畳み込みニューラルネットワーク用いたＥＣＴ信号からのキズ深さ同定

抄録

Eddy current testing (ECT) is a nondestructive inspection method for detecting cracks and defects in conductive materials such as thin heat transfer tubes of steam generator. ECT applies inverse problem analysis for crack shape estimation, but in many cases requires large CPU time and memory. In this study, an application of convolutional neural network (CNN), which is one of deep learning models, was proposed and showed the possibility of high-speed estimation of crack depth.