A novel Doppler effect reduction method for wayside acoustic train bearing fault detection systems

Research output: Contribution to journalArticlepeer-review

Abstract

Wayside acoustic detection of train bearing faults plays a significant role in maintaining safety in the railway transport system. Due to the relative movement between the train and the detection system, the collected acoustic signals are distorted by the Doppler Effect which results in frequency-domain distortion. Combining the multi-scale chirplet path pursuit (MSCPP) method, a variable digital filter (VDF), and a new motion parameter estimation method, a novel Doppler Effect reduction method is proposed. This can be used by wayside acoustic monitoring systems to improve detection system for train bearing faults, as illustrated in this paper. The MSCPP method with the build-in criterions is firstly used to estimate the instantaneous frequencies (IFs) of harmonic components in the wayside acoustic signals. Next, VDFs whose centre frequencies are the fitted IFs are constructed to exclude harmonic components. Using these, residual signals, free of strong harmonic interferences, can be obtained. At the same time, the motion parameters can be obtained by using a recently developed estimation method based on fitted IFs. The residual signal is then resampled to reduce the Doppler Effect by using the resampling time vector constructed using those estimated motion parameters. Finally, any bearing fault features can be extracted using the spectral kurtosis (SK) method. The effectiveness of the proposed signal processing method is verified by simulation and field-based experiments, as demonstrated in this paper.

Details

Original languageEnglish
Pages (from-to)112-124
Number of pages13
JournalApplied Acoustics
Volume145
Early online date10 Oct 2018
Publication statusPublished - Feb 2019

Keywords

  • Doppler Effect reduction, Train axle bearing, Wayside acoustic detection, Multi-scale chirplet path pursuit, Variable digital filter, Time-domain interpolation re-sampling