Computational data-driven based optimization of tribological performance of graphene filled glass fiber reinforced polymer composite using machine learning approach

Abstract

Data-driven computational analysis based on predictive modelling reveals novel insight into the tribological performance of the material. It also establishes co-relation in the experimental data and mechanical properties of materials. We build a predictive model using a machine-learning (ML) algorithm to predict specific wear rates by using experimental data. Using pin-on-disc tribometer, tribological tests of glass fiber reinforced epoxy composite (GFRE) incorporated with and without graphene nanoplatelets (GNPs). The test was conducted under the operating parameters of variable applied load of 5, 10, 15, 20 N, and abrading distance of 200 m at a rotating speed of 300 rpm. The experiment shows that neat GFRE has higher mass loss than GNPs incorporated GFRE sample, whereas 1 wt% Graphene nanoplatelets incorporated samples have the lowest mass loss. We used four Machine learning algorithms: Deep Neural Network (DNN), Gradient Boosting Machine (GBM), Random Forest (RF), and Extreme Gradient Boosting Machine (XGboost) to build the predictive model of experimental datasets. The Machine Learning algorithms successfully predict with an R2 value of 0.996, 0.980, 0.988, and 0.999 for DNN, GBM, RF, and XGboost, respectively. We also performed feature score analysis to find the contribution of input parameters to predicting the specific wear rate of the composites, which revealed that variation of load and density have maximum influence in determining specific wear rate prediction.