Fault diagnosis method for hydraulic pump based on order tracking

Aiming at the problem of extracting fault features from hydraulic pump under variable rotating speed, a fault diagnosis method based on order tracking was proposed. Firstly, the original signal was decomposed by ensemble empirical mode decomposition into a series of intrinsic mode function (IMF) components. Afterwards, the fault sensitive IMF was selected by combining energy spectrum and spectral kurtosis. Secondly, the selected IMF was processed by Hilbert demodulation and down-sampling. Moreover, an appropriate frequency band was chosen and the signal was further processed by wavelet cluster-based bandpass filtering. Then, the fault sensitive envelope signal was obtained. Thirdly, the instantaneous frequency was extracted from the envelop signal by wavelet ridge tracking, and the time instant of phase discrimination was further calculated. Moreover, the even angle spaced resample was performed to the envelope signal. Then, the time domain non-stationary signal was transformed to the angular domain stationary signal. Finally, the stationary signal was analyzed by order tracking, and the fault diagnosis was executed according to the order spectrum. Furthermore, an application was performed to the vibration signals of hydraulic pump under variable rotating speed. The signals were respectively acquired in four typical states, which were normal, loose slipper, slipper wear and swash plate wear. Results indicate that the proposed method can effectively extract the fault features of hydraulic pump and further realize fault diagnosis.