Bearing Fault Diagnosis of Direct-Drive Wind Turbines Using Multiscale Filtering Spectrum

Abstract

Fault diagnosis of the bearings in direct-drive (i.e., no gearbox) wind turbines is a challenging issue due to the varying shaft rotating frequency (SRF) caused by the erratic wind environment. To remove the spectrum smearing phenomena of the SRF-related components and the disturbances of the SRF-unrelated components in a measured signal, this paper proposes a novel method, called multiscale filtering spectrum (MFS), to obtain the weighted energy distribution of the monocomponent signals within a local order range based on the Vold-Kalman filter (VKF). First, the instantaneous SRF of the wind turbine is estimated from a generator current signal. Then, a VKF-based multiscale filter bank is designed according to the center frequencies corresponding to the SRF at different scales. The monocomponent signals whose frequencies are continuous multipliers of the SRF are subsequently extracted from the envelope of the measured current or vibration signal. Finally, a weighted energy spectrum is constructed within the selected order range, from which possible bearing fault characteristic orders can be identified. Simulation and experiment results show that the proposed new MFS method can enhance the characteristic orders and suppress the noise, and therefore has better performance than the traditional angular resampling method for bearing fault diagnosis of direct-drive wind turbines under varying speed conditions.