Analysis of Two-Dimensional Electromagnetic Scattering Problems Based on Generative Adversarial Network and Isogeometric Boundary Element Method

In this paper, we propose an efficient computational algorithm that integrates generative adversarial network (GAN) with the isogeometric boundary element method (IGABEM) to enhance the solution efficiency for electromagnetic scattering problems. First, based on conventional boundary element method (BEM), nonuniform rational B-splines (NURBS) are introduced and combined with classical electromagnetic scattering formulations to derive new scattering computation equations, thereby establishing the IGABEM framework. By precisely describing the geometric model using NURBS, geometric accuracy is effectively ensured, which in turn improves the accuracy of numerical computations. Second, a database is created using the IGABEM computation results for training the GAN, which is then employed to predict electromagnetic scattering problems. Finally, several numerical examples are provided to validate the accuracy of the IGABEM method and assess the reliability of the GAN predictions, with electromagnetic scattering data for selected models presented to further demonstrate the effectiveness of the proposed approach.