Machine Learning Applications in Predicting Friction Properties of Bearing Steel: A Review

This review investigates the application of machine learning (ML) techniques in predicting the tribological characteristics of bearing steel, a crucial element in determining the efficiency and lifespan of mechanical systems. Traditional methods, which rely on empirical equations and physical models, often fall short in handling complex material behaviors and variable operating conditions. ML, particularly deep learning, has become a highly effective instrument for capturing nonlinear relationships and providing accurate predictions. This paper outlines key factors influencing bearing steel friction, including material composition, microstructure, and surface treatment, and discusses the application of a range of ML algorithms, including Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Regression Trees (RT). We compare their performance across various data sets and highlight key issues such as data acquisition, model generalization, and real-time prediction. Recommendations for future research are proposed to enhance the application of ML in this field.