基于多块同位网格对管路流场的分区并行计算  

Parallel computing of pipeline flow based on multi-block collocated grid

作　　者：蒋光彪[1,2] 钟海[1,2] 殷水平[1,2] 陈胜铭[1,2] JIANG Guangbiao ZHONG Hai YIN Shuiping CHEN Shengming(College of Civil Engineering and Mechanics, Central South University of Forestry and Technology, Changsha, ttunan 410004, China Hunan Provincial Key Laboratory of Engineering Rheology, Central South University of Forestry and Technology, Changsha, Hunan 410004, China)

机构地区：[1]中南林业科技大学土木工程与力学学院,湖南长沙410004 [2]中南林业科技大学工程流变学湖南省重点实验室,湖南长沙410004

出　　处：《江苏大学学报（自然科学版）》2017年第5期536-543,共8页Journal of Jiangsu University：Natural Science Edition

基　　金：湖南省自然科学基金资助项目(14JJ2013);湖南省土木工程重点学科建设项目;中南林业科技大学人才引进项目(104-0096)

摘　　要：在基于多块同位网格和集群以突扩管路系统内部流场为算例开展的分区并行计算研究中,利用SIMPLE算法实施了可自定义分区的并行计算方法,分析了不同雷诺数和突扩比对突扩管内流动结构的影响,重点探讨了基于自定义分区的不同分区方式及核数工况下集群的并行计算效率,进一步分析负载对计算效率的影响.结果表明,网格数、核数和分区方式对集群的计算效率都有很大的影响,并且每个分区所包含的网格数越均匀即负载越均衡,并行计算效率越高.The partition parallel computing of internal flow for a sudden expansion pipe system division was conducted based on multi-block collocated grid and cluster. By SIMPLE algorithm,the custom partition method of parallel computing was implemented,and the effects of Reynolds number and sudden enlargement ratio on sudden expansion pipe flow structure were analyzed. The cluster parallel computing efficiency was detailedly discussed based on the different custom partition mode and core numeber. The influence of load on calculation efficiency was also analyzed. The results show that grid number,cores and partition way have great influence on cluster computing efficiency. When each grid partition contains the similar grid numbers or with equal load,the high parallel computing efficiency can be achieved.

关 键 词：多块同位网格 自定义分区 并行计算 突扩管路流动 CFD 

分 类 号：Q242.1[生物学—细胞生物学]