Wavelet-mathematical morphology and envelope spectrum analysis in wind power generator gearbox fault diagnosis

A lot of fault information of wind turbine gearbox exists in vibration signal in form of modulation, but vibration signal may contain plenty of random impulse noise, white noise, etc. Although wavelet de-noising has the advantage to depress the white noise, it cannot effectively suppress the impulse noise. The mathematical morphological filter has a strong ability to suppress impulse interference, but its ability to filter out white noise is weaker than the wavelet algorithm. The traditional time domain smoothing and frequency domain de-noising methods have many shortcomings. According to this feature, we use LabVIEW as the development environment to analyze the vibration signal of gearbox. The integrated wavelet-mathematical morphology is used to achieve de-noising. The envelope spectrum analysis based on Hilbert is used to demodulate out the modulation information from original signal. Then we analyze the strength and frequency to determine the gearbox fault location and the degree of its injury. We propose a method to eliminate interference noise and analyze the fault in signal. It can be easily used in many industrial fields. Experiments presented in this paper show that the wavelet-mathematical morphology can effectively suppress interference noise, and the envelope spectrum analysis can effectively diagnose the fault information of gearbox.