汇报时间：2023年8月29日（星期二）下午15:00

汇报地点：创新港 2号巨构 2-5F-036会议室

汇报人：黄婧

国际会议信息

会议名称：The 76th IIW Annual Assembly and International Conference on Welding and Joining

会议时间：July 16-21，2023

会议地点：Marina Bay Sands Convention Centre, Singapore

会议简介：The 76th Annual Assembly of International Institute of Welding (IIW) and International Conference on Welding and Joining aims to provide a platform for knowledge exchange and networking among scientists, researchers and industry experts in the field of welding and joining. Focus areas of the 18 Technical Working Units can be divided generally into Processes, Structural Integrity and Industrial Applications, and Human Factors. The Working Units operate as think tanks and engines for technical progress, focusing on current challenges in industry and research, and developing technical output to proactively support these needs. One of the tremendous strengths of the IIW is the opportunity for seamless cooperation between the different focus groups, drawing together a broad spectrum of relevant experts from around the world to focus on particular issues.

参会论文信息

Title：Acoustic Emission-based Weld Crack Leakage Monitoring with ATO Incremental Learning

Author：Jing Huang, Zhifen Zhang, Rui Qin, Guangrui Wen, Wei Cheng, Xuefeng Chen

Abstract: The leakage of pressure pipe weld cracks formed under the action of high-stress concentration and external alternating loads is a major hidden danger in the safe service of nuclear power ships. Micro structure difference and macro dynamic expansion of weld cracks cause the leakage state to drift continuously with the change of time and environment. To address above problems, this paper proposes an incremental learning strategy with adaptive threshold optimization ability. Firstly, an affinity threshold is introduced to improve its stability when separating overlapping clusters and facing different input signals. Secondly, a depth-first search algorithm is used to label the category of neurons identified by the Enhanced self-organizing incremental neural network. On this basis, the RBF neural network is trained to obtain the class labels, so as to realize online increment learning. Finally, the validity of the proposed method is verified by three well-thought-out circumferential, axial and curvilinear pipeline weld cracks. In addition, the effects of parameters on the accuracy of the labelling results are also discussed in detail. The proposed method outperforms the other three methods for all four performance indicators, enabling automatic and adaptive updating of the pipeline weld crack leakage monitoring model based on new data.

欢迎有兴趣的同学届时参加。