کنفرانس مهندسی برق ایران

صفحه اصلی / سی و سومین کنفرانس بین المللی مهندسی برق

Physics-Based Learning Approach Using Self-Terms for Electromagnetic Scattering in Multi-Object Scenarios

نویسندگان :

Arefeh Nikdast¹ Amir ahmad Shishegar²

1- دانشگاه صنعتی شریف 2- دانشگاه صنعتی شریف

کلمات کلیدی :

electromagnetic scattering،forward scattering problem،machine learning،physics-informed learning method

چکیده :

This paper presents a novel learning framework for solving electromagnetic scattering problems. While traditional full-wave numerical methods like the Method of Moments (MOM) are effective, they are computationally intensive. Machine learning algorithms offer real-time solutions but are often constrained by their training data. To enhance the performance of these learning-based approaches and avoid black-box models, we integrate physical knowledge of self-terms of scattering into our framework. Our model, designed to predict the interactions between objects, is particularly effective in scenarios involving multiple scatterers. Tested on various dielectric distributions, it achieves accuracy comparable to conventional methods and successfully addresses nonlinear problems where the Born series fails to converge. This framework seamlessly integrates physics-based insights with data-driven methods, facilitating more complex simulations.