径向基函数在动网格中的应用及可并行性研究  

Applications of Radial Basis Functions in Dynamic Mesh and Its Parallelizability

作　　者：马文鹏[1] 陆忠华[1] 胡晓东[1] 

机构地区：[1]中国科学院计算机网络信息中心超级计算中心,北京100190

出　　处：《科研信息化技术与应用》2012年第5期4-12,共9页E-science Technology & Application

基　　金：国家高技术研究发展计划(863计划)(2012AA01A304);国家自然科学基金(91130019)

摘　　要：径向基函数广泛应用于网格变形、气动外形优化设计、网格优化等领域。近年来,基于径向基函数的动网格技术得到了深入的研究和广泛的应用。本文结合计算流体力学和高性能计算的应用背景,从径向基函数对网格的变形质量和变形效率进行了总结和进一步研究:在网格变形方面,重点对比了不同基函数对同一网格运动变形能力和同一基函数对不同网格运动的适应能力;在网格变形效率方面,分析了算法在计算和存储的瓶颈所在,考虑了OpenMP和GPU这两种共享内存的加速方式,得到较好加速比。最后,分析了当网格规模增大时,动网格在分布式计算和存储模型(MPI)下的处理方法。Radial basis functions are widely used for mesh deformation, aerodynamic shape optimization, grid optimization, etc. Recently, dynamic mesh techniques based on radial basis functions are widely used and much more attention is paid to them. This paper summaries the quality and efficiency of mesh deformation with the applications of radial basis functions on both CFD （Computational Fluid Dynamics） and HPC （High Performance Computing）. In addition, a further research is conducted on the following two aspects： the abilities to satisfy the same mesh motion among different radial basis functions and the abilities to satisfy different mesh motion using one radial basis function; computation and storage bottleneck are analyzed and then parallel solutions based on shared memory models including OpenMP and GPU are considered and a good speed up is obtained. Finally the solution to large-scale dynamic mesh on the conditions of distributed computing and storage model （MPI） is discussed.

关 键 词：径向基函数 径向函数 动网格 网格变形 并行 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]