Tribological properties study and prediction of QBe2 beryllium bronze and 7075-T6 aluminum alloys based on machine learning under mixed lubrication

Abstract

Tribological properties of materials exhibit complex and non-linear correlation with working conditions under mixed lubrication. Selecting an appropriate data-driven method to predict tribological properties is important for accelerating material design and preparation. This paper investigates the tribological performance and wear mechanisms of QBe2 beryllium bronze and 7075-T6 aluminum alloy pairs under grease lubrication by using pin-on-disk friction and wear tests. The tribological performance under different conditions is further predicted using four machine learning algorithms: K-nearest Neighbors (KNN), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Random Forest (RF). The experimental and machine learning results both show that reciprocating frequency has the most significant influence. The dominant wear mechanisms include ploughing and adhesive wear. Furthermore, among the four machine learning models, SVM model performs the best in predicting tribological properties of QBe2 beryllium bronze and 7075-T6 aluminum alloy under mixed lubrication.