Comparison of Morlet wavelet filter for defect diagnosis of bearings

Condition monitoring helps to avoid unexpected failures of equipments. Rolling element bearings are critical components in rotating equipments. Vibration analysis is a common method used for defect detection and diagnosis of rotating equipments without effecting their operation. The measured vibration signal contains noise, modulations and low frequency components due to unbalance, misalignment, structural losseness etc. Since the impulses due to bearing defects are having low amplitude, it is difficult to detect and identify the location of bearing defect from raw vibration signals, especially during the initial stages of defect development. Since there are more transfer segments, detection of inner race and rolling element defects are also challenging. Morlet wavelet filter (MWF) can be used for denoising of vibration signals so that condition monitoring of bearings can be performed from the denoised signals. The parameters of the wavelet need to be optimized before denoising is performed. Two algorithms used for optimization of MWF are compared in this paper. First algorithm use Shannon entropy and kurtosis for optimization of shape factor and scale of the wavelet respectively. Second algorithm uses kurtosis for optimization of wavelet parameters. Experiments are performed to obtain vibration signal of bearings with defect induced in the rolling element. MWF optimized using the proposed methods were used to denoise the vibration signals. The filtered signals are compared and performances of the algorithms are evaluated.