Tribological performance study and prediction of copper coated by MoS2 based on GBRT method

Fabricating solid lubricating coating on the metal surface had been widely used due to excellent wear resistance. However, its tribological performance became rather complex under different working condition. In this study, we employed machine learning (ML) to predict their tribological properties after experimental investigations and molecular dynamics (MD) simulations. Firstly, copper coated by molybdenum disulfide (MoS2) was prepared with varying thicknesses. Then, their tribological properties were studied under different loads and reciprocating frequencies to explore the wear mechanism from both macroscopic scale and nano scale. Importantly, correlations between friction and wear of coatings with testing parameters were investigated by predicting Coefficient of Friction (COF) and wear rate based on ML algorithm of Gradient Boosting Regression Tree (GBRT). The results showed that the thicker coating exhibited a smaller friction coefficient and more severe wear owing to the low hardness, which was also demonstrated by experiments and MD simulations. The friction coefficient and wear increased with the increase of load, but only the friction coefficient growth with the increase of reciprocating frequency. In addition, the GBRT model can effectively predict the tribological properties of the MoS2 coating on the copper substrate and the prediction accuracy of friction coefficient and wear rate reached 94.6% and 96.3%, respectively. Furthermore, relative importance analysis revealed that load had the greatest effect both on predicting friction coefficient and wear rate.