Bearing Fault Diagnosis usingEnhanced Envelope Analysis

Abstract

Early fault detection and diagnosis of bearing is of paramount importance in wind turbines as it contributes to around 70% of gearbox and 21%-70% of generator failure. This paper presents an enhanced envelope analysis technique for bearing fault identification using vibration measurements. The raw vibration signal measured from the outer casing of the WT gearbox with faulty bearing has not only fault frequency component but also resonant frequency, discrete frequency (gear and shaft) and high frequency background noise. The present work diagnose bearing faults through a three-step process: i) using Auto- Regressive(AR) time series model to remove discrete frequency components of gear and shaft ii) fault-sensitive frequency and demodulation band selection using spectral kurtosis. The spectral kurtosis is expected to higher around the region of resonance frequencies when fault occurs otherwise the fault signal is modulated. iii) envelope analysis to obtain spectrum by Hilbert Transform (HT) and retrieval of fault information. The order of the AR model is chosen such that the residual signal exhibits maximum kurtosis and spectral kurtosis is computed through short time Fourier Transform (STFT). The enhanced envelope analysis technique for bearing fault diagnosis is verified and validated using numerical simulations and Korea Aerospace University bearing damage benchmark datasets.