This study explores the enhancement of tribological properties in epoxy composites reinforced with pineapple leaf fibre (PALF) ( Ananas comosus ) using optimization and machine learning techniques. The rationale behind this research is to develop sustainable, high-performance materials for industrial applications by utilizing bio-based fibers, which offer environmental benefits. Response Surface Methodology (RSM) with Central Composite Design (CCD) was used, involving 24 runs. Key factors included fiber weight percentage (0 wt% pure epoxy and 30 wt% PALF/epoxy), applied load (10 N and 30 N), sliding velocity (0.7 mm/s and 2 mm/s), and sliding distance (500 mm and 1500 mm). X-ray diffraction (XRD) analysis showed that pure epoxy exhibited an amorphous structure, while the 30 wt% PALF/epoxy composite displayed a peak shift to 23.9°, indicating increased crystallinity. Analysis of Variance (ANOVA) revealed that increasing fiber content improved tribological properties. The composite with 30 wt% PALF at optimized condition showed an 94 % reduction in wear rate and a 78 % decrease in the coefficient of friction (COF), with R² values of 0.9135 for wear rate and 0.9203 for COF. Three machine learning models—linear regression, gradient boosting, and Gaussian Process (GP) regression—were employed for predictions. The GP model outperformed the others, achieving R² values above 0.98. SHAP analysis identified fiber weight percentage as the most influential factor on wear rate. This study demonstrates that experimental, statistical, and machine learning methods optimize tribological properties of bio-composites for industrial applications. • Pineapple leaf fiber (PALF) reinforced epoxy composites made by hand lay-up and cured for fabrication. • X-ray Diffraction (XRD) showed a shift in the main peak to 23.9° for 30 wt.% PALF/epoxy sample. • The experiment used Response Surface Methodology (RSM) with Central Composite Design (CCD), considering fiber ratio. • Increasing the fiber content to 30 wt.% led to a 94% reduction in wear rate and a 78% decrease in friction coefficient. • Gaussian Process (GP) regression showed high predictive accuracy with a coefficient of determination (R²) > 0.98.
Louhichi et al. (Fri,) studied this question.