Abstract: Local defects in bearings give rise to periodic impulses in vibrations. Resonance demodulation is widely adopted to extract such impulse train from vibrations. In order for the resonance demodulation to be used to good effect, various optimization procedure and index have been introduced to optimize the frequency band of resonance demodulation. Unfortunately, most of the optimization procedures available are computationally intensive, whilst the indices only depict the existence and level of an impact without considering its periodical occurrence. Targeted at such deficiencies, a procedure which searches for optimal center frequency with a high resolution while bandwidth with a low resolution is presented so as to alleviate computation burden. Moreover, a new index termed crest factor of envelope spectrum is formulated as the ratio of the maximum value to the root mean square of envelope spectrum, taking into account both the level and periodicity of impulses. The proposed procedure and index is employed to optimize the parameters of complex shifted Morlet wavelet for bearing defect diagnosis, and results on bearing run-to-failure experiment and filed data are in favorable of them in comparison with kurtosis, smoothness index and Shannon entropy.

Key words: frequency band optimization, Morlet wavelet, wavelet filter, resonance demodulation