An Analytical Comparison between DWT and Hilbert-Huang-Based Methods for the Diagnosis of Rotor Asymmetries in Induction Machines

In the paper two alternative tools are applied and compared in order to diagnose the presence of rotor asymmetries in induction machines. Both tools are applied to the stator startup current. The objective is to extract the evolution during the startup transient of the left sideband harmonic associated with the asymmetry, which constitutes a reliable evidence of the presence of the fault. The first tool is the discrete wavelet transform (DWT) and its validity for the diagnosis was proven by the authors in previous works, even in cases where the classical Fourier approach does not lead to correct results. Despite its good results, some constraints remained, such as the selection of an optimal mother wavelet or the possible overlap between frequency bands associated with the wavelet signals. In the paper, an alternative time-frequency decomposition tool, the Hilbert-Huang transform (HHT), is applied to the startup current for detecting the harmonic. This tool might allow avoiding some of the limitations of the DWT, maintaining its reliability. Both approaches are applied to experimental signals obtained under various operation conditions. Finally, the results are analyzed and compared, showing the robustness of both approaches for the diagnosis of rotor asymmetries in induction machines.