现代应用物理2025，Vol.16Issue(1)：170-174,211,6.DOI:10.12061/j.issn.2095-6223.202412036
基于傅里叶神经算子的电磁智能计算研究
Research on Electromagnetic Intelligent Computation Based on Fourier Neural Operator
摘要
Abstract
In the research of electromagnetic applications,solving Maxwell's equations is a crucial step.Deep learning(DL),as an emerging and promising technology,with its powerful data-driven modeling ability,has gradually become an effective approach to solving partial differential equations(PDE).Electromagnetic problems can be addressed by realizing the direct mapping of input data to output through neural networks.In this paper,the Fourier neural operator(FNO)directly learns the mapping from parameters to the solutions of PDE and discretizes PDE in the Fourier space for their resolution.The FNO is applied to Maxwell's equations to solve complex electromagnetic problems in the two-dimensional case.Compared with traditional numerical algorithms,FNO has a computing speed that is two orders of magnitude faster,providing a new perspective and tool for handling complex electromagnetic problems in the two-dimensional context.关键词
傅里叶神经算子/偏微分方程/时域有限差分法Key words
Fourier neural operator/partial differential equation/finite-difference time-domain分类
数理科学引用本文复制引用
赵洁,周东华,冯健,方明,黄志祥..基于傅里叶神经算子的电磁智能计算研究[J].现代应用物理,2025,16(1):170-174,211,6.基金项目
国家自然科学基金资助项目(62271004,62231003,62301001,62401006) （62271004,62231003,62301001,62401006）
安徽省自然科学基金资助项目(2408085y033,2023z04020018,202423h08050007) （2408085y033,2023z04020018,202423h08050007）
安徽省教育厅资助项目(2023AH030001) （2023AH030001）
