Bearing Fault Classification Using Spectral Portrait of Motor Current Signals and Machine Learning

Procedures for diagnosing bearing failures of induction motors based on motor current signals in published methods commonly denoised signals acquired from current sensors, then extract the typical characteristics from the denoised signals, and use a classifier to discriminate the state of bearing. However, bearing faults' signatures are exposed in the frequencies of motor current (MC) signals. Therefore, this paper presents a new technique to diagnose bearing failures in electrical machines based on the spectral portrait of MC signals. This work also depicts bank filters, which are made for encoding the spectrum boundary of MC signals rather than analyzing their entire continuous frequency band. Eventually, the state of bearings is classified using a multi-class support vector machine classifier that has been trained using a collection of encoded spectral envelope characteristics. The experimental results demonstrate that the proposed technique provides a higher precision when compared with traditional methods. © 2023 IEEE.