A data-driven approach based on wavelet analysis and deep learning for identification of multiple-cracked beam structures under moving load. (1st October 2020)

• Record Type: Journal Article
• Authors: Nguyen, Thanh Q.; Vuong, Luan C.; Le, Canh M.; Ngo, Nhi K.; Nguyen- Xuan, H.
• Is Part Of: Measurement. Volume 162(2020)
• Publication Date: 2020-10-01
• Subjects: Crack -- Wavelet analysis -- Deflection -- Static element -- Dynamic element -- Deep learning; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2020.107862
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: Deflection signal's static and dynamic elements using deep learning are devised. The static elements determine quantity and location of defects in structures. The dynamic elements evaluate growth rate of defects during structural operation. Original signals were analyzed directly using wavelet analysis. This study concurrently identifies all three criteria of quantity, location and growth rate. Abstract: This article aims to identify multiple cracks of a structure under the act of moving load by using wavelet analysis based on displacement signal. Among many researches on this topic, the current study presents three new salient points. Firstly, the breakthrough is the combination of two independent algorithms in wavelet analysis of deflection signal's static and dynamic elements using deep learning. The static elements allow determining quantity as well as location of defects in structures while the dynamic elements help to assess growth rate of defects during structural operation. Secondly, wavelet analysis is conducted based on deflection signal, which is rarely used but has a great deal of information proving the existence of cracks in a structure. This signal can also be applied to various types of structures such as plate, beam, bar and pivot. Thirdly, original signals are analyzed directly using wavelet analysis without any intermediate algorithm. This gives simpler and more accurate assessment of crack status, therefore, helps to increase sensitivity as well as data accuracy during identification process, especially for structures with multiple cracks. This study concurrently identifies all three criteria of defect assessment, namely: quantity, location and growth rate. In addition, the results have high potential for practical application to most structures.