首页 | 本学科首页   官方微博 | 高级检索  
     

基于结构网格的增升构型失速特性大规模并行数值模拟
引用本文:李中武,谭伟伟,周磊. 基于结构网格的增升构型失速特性大规模并行数值模拟[J]. 数据与计算发展前沿, 2013, 4(5): 51-57
作者姓名:李中武  谭伟伟  周磊
作者单位:中国西安航空计算技术研究所, 陕西 西安 710019
摘    要:针对民机增升构型失速特性的数值模拟,我们基于贪婪负载平衡算法的剖分工具对多块结构网格进行区域分割,在某新型超级计算机系统上完成求解软件的移植、优化和测试,采用 2 亿量级的计算网格开展大规模并行计算研究,测试完成了万核级负载平衡的网格区域分割,实现了增升构型失速特性的 4 096 核数并行计算,并行效率达到 50% 以上,提高了工程应用中对复杂流动现象的数值模拟能力。数值模拟结果加深了对增升构型失速流动机理的理解,可以为增升装置设计优化提供有意义的参考依据。

关 键 词:结构网格  贪婪负载平衡  大规模并行  增升构型  失速特性  
收稿时间:2013-06-26

Massively Parallel Numerical Simulation of Stall Characteristic on High-Lift Configuration Based on Structured Grids
Li Zhongwu,Tan Weiwei,Zhou Lei. Massively Parallel Numerical Simulation of Stall Characteristic on High-Lift Configuration Based on Structured Grids[J]. Frontiers of Data & Computing, 2013, 4(5): 51-57
Authors:Li Zhongwu  Tan Weiwei  Zhou Lei
Affiliation:Aeronautical Computing Technique Research Institute, Xian, Shanxi 710019, China
Abstract:For the numerical simulation of stall characteristic on high-lift Configuration of an aircraft, we developed a CFD code, and implemented it at our supercomputing system. The grid of high-lift configuration with 0.2 billion nodes is split by WiseLBS which is based on Greedy load balancing algorithm. The original computational grid is split into 18435 blocks for insuring 10240 CPUs load balancing. Massively parallel numerical simulation was carried by using 4096 CPUs, and the parallel efficiency achieved is more than 50%. This improved our capability of dealing with the complex fluid-field. The simulation results has deepened our understanding about stall mechanism on high-lift configuration, and can be used as references for aerodynamic design optimization.
Keywords:structured grid  greedy load balancing  massively parallel  high-lift configuration  stall characteristic  
点击此处可从《数据与计算发展前沿》浏览原始摘要信息
点击此处可从《数据与计算发展前沿》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号