Fault Feature Extraction for Rolling Bearing Based on Dual Impulse Morlet Wavelet

Abstract

The common impulse feature model is oscillating and attenuated signal with a single maximum peak, while the formation principle of actual impulse feature is ignored. Considering the continuous dual impulse waveform feature of rolling bearing fault in the vibration signal and the matching effects of the single impulse waveform with Morlet wavelet, a “dual impulse Morlet wavelet” model is proposed. Through ant colony algorithm with the indicator of the maximum cross-correlation, 4 types of parameters are optimized adaptively which affect the similar degree between dual impulse Morlet wavelet and the dual impulse waveform intercepted from the bearing vibration signal. Then, the optimal model is obtained. The bearing fault experiment verification shows that the optimal dual impulse Morlet wavelet can effectively improve the analytical precision and energy concentration of impulse feature in both of time domain and frequency domain, which overcomes the disadvantages of Morlet wavelet effectively.