The use of the wavelet approximation signal as a tool for the diagnosis of rotor bar failures

The aim of this paper is to present a way for the diagnosis of rotor bar breakages in induction machines. The method is focused on the study of an approximation signal resulting from the wavelet decomposition of the startup stator current. The existence of a left sideband harmonic is used as an evidence of the rotor failure for most of the diagnosis methods based on the analysis of the stator current. Thus, a detailed description of the evolution of the left sideband harmonic during the startup transient is given in this paper; for this purpose, some physical considerations are done and a numerical model is used. In this sense, it is shown that the approximation signal of a particular level, obtained from the discrete wavelet transform (DWT) of the stator startup current, practically reproduces the time evolution of the left sideband harmonic during the startup. The diagnosis method proposed consist on checking if the selected approximation signal matches with the shape of the left sideband harmonic evolution described before. The method is checked through simulation and laboratory tests, being proved that it can be a useful tool for the rotor bar breakage diagnosis.