Motor condition monitoring by empirical wavelet transform

Bearing faults are by far the biggest single source of motor failures. Both fast Fourier (frequency based) and wavelet (time-scale based) transforms are used commonly in analyzing raw vibration or current data to detect bearing faults. A hybrid method, Empirical Wavelet Transform (EWT), is used in this study to provide better accuracy in detecting faults from bearing vibration data. In the proposed method, the raw vibration data is processed by fast Fourier transform. Then, the Fourier spectrum of the vibration signal is divided into segments adaptively with each segment containing part of the frequency band. Next, the wavelet transform is applied to all segments. Finally, inverse Fourier transform is utilized to obtain time domain signal with the frequency band of interest from EWT coefficients to detect bearing faults. The bearing fault related segments are identified by comparing rms values of healthy bearing vibration signal segments with the same segments of faulty bearing. The main advantage of the proposed method is the possibility of extracting the segments of interest from the original vibration data for determining both fault type and severity.