딥러닝과 테라헤르츠 기술을 이용한 폴리머 배관 결함 검출에 관한 연구

초록

The terahertz time-domain spectroscopy (THz-TDS) system and convolutional neural network (CNN) algorithm were used to detect a defect in a polymer tube. The THz scanning data of the normal and defective polymer tubes were obtained from the THz-TDS transmission mode. The amplitude of the THz wave transmitted by the crack of the polymer tube was decreased by scattering. It was confirmed that the crack in the polymer tube in the THz scanning image can be classified by the shading difference of the pixel. The THz image data were amplified for CNN deep learning using an augmentation technique, and the amplified THz image data were learned by the CNN deep learning algorithm by dividing classes into normal and defect datasets. As a result of deep learning, the CNN model can detect a crack in a polymer tube with an accuracy of 95% or more. Finally, it was confirmed that a defect in a polymer tube can be inspected using a noncontact and nondestructive terahertz inspection method with the CNN deep-learning algorithm.

키워드