Detectability of rotor failure for induction motors through stator current based on advanced signal processing approaches

Abstract

This paper investigated the ability of the diagnosis techniques and detectability of induction motor faults through a stator current. The proposed techniques are based on advanced signal processing tools. These methods can be classified into: high resolution approaches and time–frequency representations. Sadly, the Fast Fourier transform technique cannot give good results such as the spectral leakage, it needs a big number of measurement data samples. To address these problems, the Multiple Signal Classification technique allows for reducing the spectrum noises and to reduce the computation of signal data samples, requires less memory. However, for the diagnosis in time varying conditions, non-stationary approaches are required to diagnose and detection IM failures in variable speed operation or transient. This article is intended for a comparative study between the spectrogram, the scalogram and the Hilbert-Huang transform. In this context, the results exhibit the effectiveness of the methodology to detect induction machine fault in time varying, it is capable to detect a rotor failure. The performances of these approaches are demonstrated in simulation results using the MATLAB environment and in the experimental validation.