A Study of Mesh Quality Indicators for Computational Electromagnetics

This paper presents a study of mesh quality indicators for computational electromagnetics (CEM) methods. Taking the electric field integral equation (EFIE) as example, five geometric-based mesh quality indicators are discussed. Based on the artificial neural networks, the conditional numbers of the impedance matrix can be linked to those indicators, where it could be thus predicted as output. More than 5000 meshing data are trained, and numerical examples show that the conditional number could be accurately fitted and employed to predicate the mesh quality. This study is useful for the mesh quality evaluation and optimization of the pre-processing in EM computational methods.