Wavelet based instantaneous power analysis for induction machine fault diagnosis

The aim of this paper is to present a wavelet based approach to detect broken bar faults in squirrel-cage induction machines. This approach uses instantaneous power as a medium for fault detection. A multi-resolution instantaneous power decomposition based on wavelet transform provides the different frequency bands whose energies are affected directly by the broken bar fault. Actually, the instantaneous power has low frequency components which are difficult to localize in the frequency domain analysis unless there is a long signal acquisition of current and voltage. By the wavelet transform, the study of energy related to the low frequency band is free of noise and improves the fault diagnosis without having the slip information. This can be especially useful when the induction machine is not fully loaded. In this paper, it is shown that for these reasons the wavelet approach applied to instantaneous power is superior to the frequency domain in the case of one and three broken bars for a three-phase squirrel cage induction machine.