应用GPU加速结合压缩传感技术解宽角度电磁散射问题

Application of GPU Acceleration Combined with Compressive Sensing to Electromagnetic Scattering Problem over Wide Angle

本文基于矩量法(MOM)中阻抗矩阵与入射角度的不相关性,提出将压缩传感(CS)技术引入电磁散射问题的方法,以达到在不降低计算精度的前提下减少计算时间的目的。该方法结合了压缩感知和计算统一设备架构(CUDA)技术,通过利用图形处理器(GPU)的并行计算解决了宽角度的电磁散射问题。数值计算结果表明,该方法与传统计算方法相比,在保持精度的同时,大大降低了计算所需的时间。

Based on the feature that impedance matrix was independent from the incident angle in the meth od of moments （MOM）, this paper proposes a method that brings compressive sensing （CS） technique into electromagnetic problem in order to reduce the computing time without sacrificing much accuracy. This method combines compressive sensing with compute unified device architecture （CUDA） technique to solve the electromagnetic scattering problem over wide angle by parallel computing in graphic processing unit （GPU）. It is shown in numerical experiments that comparing with traditional algorithm the proposed method saves much computing time, and the accuracy can also be accepted.