Wayside acoustic detection of train bearings based on an enhanced spline-kernelled chirplet transform

Abstract Wayside acoustic detection is an effective and economical technology for fault diagnosis of train bearings. However, the technology has two main problems: Doppler Effect distortion, and high-level noise interference particularly harmonic interference. To solve both problems, a novel wayside acoustic detection scheme using an enhanced spline-kernelled chirplet transform (ESCT) method is proposed in this paper. Combining the spline-kernelled chirplet transform, built-in criterions, and a variable digital filter, the ESCT method is proposed for use in the extraction of the main harmonic components and corresponding instantaneous frequencies (IFs). This way, the residual signal, free of harmonic interference, can be obtained by excluding harmonic components in the raw acoustic signal using the ESCT method. The excluded harmonic components can be used to obtain motion parameters of the test train using a new estimation method. A resampling time vector can be constructed based on the estimated motion parameters. Doppler Effect in the residual signal can be reduced by using the time-domain interpolation resampling (TIR) method. Finally, spectral kurtosis (SK) is applied to extract train bearing fault features from the Doppler-free signal. By observing the Hilbert envelope spectrum of the filtered signal, train bearing faults can be detected. Comparing this approach with other schemes, the proposed solution requires comparatively little prior information and is easily applied to existing detection systems. The simulation and field experiments were conducted in this paper and results verified the effectiveness of the proposed method.