Bearing Fault Diagnosis Under Variable Speed Based on Iterative TF Curve Extraction and Demodulation

Abstract

The vibration characteristics of rolling bearings under variable speed are time-varying, which brings great difficulties to fault diagnosis. An iterative time-frequency (TF) curve extraction and demodulation based method is proposed for fault diagnosis of bearings under variable speed. The envelope of the vibration signal is firstly extracted by utilizing Hilbert transform, and the instantaneous frequency associated with each envelope component can be iteratively estimated based on curve extraction from the reassigned time-frequency spectrogram derived using synchrosqueezing transform (SST). Secondly, The instantaneous fault characteristic frequency (IFCF) of bearing is extracted, and the phase mapping function for generalized demodulation is further estimated. Thirdly, the envelope signal is generalized demodulation processed with the phase mapping function, which is computed based on the IFCF, then the time-varying component is converted into a component of constant frequency. Finally, spectrum analysis is applied to the demodulated signal to identify the bearing fault characteristics. The effectiveness of this proposed method are verified using simulation data and the bearing vibration data measured under variable speed conditions.