Data‐driven modeling for predicting tribo‐performance of graphene‐incorporated glass‐fabric reinforced epoxy composites using machine learning algorithms

Abstract

AbstractThe present article aims to investigate a comparative effect of the mechanical properties and tribological operating variables (applied load and sliding distance) on the tribo‐performance of graphene incorporated woven glass fabric reinforced epoxy (GFRE) composites. Computational and data‐driven machine learning (ML) approach has been extensively applied to examine the advancement of the tribological systems. For the study, tribo‐mechanical data, gathered from previous investigations by the present authors, have been used as well. Accordingly, a predictive model, based on the ML algorithms for predicting specific wear rates (SWR), has been developed. Further, a co‐relation between the SWR and the mechanical properties of materials has been determined. For the investigation process, three ML approaches, have been applied, viz., artificial neural network (ANN), Random Forest (RF), and Gradient Boosting Machine (GBM). Feature score analysis has been done as well revealing that variation of the sliding distance, interlaminar shear strength (ILSS) and product of tensile strength (s) and elongation (e) significantly influence the SWR prediction. The trained ML models can predict the tribo‐performance from mechanical properties variables and tribological test conditions, which is impossible with conventional two‐parameter correlations. During the present study, the coefficient of determination (R2 value) in different models was found to be 0.9883, 0.9884, and 0.9762 for the ANN, RF, and GBM, respectively. The best‐performing model was RF.