A Detection Method for Bearing Faults Using Complex-Valued Null Space Pursuit and 1.5-Dimensional Teager Energy Spectrum
De Zhu, Qingwei Gao, Dong Sun, Yixiang Lu
Abstract
Because of the nonlinear and unsteady characteristics of bearing vibration signals, the fault diagnosis of the rolling bearing in complex environments has become a challenging task. This paper deals with the bearing fault diagnosis method of vibration signals based on the complex-valued null space pursuit algorithm (NSP-CD) combined with the 1.5-dimensional Teager energy spectrum (1.5DTES). In the proposed algorithm, the bearing fault signals are decomposed by NSP-CD that is deduced by using partial differential equations and alternating direction method of multipliers, and several components that conform to the vibration signal mode are obtained. Then, the component containing the principal fault characteristic is selected according to the kurtosis value of each component. Finally, in order to accurately track the variation of amplitude and reduce noise interference, the 1.5DTES of the selected component is used to demonstrate the fault frequency of the bearing vibration signal. Experiments on vibration signals collected from electro-discharge machining with inner and outer ring faults demonstrate that the proposed method is more efficient and robust than other bearing diagnosis methods.