Detection of broken bars on induction motors using MODWT

Broken bars are one of the most common faults in induction motors. Current and vibration signals have been studied and analyzed with diverse mathematical tool, such as the Fast Fourier Transform (FFT) that allows the analysis of signals in the frequency domain, or the Wavelet Transform (WT) that performs the analysis, maintaining the characteristics of frequency and time in the transient state of the signals. In this paper we propose to use the Maximal Overlap Discrete Wavelet Transform (MODWT) wavelet coefficients and statistical parameters such as entropy and standard deviation (STD) as classifiers to determine the half and one broken bar in the Induction Motor (IM). Following this proposal, classification percentages higher than 95% were achieved to detect one broken bar with vibration signals. Also the classification percentages for half broken bar with respect to the Discrete Wavelet Transform (DWT) are exceeded.