汽车碰撞有限元仿真的并行计算及其性能研究  

Finite Element Simulation and Performance Study of Vehicle Crash Based on Parallel Computing

作　　者：亓文果[1] 金先龙[1] 张晓云[1] 李渊印[1] 李根国[2] 

机构地区：[1]上海交通大学高性能计算中心,上海200030 [2]上海超级计算中心,上海200030

出　　处：《系统仿真学报》2004年第11期2428-2431,共4页Journal of System Simulation

基　　金：获国家自然科学基金资助(60174023);教育部博士点基金资助(20010248008)

摘　　要：并行计算是提高汽车碰撞有限元仿真的计算速度和求解能力的有效途径。文章简要介绍了汽车碰撞有限元仿真并行计算的区域分解方法,给出了处理接触时的并行计算流程,分析了并行求解过程以及主要时间开销。在此基础上,研究了两种硬件体系结构和不同处理器数目对计算效率的影响。最后研究了不同的区域分解策略对整车碰撞计算效率的影响。研究结果对最终用户选择合适的硬件平台和计算软件具有指导意义。Parallel computing is an efficient approach to speed up and enhance the solving ability of finite element simulation of automobile crash. In this paper, domain decomposition for parallel computing is introduced simply, and the algorithm flowchart including contact is drawn, and the parallel computing process and primary overhead are analyzed. Furthermore, effect of two hardware platforms and the number of processors on computing efficiency are studied. At last, effect of different domain decomposition strategy on automobile crash computing efficiency is presented. To end users, research result is a guidance to choose appropriate hardware platform and computing software.

关 键 词：并行计算 碰撞 有限元仿真 区域分解 

分 类 号：U467.13[机械工程—车辆工程]