Prediction of Friction and Wear Mechanisms of Gr/ PTFE Nanocomposites Based on Multiscale Simulation and Machine Learning

ABSTRACT Polymer nanocomposites are widely used in high‐end bearings and gear components; their friction and wear mechanisms are significant for improving the service performance and lifespan of high‐end equipment. However, traditional experiments are expensive and time‐consuming and often fail to comprehensively reveal the friction and wear mechanisms of materials, posing challenges in predicting the wear scar morphology. To overcome this issue, this study employs a multiscale approach combining molecular dynamics (MD) simulations, finite element (FE) friction simulations, experiments, and machine learning (ML) to predict the friction and wear mechanisms of graphene (Gr)/polytetrafluoroethylene (PTFE) nanocomposites. Under velocities of 0.064 and 0.08 m/s and loads of 2–10 N, the optimal Gr mass fraction is 1.0 wt.%. Then this study extracts the microstructural information of the nanocomposites from the multiscale friction model (MD and FE) and couples it with experimental parameters to construct an ML prediction model based on an artificial neural network algorithm. By leveraging the trained nonlinear input–output relationships, the model successfully predicts the friction coefficient, wear amount, and wear scar morphology. The predicted wear amount deviates from experimental values by only 2.8%, and the wear scar depth profile shows 96.8% overlap. Furthermore, its reliability was experimentally verified. The results demonstrate that the coefficient of friction (COF) initially increases with load, then decreases, while wear amount rises monotonically; these trends are validated by experimental errors of 2.35%–13.36% across different loads. This validated multiscale prediction framework provides actionable insights for the wear‐resistant design and lifespan prediction of Gr/PTFE components in engineering applications.