An Adaptive Demodulation Band Segmentation Method to Optimize Spectral Boundary and Its Application for Wheelset-Bearing Fault Detection

Abstract

Wheelset-bearing working states directly affect the stability of bogies, and existing defects may threaten the running safety of high-speed trains. Thus, the fault detection of wheelset-bearing is of great importance. In this paper, a novel wheelset-bearing fault detection method, named adaptive autocorrelated kurtogram (AAK), is proposed based on the developed autocorrelated kurtosis and the presented spectral segmentation method. Autocorrelated kurtosis is designed to reduce the unrelated components interference and increase the signal-to-noise ratio (SNR). The spectral segmentation method is proposed to obtain the fault zone as completely as possible. Based on the feature that rotation frequency is related to bearing fault frequency, the rotation frequency-based window size selection and extension strategy is proposed. With different frequency levels, the AAK is formed. The proposed method is validated by simulated and experimental data. The results show that this method can automatically and adaptively search for a reasonable demodulation bandwidth according to the fault information feature and resonance region position. The method can estimate the center frequency and bandwidth and avoid unrelated component interference. The AAK can provide accurate detection results and possesses excellent performance. Thus, it is suitable for wheelset-bearing fault detection.