Data-driven framework for prediction of abrasive wear performance of recycled thermoplastic-based biocomposites

The present research was sought to propose a comprehensive machine learning based predictive framework to model and predict the abrasive wear response of polymeric biocomposites. The evaluations of abrasive wear rate of nine different specimens using a dry sand rubber wheel apparatus under five different conditions of applied loads and sliding speeds has indicated nonlinearity in wear response. In order to effectively capture this nonlinearity and enhance the predictive capability, nonlinear models were implemented. The applied load, sliding speed, polymer type, sand content, density, and compressive strength were the input features whereas abrasive wear volume was the target output. The model performance was evaluated by the coefficient of determination, root mean square error, and mean absolute. The nonlinear models have demonstrated better predictions than the linear models. The applied load and sliding speed were the most significant factors for abrasive wear. The ML framework has effectively optimized the abrasive wear performance of biocomposites.