Improved Time Synchronous Averaging and Its Application in Data-Driven Rotor Fault Diagnosis

Abstract

During the assembly and operation of rolling bearings, the inner ring and outer ring are prone to misalignment, which will cause eccentric wear and seriously affects the performance and service life of the bearing. In this paper, the NARX (Nonlinear Auto-Regressive with exogenous inputs) model of a bearing-rotor system is established to evaluate the severity of eccentric wear. Effective data preprocessing is a prerequisite for accurately establishing a NARX model. Considering no large fluctuations in rotational speed and load, and strong autocorrelation of vibration signal, an improved TSA (Time Synchronous Averaging) is proposed, which is combined with the Pearson correlation function. This method can effectively determine the number of time delay points, automatically compensate the phase, and avoid the accumulation of phase errors. In the absence of a tachometer, compared to conventional TSA, the method proposed can reduce the attenuation of high-frequency components and avoid distortion of vibration signals. Finally, the diagnosis approach combined with the improved TSA is determined and verified by the experiment. The improved TSA and data-driven fault diagnosis method proposed in this paper provide a theoretical reference for assembly performance detection and operation state monitoring of the bearing-rotor system.