A parameter optimized variational mode decomposition method for rail crack detection based on acoustic emission technique. Issue 4 (4th July 2021)

• Record Type: Journal Article
• Authors: Zhang, Xin; Sun, Tiantian; Wang, Yan; Wang, Kangwei; Shen, Yi
• Is Part Of: Nondestructive testing and evaluation. Volume 36:Issue 4(2021)
• Page Start: 411
• Page End: 439
• Publication Date: 2021-07-04
• Subjects: Rail crack detection -- acoustic emission -- variational mode decomposition -- optimised index -- permutation entropy; Non-destructive testing -- Periodicals; 620.112705
• DOI: 10.1080/10589759.2020.1785447
• Languages: English
• ISSNs: 1058-9759

Abstract

ABSTRACT: An important issue in analysing signals by the variational mode decomposition (VMD) algorithm is to confirm the number of modes and the balance parameter. In the applications of fault detection, most studies optimise parameters by the characteristics to extract fault information perfectly. Then, the results are used for subsequent operations such as fault analysis and classification. However, the optimal methods aiming at extracting fault information are not completely applicable to detect whether the fault occurs for different signal segments. To address this issue, this paper proposes a parameter optimised VMD method, and it is used to analyse acoustic emission (AE) signals from actual operating railway environment. Firstly, an optimised index is constructed based on a universally ideal decomposition result, which completely decomposes the signal without mode mixing and over-decomposition. Then, the VMD parameters are searched by the particle swarm optimisation (PSO) algorithm using the maximum index as the optimisation fitness function. Meanwhile, the permutation entropy feature of modes obtained by the optimised parameters is extracted to detect rail crack signals. After that, the proposed method is further analysed based on two different AE signals. Finally, the detection results are analysed and demonstrate the effectiveness of the proposed method.