Intelligent fault detection of electric motor bearing using wavelet packet transform and ED-RBFNN

Abstract

The ubiquitous nature of electric motors across diverse industries highlights the critical role of components like rolling bearings in ensuring reliable operation. Bearing defects are a leading cause of rotating machinery failure, underscoring the importance of effective diagnostic methods. This research presents an intelligent system for rolling bearing fault detection in electric motors. The proposed system utilizes vibration analysis and Daubechies wavelet packet decomposition to extract approximate and detail signals across three levels. Innovatively, this research combines Wavelet Packet Transform (WPT) with an Enhanced Dynamic Radial Basis Function Neural Network (ED-RBFNN) for intelligent fault detection. The combination of WPT and ED-RBFNN provides adaptive and accurate fault detection; WPT decomposes signals to reveal subtle fault indicators, while ED-RBFNN dynamically learns and classifies these patterns, adapting to non-static conditions and offering robust feature extraction and dynamic learning. Extracted features, both in the time and frequency domains, are subjected to a distance evaluation method for feature selection, resulting in a feature vector that serves as input to the ED-RBFNN. The system, trained and tested on separate datasets, demonstrates high reliability and accuracy in detecting the presence and location of rolling bearing faults, effectively distinguishing between healthy and defective bearings. This robust solution for predictive maintenance in industrial settings serves as a valuable tool for minimizing downtime and improving the operational lifespan of rotating machinery, contributing to the development of advanced predictive maintenance techniques and ensuring reliable and efficient industrial operations.