A Two-Step Method for Efficient Simulation of Eddy Currents and Magneto-Statics in Heterogeneous Domains
A two-step method is presented that efficiently couples eddy current effects in a part of the computational domain with magneto-static effects in the remaining domain, enabling accurate and computationally efficient simulations of electromagnetic phenomena.
맥스웰 방정식의 비학습 그래프 신경망 메시지 전달을 통한 해결
그래프 신경망의 고유한 특성을 활용하여 맥스웰 방정식의 이산화된 형태를 직접 표현하고 이를 통해 효율적으로 전자기파 전파를 모사할 수 있다.
Efficient Numerical Solution of Maxwell's Equations using Non-Trainable Graph Neural Networks
A graph neural network (GNN) with static and pre-determined edge weights can efficiently solve Maxwell's equations with the same accuracy as conventional computational electromagnetics methods, while providing significant computational time gains.
Adaptive Mesh Refinement for Capturing Fine-Scale Features in Leaky Modes of Microstructured Optical Fibers
The paper presents an adaptive algorithm based on dual-weighted residual error estimation to efficiently capture fine-scale features in the leaky modes of complex microstructured optical fibers. The algorithm automatically detects and refines regions with high error contributions to accurately resolve the critical mode characteristics, including confinement losses.
© 2024 by Linnk AI