A Unified Frequency-Wise Electromagnetic Modeling Framework With Improved Generalizability and Scalability

Electromagnetic (EM) modeling accelerates the design and optimization of EM structures by predicting their frequency-dependent features such as <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$|S_{11}|$</tex-math> </inline-formula>. Current methodologies constrain each model to a fixed frequency range at specified points, thereby necessitating the training of separate models for different frequency conditions, which limits generalizability and scalability in real-world applications. This article introduces a novel EM modeling framework with improved generalizability and scalability. It significantly enhances modeling accuracy for frequency conditions not previously encountered. The model can be easily scaled up by incorporating new data with variable frequency conditions, thereby further improving the generalizability. Our method incorporates a frequency-wise learning strategy that enforces a robust understanding of the frequency-dependent working mechanism of EM structures. We demonstrate the effectiveness of our approach through multiple implementations with variable frequency conditions. The comparative results validate the improved generalizability and scalability, showcasing its potential to simplify and enhance EM design processes.