0.9524–1.000), outperforming linear regression and the machine-learning approaches, while low RMSE (0.000019–0.000047 g) and MAE (0.000014–0.000044 g) values confirmed good dataset stability. Shapiro–Wilk tests (p > 0.314) and QQ-plot analysis verified residual normality and model reliability. Microstructural observations revealed that higher chill levels produced finer dendritic spacing, lower porosity, and more uniform reinforcement distribution, resulting in reduced wear and lower speed sensitivity. Overall, chill-controlled solidification significantly improved wear resistance and stabilized wear progression, providing a reliable framework for optimizing LM13-based hybrid composites for tribological applications.