Determination of the Presence or Absence of Defect for Laser Ultrasonic Visualization Testing Using Transfer Learning

LUVT (Laser Ultrasonic Visualization Testing) is known as an effective experimental technique for detecting surface defects. It can visualize ultrasonic wave propagation on a defect surface. Inspectors typically determine the presence or absence of defects by visually observing the propagation of ultrasonic waves and checking for the generation of scattered waves. If AI is used to automate this visual judgment made by inspectors, it could reduce the workload of the inspectorate. In this study, we attempt to automate the LUVT inspection by incorporating transfer learning into deep learning, which is the basis of AI. The effectiveness and other aspects of the proposed method are discussed by presenting several defect detection results.