Wavelet Analysis for Effective Detection of Broken Rotor Bar in Induction Machine

Early detection of the induction machines (IM) faults reduces maintenance costs and prevents the unscheduled downtimes. The key for successful fault detection of IM is achieved by suitable condition monitoring and signal processing (SP). The Motor Current Signature Analysis (MCSA) is considered the most popular fault detection method in any IM. However without an effective SP method, the fault detection fails. Basic SP techniques which enhance the fault signature and suppress the dominant system dynamics and noise must be considered. Among different SP methods the wavelet transform is a powerful tool to analysis of non-stationary signals. Simultaneous time and frequency analysis can be performed by this technique, which decomposes original signal into different frequency levels. However, there are different families in wavelet analysis which the efficiency of encoding, denoising, compressing, decomposing and reconstructing signals is not the same. Meanwhile the obtained results are unique to the selected family. It is therefore desirable to select the powerful wavelet family which produces the best results for the signal being analyzed. The current signals have been processed by different families of wavelet and results have been compared for effective detection of broken rotor bars in IM.

Induction machine, Fault detection, Broken rotor bar, Motor Current Signature Analysis, Wavelet analysis