Speaker
Description
Spatial homogeneity is one of the most important requirements of REBCO coated conductors (CCs) for practical applications. We demonstrated automatic detection of local obstacles in a PLD processed long length REBCO CC by introducing image classification based on machine learning into reel-to-reel scanning Hall-probe magnetic microscopy [1, 2]. This allows us to classify magnetic images including obstacles from thousands of images taken from long CC tape. In this study, we further explore this approach for detecting the obstacle more directly as an object, which enables us to analyze the features of the obstacles including size and position in the REBCO CCs. Influence of measurement conditions on the detection was also examined. We found that the spatial resolution is the most important conditions for the detection, and it should be selected the same or similar value between training data and input data so as to realize reliable detection. When we trained an object detection model by using data set from PLD processed REBCO CC from a company, the model can detect obstacles with similar performance even in the other PLD processed REBCO CC from different manufacture. We also studied the influence of these obstacles on the critical current fluctuation of the CCs from the viewpoint of statistics. This method allows us to analyze spatial homogeneity more in detail, which is hardly possible by the conventional methods, and can compare the quality of the CCs and/or clarify the influence of the process conditions. Namely, it can lead useful insights for both CC users and manufacturers.
Acknowledgements: This work was supported by Moonshot R&D - MILLENNIA Program Grant Number JPMJMS24A2 and JSPS KAKENHI Grant Number JP24H00320, JP23K13368.
[1] K. Higashikawa et al., SUST vol. 33, No. 6, June 2020, Art no. 064005, doi: 10.1088/1361-6668/ab89ef
[2] N. Somjaijaroen; T. Kiss; K. Imamura; K. Higashikawa, TEEE TAS, vol. 32, no. 6, pp. 1-4, Sept. 2022, Art no. 6601504, doi: 10.1109/TASC.2022.3156541
