Autocorrelation Based Feature Extraction for Bearing Fault Detection in Induction Motors

Abstract

Bearing defects are one of the most common types of faults that occur in induction motors. For smooth running of industrial processes, accurate and fast detection of bearing faults is necessary. Majority of the bearing faults which occur in induction motors are detected using vibration signals analysis. Therefore, proper feature extraction from vibration signals is necessary to develop a reliable diagnostic system for condition monitoring of induction motors. Considering the aforesaid fact, autocorrelation based feature extraction technique is proposed for automated detection of bearing defects in induction motors. The vibration signals of healthy and different faulty bearings recorded from the drive end accelerometers were procured and autocorrelation was done to measure their self-similarity. From the resultant autocorrelation sequences, four statistical features were extracted and using analysis of variance test, their statistical significance was examined. Then the selected features were fed to random forest and k-nearest neighbor to classify bearing defects. Four classification tasks were performed and it was observed that the present approach can differentiate different bearing defects accurately with a very high degree of classification accuracy.