Prediction of Polymer Interface Wear Amount Based on Noise Emission and Machine Learning Regression

Abstract This study explores the complex nonlinear relationship between wear and noise, expanding traditional tribological methods. We conducted friction and wear experiments using two polymers and six metals across a wide temperature range, focusing on the noise generated at the friction interface. The research analyzes the wear mechanisms of polymer–metal tribopairs at low temperatures and establishes a model to clarify the relationship between wear and noise. We employed three machine learning algorithms—Light Gradient Boosting Machine (LightGBM), Adaptive Boosting (AdaBoost), and Extremely Randomized Trees (Extra Trees)—to develop a regression model that correlates noise emission with the amount of wear, enhancing feature selection and model robustness through Kernel Principal Component Analysis (KPCA).