Application of taguchi design and machine learning techniques for optimising the tribological performance of Al/B4C/ZrO2 hybrid composite

Despite growing interest in aluminium-based hybrid composites for tribological applications, a clear understanding and optimisation of the wear behaviour of Al/B 4 C/ZrO 2 hybrid composites fabricated by powder metallurgy remains limited, particularly with respect to the combined influence of reinforcement composition and operating parameters. The research examines frictional behaviour and the parametric enhancement of an aluminium-based hybrid composite (HC) strengthened with boron carbide and zirconia particles, produced via powder metallurgy techniques. Composites containing 2, 5, and 8 wt% reinforcements were produced through high-energy ball milling, followed by compaction at 700 MPa and sintering at 750°C under argon atmosphere to ensure uniform particle dispersion and strong interfacial bonding. Dry sliding wear experiments were performed using a pin-on-disc tribometer at normal loads of 10–30 N, with sliding speeds of 0.5–1.5 m/s and sliding distances of 500–1000 m. Experimental results demonstrated significant dependence of wear loss on both operating parameters and reinforcement composition. Taguchi-based optimisation determined the lowest wear rate at a load of 20 N, a sliding speed of 1.5 m/s, a sliding distance of 1000 m, and 5 wt% B 4 C, and 8 wt% ZrO 2 . Analysis of variance indicated that sliding velocity was the most influential parameter, contributing 44.94%, followed by sliding distance at 24.39% and applied load at 11.42%, whereas ZrO 2 and B 4 C accounted for 10.28% and 4.75%, respectively, with an experimental error of 4.21%. Linear regression achieved 82.81% predictive accuracy, whereas Random Forest and Polynomial Regression improved R 2 to beyond 0.95 with minimal prediction errors. The integrated statistical and machine-learning framework provides reliable multi-parameter optimisation of hybrid aluminium composites for advanced tribological applications.