渦電流探傷試験の高度化のためのニューラルネットワークの適用可能性の検討　?ニューラルネットワークを応用した欠陥深さと長さの同定?

英字タイトル:
Feasibility study of neural network technology applied to advanced eddy current testing -Depth and length sizing of defects using neural networks-
著者:
周新武高木敏行内一哲哉（東北大）
発刊日: 2019年7月24日
公開日: 2021年9月30日

カテゴリ: 第16回

キーワードタグ:
Artificial intelligenceDefect sizingEddy current testingNDTNeural network

概要

This paper proposes a neural network to achieve automated data analysis target. A neural network, which is commonly used as an artificial intelligence technology, possesses excellent feature recognition and logistic regression ability, which are very important to implement automated data analysis. In this paper, the principle and characteristics of the neural network are presented. A neural network is established to discern the depth and length of slits automatically, and it is verified whether the ANN can work in ECT data analysis or not. According to the discussion and verification, it is evident that the trained neural network can accurately and efficiently offer quantitative analysis of defects.