Tribological performance of Al5052/B₄C/Granite hybrid composites: machine learning prediction and multi-response optimization.

This study investigates the sustainable wear performance of Al5052/B₄C/granite hybrid composites produced by the stir casting process. Wear tests were executed on a pin on disc by varying load from 20 N to 50 N and sliding speed from 2 m/s to 3 m/s to evaluate the specific wear rate (SWR) and coefficient of friction (COF). From the results, the lowest specific wear rate of 0.201 × 10⁻³ mm³/Nm was obtained at 20 N load, 2 m/s sliding speed, and 4 wt% granite reinforcement. SEM analysis of worn surfaces revealed small cracks and minimal wear debris, indicating low fatigue wear and better resistance to surface degradation. ANN, ANFIS, and LightGBM models were also used to predict wear performance. LightGBM showed the highest accuracy, with R² values of approximately 0.9992 for SWR and 0.9987 for COF. ANOVA confirmed that sliding speed was the most significant factor affecting both responses. The Composite Desirability Function (CDF) was employed to identify the best parameter combination for reducing both SWR and COF. According to CDF results, the optimum conditions were 32.12 N load, 2.0 m/s sliding speed, and 6 wt% granite reinforcement. Under these settings, the predicted minimum COF was 0.4129 with desirability of 0.92346, and the predicted minimum wear rate was 0.3229 with desirability of 0.93323. Overall, this study demonstrates that granite powder can be effectively used as a sustainable reinforcement to improve the wear resistance of Al5052-based hybrid composites for industrial tribological applications.