紀要論文

深層学習を用いた永久磁石磁化推定手法に関する基礎検討

A FUNDAMENTAL STUDY ON ESTIMATION METHOD FOR MAGNETIZATION DISTRIBUTION IN PERMANENT MAGNET USING DEEP LEARNING

jpn

24368083

https://doi.org/10.15002/00025341

法政大学大学院理工学研究科

法政大学大学院紀要. 理工学研究科編

63

1

7

2022-03-24

Version of Record

ヘッダー部分に誤記 ; (誤) 法政大学大学院紀要理工学・工学研究科編 (正) 法政大学大学院紀要理工学研究科編

Since the performance of a permanent magnet synchronous motor depends on the magnetization distribution in the permanent magnet, it is essential to estimate the magnetization distribution at the previous stage of the practical design. When the magnetization is nondestructively identified by the electron back-scattered diffraction patterns or X-ray diffraction, the estimation time becomes so long. In contrast, an effective method, which is based on the minimization of the measured and calculated fluxes, has been proposed. The method based on mathematical programming has been proposed as a nondestructive estimation method. Although the method can estimate the magnetization distribution fast, this method has the possibility to derive unphysical magnetization distribution when some parameters are not adequately adjusted. Subsequently, to suppress the non-physical magnetization, an estimation method for magnetization distribution sing deep learning is proposed to suppress the nonphysical magnetization.

Article