Optimization and prediction of wear in LM25–B₄C/EN31 tribopair using Taguchi method and machine learning model

Abstract

This study aimed to analyse the interfacial contact between LM25-B4C composite and steel EN 31 tribopairs via pin-on-disc wear tests. The Taguchi approach was employed to design the pin-on-disc experiments to evaluate the tribological properties of the material system. The optimal conditions for the lowest wear rate were identified at 12 wt.% reinforcement, a low load of 19.62 N, and a sliding distance of 1200 m.  SEM micrographs show formation of a homogeneous α-aluminium dendrites network structure developed due to super-cooling of casting during solidification. EDS spectrum of 12 wt.% B4C composite confirmed high content of ceramic elements which improved mechanical and wear properties. Coefficient of friction (COF) analysis showed a higher value reduction with 12 wt. %B4C additC addition, and its percentage reduction increased at higher load and with increase in sliding distance in all composites. Wear mechanisms transition from two-body abrasion and ploughing in the base alloy and lower-reinforced composites to three-body abrasion and plastic deformation in the 12 wt.% B₄C composite. This transition is attributed to the high compressive strength of ceramic particles, which restrains severe delamination wear. Furthermore, four machine learning models predicted wear and COF, with polynomial regression performing best for wear rate and linear regression for COF.