In recent years, high accuracy nondestructive testing has been demanded. When the eddy current flaw signal is displayed as an image in two-dimensional eddy current testing, the flaw image becomes blurred in principle. In this study, the authors propose a method of applying a neural network to processing the eddy current testing signal. For the two-dimensional signal processing, the authors have restored the flaw image from blurred flaw signals of pancake coil and rotating uniform eddy current probe. The experiments of two-dimensional signal processing have shown that the neural network can restore a clear flaw image. It was also confirmed that the rotating uniform eddy current probe provides a better flaw image than a pancake coil.

抄録全体を表示