A novel acoustic emission source location method for crack monitoring of orthotropic steel plates. (15th February 2022)

• Record Type: Journal Article
• Authors: Li, Dan; Nie, Jia-Hao; Ren, Wei-Xin; Ng, Wee-Hoe; Wang, Guo-Hua; Wang, Yang
• Is Part Of: Engineering structures. Volume 253(2022)
• Publication Date: 2022-02-15
• Subjects: Orthotropic steel plates -- Crack location -- Acoustic emission -- Empirical wavelet transform -- Long short-term memory neural network; Structural engineering -- Periodicals; Structural analysis (Engineering) -- Periodicals; Construction, Technique de la -- Périodiques; Génie parasismique -- Périodiques; Pression du vent -- Périodiques; Earthquake engineering; Structural engineering; Wind-pressure; Periodicals; 624.105
• DOI: 10.1016/j.engstruct.2021.113717
• Languages: English
• ISSNs: 0141-0296

Abstract

Highlights: An AE-based crack location method is proposed for orthotropic steel plates. Empirical wavelet transform extracts distinct dispersive modes of AE signals. Long short-term memory model relates arrival time differences to source coordinates. It is proved to outperform other AE source location methods by field experiment. Abstract: As the fatigue problem of orthotropic steel plates becomes increasingly severe in medium- and long-span bridges, structural health monitoring of the plates that could identify cracks timely is in great demand. Considering that orthotropic steel plates are of multiple structural elements and complex geometries, it is challenging to locate the cracks accurately. To this end, a novel acoustic emission (AE) source location method based on empirical wavelet transform and long short-term memory neural network was developed for crack monitoring of orthotropic steel plates. Empirical wavelet transform adaptively decomposed prominent components contained in the raw AE signal, of which the one with distinct dispersive modes was selected as the characteristic component to more precisely determine the arrival time through Akiake information criteria. The long short-term memory model established a more exact relationship between the arrival time differences of AE signals at multiple sensors and the source coordinates. With the help of empirical wavelet transform and long short-term memory model, it significantly eliminated the influence of dispersion, reflection, scattering, multiple propagation paths and noise, and thus guaranteed higher crack location accuracy. Through field experiment on a long-span suspension bridge, where cracks were simulated by artificial AE sources, a comprehensive comparison was carried out among traditional time-of-arrival method, delta-T mapping method, improved delta-T mapping method based on Akiake information criteria and Gaussian process, a counterpart of the proposed method based on empirical wavelet transform and Gaussian process, as well as the proposed method based on empirical wavelet transform and long short-term memory neural network. The location errors and their variations of the proposed method were found far less than those of the other existing methods. The location accuracy could be further improved by increasing the grid resolution of pencil lead break pre-tests. The results demonstrated the feasibility and superiority of the proposed crack location method for field applications on large-scale complex structures like orthotropic steel plates.