VIBRATION SIGNAL ANALYSIS AND FEATURE EXTRACTION

Abstract

Use of vibration spectrum for fault diagnosis of electrical machines is very well known. By monitoring the magnitude of fault characteristic frequencies present in the vibration spectrum, it is possible to predict the supply condition, mechanical condition and above all, the bearing condition of the rotating electrical machine. However, while spectrum analysis of vibration signal using Fast Fourier Transform (FFT) has the potential for providing a major improvement over time based analysis, there are many examples where FFT does not provide all insight needed to identify the incipient fault developing in the motor. In the present work, an attempt was made to use Wavelet Transform as an alternative tool for identification of incipient fault in the machine Various Wavelet families are tested and it is found that Debauchee Wavelet provides strong fault information, and hence can be used for determination of various faulty conditions like bearing faults, single phase condition and unbalanced voltage supply in the motor. The RMS values of decomposed signals were calculated. Comparison of RMS values of decomposed signals for healthy machine and machine under fault have been made. iii