Physics-Informed Neural Networks with Dynamic Sampling Method for Solving Rectangular Waveguide Problems

In this work, a novel approach for solving electromagnetic problems is proposed by using physics-informed neural networks and dynamic point sampling. The proposed method enhances the efficiency and accuracy of electromagnetic simulations within the PINN framework by adapting the distribution of residual points during the training process. To demonstrate the effectiveness of the proposed method, a rectangular waveguide is considered. The final results verify the validity of the proposed method in solving electromagnetic problems.