Epoxy composite reinforced with jute/basalt hybrid – Characterisation and performance evaluation using machine learning techniques

Abstract

Epoxy resins, prized for their versatile properties, are derived from bio-based materials, contributing to sustainability and eco-friendliness in both production and application. This study focuses on the application of gradient boosting machine learning techniques in the field of machining to predict the surface roughness and also the contour based experimental validation of the numerical results. The turning experiments, conducted via Taguchi's L₂₇ array, aimed to explore the effects of depth of cut, feed rate, and spindle speed. Higher spindle speeds, lower feed rates, and shallower cuts led to smoother surfaces in turned jute/basalt epoxy composites. Machine learning models (Gradient Boosting Machine, AdaBoost, and XGBoost) were then used to predict surface roughness. Amongst these, XGBoost outperformed GBM and AdaBoost, exhibiting maximum and average prediction errors of 3.78 % and 2.24 %, respectively. XGBoost accurately predicted 2D surface roughness contours that closely matched experimental contours for training and test cases. Taguchi's Orthogonal Matrix identified minimum surface roughness values as 0.773 μm (experimental), 0.800 μm (GBM), 0.880 μm (AdaBoost), and 0.774 μm (XGBoost). All were achieved at 1500 rpm spindle speed, 0.05 mm/rev feed rate, and 0.3 mm depth of cut.