LAUSR

Abstract The wheelset-bearing system is one of the most critical subsystems in a high-speed train. Detecting faults is significant for ensuring the operational safety of a high-speed train. The key to detect a wheelset-bearing system fault is to extract the weak and periodic impulse responses (WPIRs) from the collected acceleration signals of the axle box. To improve the poor performance of the tunable Q-factor wavelet transform (TQWT) in extracting WPIRs, a multi-Q-factor (MQ), multi-level (ML) tunable Q-factor wavelet transform (MQML-TQWT) with an almost constant bandwidth characteristic is proposed. The optimal decomposition bandwidth and decomposition level of the MQML-TQWT are automatically selected based on the maximal envelope spectral kurtosis (MESK). The MQML-TQWT with the optimal decomposition bandwidth and decomposition level is used to extract WPIRs. The proposed method is verified by simulated signals and vibration signals measured from a wheelset-bearing system test bench. The results indicate that the MQML-TQWT effectively extracts WPIRs that may be associated with early faults in the wheelset-bearing system and is superior to the existing TQWT in extracting WPIRs.