Intelligent retrieval of wear rate prediction for hypereutectoid steel

Machine learning (ML) has emerged as an excellent technique for predicting the wear properties of solid materials. The present work investigates the wear performance of 0.92% C and 1.57% C hypereutectoid steels under various operating conditions (sliding speed, normal pressure, and sliding distance). ML algorithms such as random forest (RF), linear regression (LR), Ada boost (AdaB), gradient boost (GB), gaussian process regression (GPR), k-nearest neighbor (KNN), and support vector machine (SVM) were applied to the outcomes of the experiments to predict the wear rate. The seven ML algorithms were ranked in order of prediction accuracy: GB, RF, AdaB, GPR, SVM, KNN, and LR. GB had high-caliber results in R2 (training and test), MAE, and RMSE out of the applied models. The worn surfaces and debris showed an oxidative wear mechanism of 0.92% C and an abrasive wear mechanism of 1.57% C. The results might help in the creation of hypereutectoid steels with measured wear properties; hence it speeds up the development of new operational hypereutectoid steels.