研究目的
To analyze the electromagnetic scattering and radiation problems using an efficient time domain simulation algorithm based on the discontinuous Galerkin time domain (DGTD) method and GPU parallelization acceleration technique.
研究成果
The DGTD method, when accelerated with GPU, provides a reliable and fast algorithm for solving a variety of electromagnetic problems, demonstrating significant speedup over CPU-based methods. It combines the advantages of FDTD and FEM methods, offering high parallel efficiency and adaptability to unstructured meshes.
研究不足
The study is limited by the computational resources (specific GPU and CPU models used) and the complexity of the meshing for very large or intricate structures. The accuracy and performance are dependent on the polynomial order and mesh quality.
1：Experimental Design and Method Selection:
The study employs the DGTD method for solving Maxwell's equations, focusing on its implementation on GPU platforms using CUDA for acceleration. The methodology includes local-element differentiation, surface integrals, and Runge-Kutta time-integration assembly.
2：Sample Selection and Data Sources:
The benchmarks include a dielectric sphere and a PEC almond for RCS prediction, and a helix antenna for directivity pattern analysis.
3：List of Experimental Equipment and Materials:
NVIDIA GTX780M GPU, Intel Xeon i-7 4990MQ CPU, and various computational domains with specified mesh sizes.
4：Experimental Procedures and Operational Workflow:
The process involves mesh generation, setting up the DGTD method on GPU, applying excitation, and calculating electromagnetic fields and parameters like RCS and directivity patterns.
5：Data Analysis Methods:
Performance metrics include simulation time, memory usage, and accuracy comparisons with analytical and numerical methods like FIT, MoM, and FEM.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
