Electronic Science and Technology ›› 2022, Vol. 35 ›› Issue (4): 40-46.doi: 10.16180/j.cnki.issn1007-7820.2022.04.007

Previous Articles     Next Articles

Track Disease Diagnosis Method Based on VMD and BP Neural Network

Li HUA,Jian YANG,Tianchen YUAN,Ruigang SONG   

  1. School of Urban Rail Transit,Shanghai University of Engineering Science,Shanghai 201620,China
  • Received:2020-12-01 Online:2022-04-15 Published:2022-04-15
  • Supported by:
    National Natural Science Foundation of China(5207052806);Shanghai Natural Science Foundation(19ZR1421700);Postgraduate Innovation Project Fund of Shanghai University of Engineering Science


In view of the problem of difficulty in extracting disease features from non-linear and unsteady sleeper vibration signals, this study proposes a track disease feature extraction method based on variational modal decomposition and multi-scale permutation entropy, and adopts the BP neural network disease diagnosis model to perform disease identification. The variational modal decomposition method is used to decompose the collected vibration acceleration signals to obtain several eigenmode components. The multi-scale permutation entropy value of these eigenmode components is calculated and used as the high-dimensional feature vector of the track disease to realize the noise reduction of the sleeper vibration signal and the extraction of the disease feature. Through the establishment of a BP neural network disease diagnosis model, high-dimensional feature vectors are input into the BP network for training, fitting, and verification, and compared with the method of combining empirical mode decomposition and BP neural network. The analysis results show that the proposed method has a higher recognition accuracy rate and can effectively diagnose disease.

Key words: variational modal decomposition, multi-scale permutation entropy, BP neural network, eigenmode component, noise reduction, high-dimensional feature vector, empirical mode decomposition, disease diagnosis

CLC Number: 

  • TN911.7