Classification of Elastic Wave for Non-Destructive Inspections Based on Self-Organizing Map

• Published in: Sustainability Vol. 15; no. 6; p. 4846
• Main Authors: Nakamura, Katsuya, Kobayashi, Yoshikazu, Oda, Kenichi, Shigemura, Satoshi
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.03.2023
• Subjects: Algorithms; Analysis; Classification; Elastic waves; Localization; Neurons; Nondestructive testing; Picking; Self organizing maps; Signal classification; Sustainability; Time measurement; Tomography; Wave propagation; Wave velocity; Waveforms
• ISSN: 2071-1050; 2071-1050
• DOI: 10.3390/su15064846

An arrival time of an elastic wave is the important parameter to visualize the locations of the failures and/or elastic wave velocity distributions in the field of non-destructive testing (NDT). The arrival time detection is conducted generally using automatic picking algorithms in a measured time-history waveform. According to automatic picking algorithms, it is expected that the detected arrival time from low S/N signals has low accuracy if low S/N signals are measured in elastic wave measurements. Thus, in order to accurately detect the arrival time for NDT, the classification of measured elastic waves is required. However, the classification of elastic waves based on algorithms has not been extensively conducted. In this study, a classification method based on self-organizing maps (SOMs) is applied to classify the measured waves. SOMs visualize relation of measured data wherein the number of classes is unknown. Therefore, using SOM selects high and low S/N signals adequately from the measured waves. SOM is validated on model tests using the pencil lead breaks (PLBs), and it was confirmed that SOM successfully visualize the classes consisted of high S/N signal. Moreover, classified high S/N signals were applied to the source localization and it was noteworthy that localized sources were more accurate in comparison with using all of the measured waves.