Fault diagnosis for railway track circuit based on wavelet packet power spectrum and ELM

Abstract

For enhancing the troubleshooting efficiency of a track circuit, a fault diagnosis method for the track circuit is proposed in this paper. First, a locomotive signal induced voltage model is established based on the transmission-line theory. Then, cases of the induced voltage envelope signals, when the track circuits are in the normal and fault conditions, respectively, are simulated. Next, a three-layer wavelet packet is adopted to decompose the induced voltage envelope signals and power spectrum analysis for the detail signal is realized. 16 time-domain indices of the β power spectrum including the standard deviation, variance, kurtosis value, and the variable coefficient are used as the failure features. Then, the information fusion of the time domain features is implemented using the principal component analysis (PCA) technology. Finally, the fusion features are input to an extreme learning machine (ELM) model to identify the failures. Case analyses show that the fault diagnosis method proposed in this paper can obtain a high accuracy and provide a scientific basis for the on-site maintenance of the track circuit.