Application of multifractal detrended fluctuation analysis in fault diagnosis for a railway track circuit. Issue 1 (2nd January 2018)

• Record Type: Journal Article
• Authors: Wang, Zicheng; Zhang, Yadong; Guo, Jin; Su, Lina
• Is Part Of: Transactions. Volume 25:Issue 1(2018)
• Page Start: 44
• Page End: 55
• Publication Date: 2018-01-02
• Subjects: Fault diagnosis -- track circuit -- multifractal detrended fluctuation analysis -- extreme learning machine; Engineering -- Periodicals; Engineering -- China -- Hong Kong -- Periodicals; Engineering; China -- Hong Kong; Periodicals; 620
• DOI: 10.1080/1023697X.2017.1409664
• Languages: English
• ISSNs: 1023-697X

Abstract

ABSTRACT: The outdoor equipment failures of track circuits are usually not easy to be detected. In addition, the location of outdoor equipment failures can cause trouble for on-site maintainers. To solve the problem, this paper proposes a novel method for fault diagnosis of railway track circuits based on multifractal detrended fluctuation analysis (MF-DFA). Firstly, a locomotive signal induced voltage model was established based on the uniform transmission-line theory. The locomotive signal amplitude envelope (LSAE) signals of the track circuit in the normal and fault conditions were solved out. Through this model, the influence mechanism of track circuit faults on the LSAE signals was revealed. On the basis of MF-DFA, the generalised Hurst exponents and multifractal spectra of the LSAE signals were obtained. Then the six-dimensional vectors extracted from the multifractal spectra were used as the fault features. Finally, these features were input to the extreme learning machine (ELM) to identify faults. The fault diagnosis accuracy using the method proposed in this paper reached 94.2949% after k-fold cross validation. The results indicated that MF-DFA had obvious advantages in the application of track circuit fault diagnosis.