Abstract <p>The problem of plane electromagnetic wave diffraction by thin perfectly conducting screens with perturbed geometry is considered. The numerical solution based on the method of moments is developed. The surface current is approximated with Rao–Wilton–Glisson (RWG) basis functions. The numerical solution is obtained with a fast domain-cutting algorithm. This allows to significantly reduce computational costs for multiple screen configuration changes by using a pre-calculated moment matrix. The parallel algorithm using computing on graphics processing units (GPUs) with CUDA technology is implemented. The numerical results illustrating the influence of the shape, size and position of cutouts in the screen on the distribution of surface currents are presented. A comparative analysis of the running time of sequential and parallel algorithms is presented. It is shown that the proposed algorithm provides a significant acceleration and can be effectively used for large series of computational experiments and analysis of surface current behavior on complex shape screens.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Study of Surface Currents on Thin Screens with Perturbed Geometry: A Method of Moments Approach with GPU Acceleration

  • D. Kh. Giniyatova

摘要

Abstract

The problem of plane electromagnetic wave diffraction by thin perfectly conducting screens with perturbed geometry is considered. The numerical solution based on the method of moments is developed. The surface current is approximated with Rao–Wilton–Glisson (RWG) basis functions. The numerical solution is obtained with a fast domain-cutting algorithm. This allows to significantly reduce computational costs for multiple screen configuration changes by using a pre-calculated moment matrix. The parallel algorithm using computing on graphics processing units (GPUs) with CUDA technology is implemented. The numerical results illustrating the influence of the shape, size and position of cutouts in the screen on the distribution of surface currents are presented. A comparative analysis of the running time of sequential and parallel algorithms is presented. It is shown that the proposed algorithm provides a significant acceleration and can be effectively used for large series of computational experiments and analysis of surface current behavior on complex shape screens.