Investigation of tribological behaviour of B ₄ C–graphene glass fiber composites using Taguchi, multi-criteria decision making, and machine learning approaches

The current research has focused on developing epoxy-based glass fiber composites reinforced with boron carbide (B 4 C) and graphene (Gr) particles processed through a compression moulding process for enhanced tribological performance. Through Taguchi’s L 32 experimental design, the experimental layout has been prepared by varying material configuration, sliding speed, normal load, and sliding time to observe their effect on wear rate and coefficient of friction. To address the multi-response problem, criteria weighting have been evaluated through Shannon’s entropy and CRITIC techniques integrated with TOPSIS and CoCoSo MCDM methods. Both methods have identified the optimal condition for minimizing wear and CoF as the hybrid filler composite tested at 600 rpm, 20 N, and 4 min. Sensitivity analysis across the four weighting combinations indicated a strong correlation (≈0.9) between the CRITIC TOPSIS and CRITIC CoCoSo rankings. Multivariate ANOVA has ensured and emphasized the strong interaction of material configuration and speed over both the responses studied. The AdaBoost algorithm has exhibited superior predictive accuracy, with an R 2 of 0.99 among the machine learning models. SEM analysis revealed fiber breakage, matrix cracking, void formation, and debris formation. The findings establish the potential of composites developed with hybrid fillers for applications demanding good tribological performance.