Researching the fault diagnosis of hydraulic pump by wavelet analysist

The paper have a compare in removing noise between the traditional method and wavelet method. The former is applied for signal in the frequent field, whose variance can not be given at certain time and any sudden change at time axis will affect the whole signal wave chart. Above methods only are reasonable to calm-signal, except for non-calm signal. The latter could handle the signal between in time field and frequent field at simultaneity. Through decomposition the signals with wavelet, holding the wavelet transform result of frequency band you choice, reconfiguration the measured signals it is easy to strengthen the failure signals, so it can different the sudden change and noise efficiently to disperse signal noise. Directing to above problems, This paper indicates the primary failure characteristics of hydraulic pump are fatigue crackle, surface abrasion and gas-corroding. These errors could destroy the oil membrane between portplate and slipper wear and occurs friction pairs., The friction brings the additional vibrant signal, so we have to mount the sensor to gain the fault characteristics in the hydraulic pump shell. Through decomposition and reconfiguration the measured signals with wavelet, it is easy to eliminate the noise and to strengthen the failure signals effectively. The experimental results indicate that the wavelet analysis can divide the signals into multi-frequency band when portplate and slipper wear occurred among friction pairs in hydraulic pump. According to the signal characteristics, the frequency band was selected adaptively so as to promote the failure information, then the fault diagnosis was realized based on small SNR(Signal-to-Noise Rate)signals. The experimental results indicate that the method is feasible.