Feature Knowledge Based Fault Detection of Induction Motors Through the Analysis of Stator Current Data

The fault detection of electrical or mechanical anomalies in induction motors has been a challenging problem for researchers over decades to ensure the safety and economic operations of industrial processes. To address this issue, this paper studies the stator current data obtained from inverter-fed laboratory induction motors and investigates the unique signatures of the healthy and faulty motors with the aim of developing knowledge based fault detection method for performing online detection of motor fault problems, such as broken-rotor-bar and bearing faults. Stator current data collected from induction motors were analyzed by leveraging fast Fourier transform (FFT), and the FFT results were further analyzed by the independent component analysis (ICA) method to obtain independent components and signature features that are referred to as FFT-ICA features of stator currents. The resulting FFT-ICA features contain rich information on the signatures of the healthy and faulty motors, which are further analyzed to build a feature knowledge database for online fault detection. Through case studies, this paper demonstrated the high accuracy, simplicity, and robustness of the proposed fault detection scheme for fault detection of induction motors. In addition, with the integration of the feature knowledge database, prior knowledge of the motor parameters, such as rotor speed and per-unit slip, which are needed by the other motor current signature analysis (MCSA) methods, is not required for the proposed method, which makes it more efficient compared with the other MCSA methods.