[23p-E203-1~12] 11.3 Critical Current, Superconducting Power Applications

[23p-E203-9] Automatic Detection of Local Defects in Long HTS Tapes by Image Recognition Using Deep Learning in Reel-to-Reel Magnetic Microscopy

〇Takanobu Kiss¹, Kazutaka Imamura¹, Natthawirot Somjaijaroen¹, Zeyu Wu¹, Kohei Higashikawa¹ (1.Kyushu Univ.)

Keywords:HTS tape, image recognition, magnetic microscopy

As a requirement of high temperature superconducting wire, the uniformity of local I_c has been recognized more and more its importance in recent years. In the scanning magnetic microscopy, it has the advantage that a local defect in the tape plane can be detected in two dimensions. In this study, reel-to-reel high-speed scanning Hall-probe magnetic microscopy (RTR-SHPM) was carried out for a 100-m-long Bi-2223 taped wire, and by applying deep-learning based image recognition to the observed image by the RTR-SHPM, we succeeded in detecting local defects automatically without depending on visual observation.