Wheelset bearings are the core components of the high-speed train running gear. Due to their long service time and harsh working environment, the bearings bear complex loads while running fast, which can easily cause fatigue wear, peeling, scratches, and other bearing failures, which can bring great instability to the normal and safe operation of trains. Extracting the bearing fault information concealed by interference from the measured signal is a key step to realize the fault diagnosis of high-speed train wheelset bearings. Aiming at the remaining shortcomings of the traditional empirical wavelet transform frequency band division, an empirical wavelet transform optimization method based on the trend of the frequency spectrum is proposed. Mathematical morphology closing filtering can extract the fault center frequency position of the frequency spectrum. Based on this, this paper optimizes the empirical wavelet transform method. According to the fault distribution characteristics of the bearing signal envelope spectrum, the significant fault index is proposed as the evaluation parameter of the frequency band fault information, and it is applied to the effective frequency band screening of the actual engineering signal, which improves the robustness of the method. The validity of the method is verified by bench test signals and real vehicle signals.

Wheelset bearings; Empirical wavelet transform; Closing operation; Morphological filtering; Fault diagnosis