Bearing Fault Detection in Induction Motors Employing Difference Visibility Graph

In this contribution, a novel technique to identify rolling bearing faults in induction motors is proposed employing difference visibility graph (DVG) analysis of vibration signals. To this end, vibration signals corresponding to healthy as well as different categories of bearing faults have been acquired using accelerometers located at the drive end. The acquired signals have been subsequently converted to connected graphs using the proposed DVG technique. From the obtained DVG adjacency matrices, four well-known network features have been extracted. Following that, one way analysis of variance (ANOVA) test has been carried out to measure the distinguishing ability of the extracted features between different classes. A multiclass classification task implementing support vector machines (SVM) classifier has been performed to discriminate between different categories of rolling bearing faults. The proposed methodology has been observed to return very high classification accuracy in distinguishing different categories of rolling bearing faults. The proposed technique can be potentially implemented to build an automated diagnostic system for identification of bearing faults.