Identification and Fault Diagnosis of Rolling Element Bearings Using Dimension Theory and Machine Learning Techniques

Abstract

The study presents the classification of bearing fault types occurring in rotating machines using machine learning techniques. Recent condition monitoring demands all-inclusive but precise fault diagnosis for industrial machines. The utilization of mathematical modeling with machine learning may be combined for fine fault diagnosis under different working conditions. The current study presents a blend of dimensional analysis (DA) and a K-nearest neighbor (KNN) to diagnose faults in industrial roller bearings. Vibrational responses are collected for several industrial machines under diverse operational conditions. Bearing faults are identified using the DA model with 3.62% error (avg) and classified using KNN with 98.67% accuracy. Comparing the performance of models with experimental and artificial neural networks (ANN) validated the potential of the current approach. The results showed that the KNN demonstrates superior performance in terms of feature prediction and extraction of industrial bearing.