三维FDTD亚网格技术的研究

发布时间：2018-05-11 02:09

本文选题：亚网格技术 + 时域有限差分法　；参考：《西安电子科技大学》2015年硕士论文

【摘要】：本文主要分析了两类亚网格技术的具体实现方式以及其性能的提升方案。作为一种简便有效的计算方法,时域有限差分方法(Finite-Difference-Time-Domain,FDTD)在计算目标电磁散射特性方面有着天然的优势,通常FDTD方法在模拟目标外观特征时采用的都是具有固定尺寸的均匀网格,但这种处理方式在面对具有精细结构的电大尺寸目标时,必然会带来内存的大量消耗和计算效率的下降。基于此,一种有效的处理方案—亚网格技术应运而生,即对目标主体部分采用较粗的网格,精细部分采用较细的网格,这样的网格剖分方式能将所有的资源都物尽其用,大大提升计算的性能。不同的粗细网格边界处理方法诞生了不同的亚网格技术,本文就首先分别详细介绍了基于波动方程和线性插值的两类亚网格实现原理,并给出了相应的散射与辐射算例,数值结果表明在这两种方法下,粗细网格边界都有着很低的反射率,程序稳定且结果正确有效。接着简要说明了几种常见的FDTD建模方法,包括简单几何建模、型值点法、三角面元,并重点分析了如何将亚网格技术同投影求交法结合起来实现亚网格复杂模型的建模,并给出相应的电磁散射算例表明该方法的正确性。为了进一步提升亚网格计算的性能,以适应大规模并行计算的现实需求,本文在最后探讨了基于MPI和OpenMP的亚网格并行实现方式,大大提升了亚网格技术的计算能力。
[Abstract]:In this paper, the implementation of two subgrid technologies and their performance enhancement schemes are analyzed. As a simple and effective calculation method, Finite-Difference-Time-Domain-FDTD (FDTD) has a natural advantage in calculating the electromagnetic scattering characteristics of a target. In general, the FDTD method uses uniform mesh with fixed size when simulating the external features of the target. However, in the face of electrically large size targets with fine structure, this method will inevitably lead to a large amount of memory consumption and a decrease in computational efficiency. Based on this, a kind of effective processing scheme, sub-grid technology, emerges as the times require, that is, the coarse grid is used for the main part of the target body, and the fine part is used for the finer grid. This kind of mesh generation method can make the best use of all the resources. Greatly improved computing performance. Different methods of coarse and fine mesh boundary processing have different subgrid techniques. In this paper, the realization principles of two submeshes based on wave equation and linear interpolation are introduced in detail, and the corresponding scattering and radiation examples are given. Numerical results show that both methods have low reflectivity and the program is stable and the results are correct and effective. Then several common FDTD modeling methods are briefly introduced, including simple geometric modeling, value point method, triangular plane, and how to combine subgrid technology with projection intersection method to realize the modeling of complex subgrid model is analyzed. An example of electromagnetic scattering is given to show the correctness of the method. In order to further improve the performance of sub-grid computing to meet the practical needs of large-scale parallel computing, this paper finally discusses the implementation of sub-grid parallel based on MPI and OpenMP, which greatly improves the computing capability of sub-grid technology.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：O441.4

【参考文献】