Fast Analysis and Design for 3D-Structured Magnetic Components Using Surrogate Model from Transfer Learning

This paper presents a new fast method to generate an accurate surrogate model for 3D-structured magnetic components using transfer learning (TL). By using the proposed TL method, the small amount of the training datasets obtained from the 3D finite element analysis is required for the accurate surrogate model. In this study, the surrogate models are constructed by the proposed and conventional models, and then the accuracy and number of the training datasets of the surrogate models are compared between two models to validate the effectiveness of the proposed method.