Electromagnetic Field Analysis using Physics Informed Neural Network considering Eddy Current

被引：0

作者：

Han, Ji-Hoon ^{[1
]}

Park, Jong-Hoon ^{[1
]}

Song, Seung-Min ^{[1
]}

Hong, Sun-Ki ^{[2
]}

机构：

[1] Hoseo Univ, Dept Informat Control Engn, Asan, South Korea

[2] Hoseo Univ, Dept Syst & Control Engn, Asan, South Korea

来源：

2024 IEEE 21ST BIENNIAL CONFERENCE ON ELECTROMAGNETIC FIELD COMPUTATION, CEFC 2024 | 2024年

基金：

新加坡国家研究基金会;

关键词：

PINN; Electromagnetic analysis; Deep learning; Eddy current; Transfer learning;

D O I：

10.1109/CEFC61729.2024.10585621

中图分类号：

TP39 [计算机的应用];

学科分类号：

081203 ; 0835 ;

摘要：

The exploration of PINN (Physics Informed Neural Network) in research is still in its nascent stages globally, with a notable dearth of studies focusing on electromagnetic field analysis. In response to this gap, this paper introduces a novel approach for eddy current analysis employing a transfer learning-based discrete differentiation method within the framework of a PINN. While discrete differentiation methods offer the advantage of low model complexity and rapid analysis, they encounter challenges in eddy current analysis due to the need for learning at each time step. This paper addresses these challenges through the application of transfer learning techniques. Our findings demonstrate that the proposed method significantly reduces the total analysis time in time-dependent scenarios compared to traditional Finite Element Method (FEM) approaches.

引用

页数：2