Identification of the high SNR frequency band for bearing fault signature enhancement

Enhancement of the vibration signals measured from faulty bearings is the major step towards a successful fault detection and diagnosis. Bandpass filtering has shown to be an effective de-noising approach. Though simple, this method requires prior knowledge of the dominant resonance frequencies excited by the fault impacts. In this paper an on-line resonance frequency estimation algorithm is presented. This approach exploits the effect of variable shaft rotational speed on the measured vibration and applies a reliable wavelet based instantaneous frequency calculation method to find the proper center frequency for the bandpass filter. The proposed algorithm is evaluated using the vibrations measured from a faulty bearing and further validated with the results obtained based on the bearing impact analysis.