Integrated response surface and machine learning framework for tribological optimization of LM24-based hybrid composites

This scientific research is focused on the dry sliding wear characteristics of an LM24 aluminum alloy-based hybrid composite reinforced by 2.5 wt% fly ash and 7.5 wt% ZrB 2 fabricated by the liquid stir casting technique. Quantitative microstructural characterization confirms the homogeneous dispersion of reinforcing particles, significant grain refinement (average grain size reduced by 46.7% to 43.7 μm), and minimal defect formation with a measured porosity of only 1.97%. Relative to the unmodified LM24 alloy, the performance of the hybrid composite is higher with a 7.14% increase in density (2.881 g/cc), a 47.67% improvement in hardness (127 ± 2.3 BHN), a 12.43% improvement in ultimate tensile strength (175.531 ± 3.46 N/mm 2 ), and a 34.09% increase in yield strength (118 ± 2.31 N/mm 2 ), albeit a 10.14% decrease in elongation (3.19%). The effects of applied load, sliding distance, and sliding speed on Specific Wear Rate (SWR) and CoF were studied with the help of Response Surface Methodology (RSM), and statistically robust predictive models were obtained for both response variables. To corroborate the statistical trends identified by RSM within the tested experimental domain, a comparative Machine Learning (ML) framework was deployed as a robust algorithmic check to verify the consistency of the identified parameter hierarchy. Multi-objective optimization by the desirability function resulted in optimal operating conditions of 10.011 N applied force, 589.578 m sliding distance, and 1.089 m/s sliding speed. Predictions from the trained ML models at these optimized parameters were then compared against the RSM results to test the robustness of the identified parameter hierarchy, serving as a comparative cross-validation.