GPU-Based Hybrid Method for Electromagnetic Scattering of.pdfVIP

2nd International Conference on Electrical, Computer Engineering and Electronics (ICECEE 2015) GPU-Based Hybrid Method for Electromagnetic Scattering of Electrically Large Objects ZHU Yan-ju 1,a, XIE Shu-guo1 1School of Electronic and Information Engineering, Beijing University of Aeronautics and Astronautics, Beijing 100191, China azhuyanju1309@163.com Keywords: Compute Unified Device Architecture(CUDA); Graphical-Electromagnetic Computing (GRECO); Graphics Processing Unit (GPU); Modified Physical Optics (MPO). Abstract. A GPU-based parallel hybrid method is proposed to accelerate the electromagnetic scattering from electrically large complex structures in high frequency. The MPO and the GRECO hybrid method can simplify the computations of the diffraction fields of the wedges by modifying the surface-normal vector of the target to redefine the surface equivalent currents. By using GPU-based reduction operation, we design a fast parallel summing method for the hybrid integrals over illuminated triangles. The hybrid method is validated by comparing the numerical results with those obtained through the serial computing program on a central processing unit, as well as through FEKO method of moments, which showed good agreement. Introduction With the development of the stealth technology, electromagnetic scattering of the electrically large targets in high frequency is very important for the national defense, aerospace and remote sensing, etc.. To obtain the scattered field for the complex target, the computation and memory are in high demand. Then, asymptotic methods