Selection of Optimal Mother Wavelet for Fault Analysis in Induction Motor Using Stator Current Waveform

Abstract

Current signature analysis is used successfully for induction motor fault detection. FFT based spectrum analysis of current signal is applied for fault detection but it cannot be applied for detection of non-stationary signals. Signal analysis using Wavelet transform provides time and frequency domain information both of signals. It is widely used for signaled noising, signal compression and features extraction. Here, mother wavelet has been selected from different discrete wavelet transform (DWT) families using wavelet coefficients of three phase stator current signals. The data samples of stator current are collected from different faulty (electrical and mechanical faults) induction motors to perform the analysis. Different mother wavelets (db(3–10), sym(3–8), coif(1–5)) are employed for signal decomposition of the distorted stator current signals of different faulty induction motors which 198are called the wavelet coefficients of the signals. The decomposed signals are assembled back using reconstruction program and square root of mean square error (RMSE) are found for every mother wavelet. The correlation coefficients are also calculated for each mother waveform individually from the original signal and the coefficients of different mother wavelets. The optimal mother wavelet are selected comparing the values of RMSE and correlation coefficients of different mother wavelet. The optimal mother wavelet are found among the 19 mother wavelets which has smallest RMSE value and largest value of correlation coefficient for all faulty conditions and for all phases.