Scholarlink

This work proposes an unsupervised Physics-Informed Neural Network (PINN) framework with Fourier feature encoding to solve partial differential equations (PDEs) describing magnetic field distributions under current excitation. A three-output network is designed to predict magnetic vector potential and magnetic field components, addressing media discontinuities. Two fundamental benchmark models are constructed to demonstrate the effectiveness of the method. Additionally, a relay model is introduced for validation, confirming the framework's accuracy and generalization in practical multi-domain electromagnetic problems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Physics-Informed Neural Networks for Solving Static Magnetic Problems

  • 会议论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年

查看全文
  • Youkang Hu,
  • Yuqiu Sun,
  • Shenglei He,
  • Tao Xu,
  • Lin Qin,
  • Wei Xu

摘要

This work proposes an unsupervised Physics-Informed Neural Network (PINN) framework with Fourier feature encoding to solve partial differential equations (PDEs) describing magnetic field distributions under current excitation. A three-output network is designed to predict magnetic vector potential and magnetic field components, addressing media discontinuities. Two fundamental benchmark models are constructed to demonstrate the effectiveness of the method. Additionally, a relay model is introduced for validation, confirming the framework's accuracy and generalization in practical multi-domain electromagnetic problems.