Tribological properties study and prediction of PTFE composites based on experiments and machine learning.

The tribological properties of materials exhibit a complex and non-linear correlation under varying operational conditions. Therefore, prioritizing a data-driven approach to predict service capability for accelerating material design and preparation is imperative in advancing tribology. The investigation was conducted to analyze the tribological performance and wear mechanism of PTFE composites. The machine learning (ML) approach was concurrently employed to predict tribological properties under diverse operational conditions. The gradient boosting regression (GBR) model demonstrated excellent predictive performance, with R2 of 82% and 91% for the friction coefficient and wear rate, respectively. Furthermore, Pearson correlation coefficient indicated that temperature and speed has a greater impact on friction coefficient and wear rate when compared to load. • The gradient boosting regression (GBR) model demonstrated outstanding predictive performance. • Pearson correlation coefficient indicated that temperature and speed had a greater impact. • With the increase in speed, the friction coefficient and wear rate exhibited a decreasing trend. [ABSTRACT FROM AUTHOR]