A Fault Diagnosis Method for Train Plug Doors Based on MNPE and IPSO-MSVM

• Published in: 2018 International Conference on Control, Automation and Information Sciences (ICCAIS) pp. 467 - 471
• Main Authors: Sun, Yongkui, Xie, Guo, Cao, Yuan, Wen, Tao
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2018
• Subjects: Automation; Fault diagnosis; Feature extraction; multi-scale permutation entropy; Plugs; PSO; Support vector machines; SVM; Time series analysis; train plug doors; Training
• ISSN: 2475-7896
• DOI: 10.1109/ICCAIS.2018.8570324

Train plug doors play a vital role to keep train operation safe and reliable. Taking the advantages of sound signals based fault diagnosis method into consideration, a novel fault diagnosis method for train plug doors based on multiscale normalized permutation entropy (MNPE) and an improved particle swarm optimization based multi-class SVM (IPSO-MSVM) is proposed. First, empirical mode decomposition (EMD) is used to decompose each sound signal into a series of intrinsic mode functions (IMFs) and a residue for stationary processing. Then, MNPE features are extracted from the IMFs. And Fisher discrimination criterion is utilized to obtain the most significant features as feature vectors. Further, an improved PSO (IPSO) is used to search for the optimal parameters of multi-class SVM. Finally, the IPSO-MSVM model is trained and verified using training set and test set, respectively. The identification accuracy of the proposed method reaches 90.54%, which is higher than backpropagation (BP) neural network classifier and 1 Nearest Neighbor (1NN) classifier, indicating the proposed method for fault diagnosis on train plug doors is feasible.