On-Board Identification Algorithm for Rail Weld Irregularity of Heavy-Haul Railway Based on Rail Weld Joint Characteristics

The rail weld joint is a weak point in the track of heavy haul railways. The detection method based on the wheelset dynamic response of freight vehicles can promptly evaluate the health status of rail weld joints, which is crucial for ensuring the safe operation of heavy haul railways. A freight vehicle-ballasted track coupling model with the rail weld joint, verified by measured data, is established. The influence of rail weld irregularity on wheelset vibration response is thoroughly analyzed in both time and frequency domains, providing a theoretical foundation for the on-board identification algorithm. The proposed rail weld irregularity identification algorithm includes localization and wave depth detection. In the localization algorithm, the characteristic waveform of the rail weld joint is introduced to construct the Weld Impact Index (WII) based on a scaled wavelet-like base function, which addresses the fluctuation of the existing track impact index and enhances discernment under diverse conditions. The threshold segmentation ofWIIis employed to accurately identify and locate the rail weld irregularity in the wheelset vibration acceleration signal. In the wave depth detection algorithm, an Empirical Fourier Decomposition algorithm optimized by kurtosis distribution index (Q-EFD) is proposed to extract the feature component caused by rail weld joint irregularity from the wheelset vibration response. The kurtosis distribution index evaluates the concentration of impact in all components. The parameter adaptation of the Empirical Fourier Decomposition is based on the kurtosis distribution index and the correlation between the reconstructed signal and the fault waveform. Then, a feature vector containing amplitude and shape features of the feature component is constructed, and the nonlinear classifier KNN is used to classify the depth of rail weld irregularity. Simulation data and field data confirm the advantages ofWIIand Q-EFD in detecting rail weld irregularity.