Artificial neural network interpolation for magnetic field mapping in an air-core HTS quadruple magnet

Abstract

An air-core high-temperature superconducting (HTS) quadruple magnet is currently being designed and fabricated at the Applied Superconductivity Laboratory, Yonsei University, in collaboration with Uiduk University. It is composed of eight double-pancake racetrack coils, each wound with 4-mm REBCO tapes manufactured by SuNAM Co., Inc. The Metal-as-Insulation technique will be used to reduce no-insulation charging delays. The gradient, uniformity and effective length of the initially-designed air-core quadruple magnet are 3.1 Tm−1, below 0.5%, and 209.0 mm, respectively. However, the magnet to be fabricated will be different from the designed magnet owing to manufacturing uncertainties. Therefore, magnetic field measurements must be performed in the volume of interest of the magnet. As it is impossible to measure the magnetic field at all points, it is crucial to obtain a mapping algorithm that can calculate the magnetic field in the volume of interest based on the measured field at several points. Herein, a mapping method using the discrete Fourier transform and artificial neural network interpolation was presented. To verify the mapping method, a mapping simulation of the designed magnet was performed and analyzed.