Impact of different node distributions on the meshless Radial Point Interpolation Method in time-domain electromagnetic simulations

Abstract

This paper investigates the effect of different node distributions on the accuracy of electromagnetic simulations performed with meshless methods. As a test case, a rectangular waveguide truncated with perfectly-matched layers is simulated using the three-dimensional meshless Radial Point Interpolation Method (RPIM) in the time domain. For the discretization of the geometry, different strategies of node distribution are utilized, namely a uniform grid distribution, a cylindrical distribution, and disturbed grid distributions with random displacements amounting to 5% and 10% of the grid average node distance. All distributions are generated with a similar node density, and the results are compared in terms of phase and amplitude of the propagating wave. The application of RPIM in all cases demonstrates similar levels of error, which indicates the robustness of this meshless algorithm with respect to different node distribution strategies.