文献类型：期刊文献

英文题名：A novel baseline-free defect detection and localization method of welded steel plate based on reciprocity loss

作者：He, Tianhao[1]; Xing, Gailan[1]; Li, Yong[1]; Li, Qinfei[1]; Zhou, Shaoping[1]

机构：[1] Key Lab of Pressure Systems and Safety, East China University of Science and Technology, Shanghai, 200237, China

年份：2020

卷号：32

期号：2

外文期刊名：Measurement Science and Technology

收录：EI(收录号：20205109635211)

语种：英文

外文关键词：Structural health monitoring - Surface waves - Welding - Defects - Ultrasonic testing - Plates (structural components) - K-means clustering - Nondestructive examination

摘要：In recent years, ultrasonic Lamb waves have been widely applied to the field of non-destructive testing of plate-like structures as they have outstanding advantages, such as low attenuation, high sensitivity, and wide detection range. Current studies about defect-detection of plate-like structures using Lamb waves mostly focus on non-weld plate-like structures, and defect-detection methods are based on baseline data. This paper proposes a novel baseline-free damage inspection method of welded plate-like structures, which is based on the principle of reciprocity loss and combines the OPTICS and K-means intelligent clustering algorithms to achieve accurate defect localization. In order to verify the location accuracy of the clustering defect localization algorithm, this paper performs comparative experiments between the ellipse imaging algorithm and the clustering algorithm, which use baseline data as health signals. The comparative experimental results show that the single-defect location accuracy of the clustering algorithm is greatly improved compared with the traditional ellipse algorithm. Moreover, in order to verify the validity and feasibility of the baseline-free method, this paper applies this method to obtain characteristic signals and combines the clustering algorithm to locate both single-defect and double-defects. The experimental result of baseline-free method shows that this method can successfully detect and locate multiple defects, which gets rid of the dependence of baseline data. ? 2020 IOP Publishing Ltd.