This paper presents the surrogate model for magnetic force analysis which is constructed through physics-aware machine learning. The proposed surrogate model is expected to be useful for modeling, optimal design, uncertainty quantification and data assimilation because of high computing efficiency in construction. In the proposed method, the physical property—here, evaluated approximately by a magnetic circuit—is embedded in machine learning. The proposed surrogate model is shown to provide accurate prediction of magnetic force even with small training data.
Hajime Igarashi (Wed,) studied this question.