Intelligent Global Error Estimation for Mesh Discretization in Method of Moments

Traditional methods are insufficient for the effective assessment of mesh quality prior to performing electromagnetic field calculations using the method of moments (MoM). To address this problem, we propose an intelligent global error estimation algorithm for mesh discretization in MoM. The algorithm utilizes a neural network to effectively capture and model the complex nonlinear relationship between point cloud-based mesh data and global mesh discretization errors. To train the neural network, we developed a custom electromagnetic scattering error dataset and proposed a technique for encoding unstructured mesh as point cloud data to effectively capture mesh features. Experimental results demonstrate the proposed neural network framework's efficacy in accurately assessing global errors in electromagnetic scattering.