Abstract：In view of the nonlinear,non-stationary and complex modulation of the bearing vibration signal in a strong noise background,a new method is proposed for fault diagnosis based on wavelet packet entropy and autocorrelation function. Firstly,wavelet packet de-noising signal is performed by means of the signal of wavelet packet entropy. Then,the interference of noise to the characteristic frequency is suppressed by the method of correlation function analysis,and the energy operator demodulation of the signal is carried out,so as to realize the amplitude and frequency information of the fault signal of the bearing. The experimental analysis of mechanical fault vibration signal shows that the noise reduction effect is not ideal and the accuracy and the ordinary Hilbert demodulation method cannot meet the needs of the diagnosis compared with the wavelet packet analysis. The proposed method can effectively process the fault frequency information and realize the inference of fault types.

Keyword：wavelet package entropy;autocorrelation function;energy operator;modemod;fault diagnosis;