Wavelet Detectors for Extraction of Characteristic Features of Induction Motor Rotor Faults

The paper describes the issue of the generation of low bandwidth detection wavelet filters for the purpose of induction machines diagnostics. The solution for this problem should be characterized by good resolution both in frequency and in time domain this is because sudden and very short lasting load changes or sudden changes of control signals dictated by technology demands, cause effects in currents, voltages, vibrations of others diagnostic signal of damage machines which are of short duration. In the time domain, these effects are similar to the effects of variations of machine parameters or variation of supply system parameters which impede or even prevent an the assessment of the machine's condition. In the frequency domain, a non-stationary signal from the machine or the inverter's transient state usually becomes fuzzy in the spectrum. The approach proposed by the authors is based on time-frequency analysis of signals with non-parametric analysis of the faults' identification. Wavelet decomposition has been used, with mother wavelet active generation and choosing the optimal level of decomposition. It has been proven that a proper selection of mother wavelet for a particular signal corresponding to a specific machine's fault increases the effectiveness of fault detection. Inappropriate choice of a mother wavelet and decomposition level results in fuzzitication of the spectrum or can cause its nonlinear deformation, which impedes or even prevents achieving a proper diagnosis.