LSTM Based Bearing Fault Diagnosis of Electrical Machines using Motor Current Signal

Abstract

Rolling element bearings are one of the most critical components of rotating machinery, with bearing faults amounting up to 50% of the faults in electrical machines. Therefore, the bearing fault diagnosis has attracted attention of many researchers. Typically, the bearing fault diagnosis is performed using vibration signals from the machine. In addition, by using deep learning algorithms on the vibration signals, the fault detection accuracy close to 100% can be achieved. However, measurement of vibration signals requires an additional sensor, which is not present in majority of the machines. Nevertheless, with an alternative approach, the stator current can be used for diagnosis. Therefore, the paper emphasizes on the diagnosis of bearing faults using the stator current. The diagnosis requires signal processing for the fault signature extraction that is buried underneath the noise in the current signal. The paper uses the Paderborn University damaged bearing dataset, which contains stator current data from healthy, real damaged inner raceway and real damaged outer raceway bearings with different fault severity. For fault signature extraction from the current signals, the redundant frequencies in the signals are filtered, then from the filtered signals eight features are extracted from the time and time-frequency domain by using the wavelet packet decomposition (WPD). Then, using these features and the well known deep learning algorithm Long Short-Term Memory (LSTM), bearing fault classification is made. The deep learning LSTM algorithm is mostly used in speech recognition due to its time coherence, but in this paper, the ability of LSTM is also demonstrated with the fault classification accuracy of 96%, which outperforms most of the present algorithms that perform bearing fault diagnosis using stator current. The method developed is independent of the speed and the loading conditions.