Integrated Model and Machine Learning-Based Approach for Diagnosis of Bearing Defects

Abstract

Rolling element bearings are the furthermost precarious mechanical components and used in almost all types of rotating machineries. Due to continuous operation, fatigue stresses are developed over the bearing contact surfaces which result, defects emerging over the rolling element bearing surfaces. For the continuous and smooth operation of rotating machines, it is necessary to detect these defects at their early development stage to avoid the catastrophic failure of rotating machines. In this study, an integrated model and data driven-based methodology for the diagnosis of bearing defect is presented. Data have been collected from the simulation of the mathematical model as well as from experiment. Further, statistical time domain features are calculated from the data to train the machine learning technique, i.e., artificial neural network, support vector machine and decision tree. The features vector is prepared by combining the features extracted from both simulation and experimental data. It has been found that integration of model and machine learning-based technique provides good diagnosis efficiency. © Springer Nature Singapore Pte Ltd. 2020.