首页 | 本学科首页   官方微博 | 高级检索  
     检索      

结构静力有限元分层并行计算方法
引用本文:苗新强,金先龙,丁峻宏.结构静力有限元分层并行计算方法[J].力学学报,2014,46(4):611-618.
作者姓名:苗新强  金先龙  丁峻宏
作者单位:1.上海交通大学机械系统与振动国家重点实验室, 上海 200240
基金项目:国家高技术研究发展计划(2012AA01A307);国家自然科学基金(11072150,61073088)资助项目~~
摘    要:根据分布式存储并行计算机的体系结构特点,提出了一种结构静力有限元分层并行计算方法. 该方法在两级分区两次缩聚策略的基础上不仅实现了大量数据的分布式存储,提高了数据的内存访问速率;而且实现了计算过程的三层并行,有效提高了通信效率;此外,它还进一步降低了界面方程的规模,大幅度减少了界面方程的求解时间. 因此,它能够充分利用分布式存储并行计算机的体系结构特点提升大规模并行计算效率. 最后通过典型数值算例验证了该方法的正确性和有效性. 

关 键 词:分布式存储并行计算机    分层并行计算    区域分解    缩聚
收稿时间:2013-10-10

A HIERARCHICAL PARALLEL COMPUTING APPROACH FOR STRUCTURAL STATIC FINITE ELEMENT ANALYSIS
Institution:1.State Key Laboratory of Mechanical System and Vibration, Shanghai Jiaotong University, Shanghai 200240, China2. School of Mechanical Engineering, Shanghai Jiaotong University, Shanghai 200240, China3. Shanghai Supercomputer Center, Shanghai 201203, China
Abstract:According to the architecture characteristics of distributed memory parallel computers, a hierarchical parallel computing approach for structural static finite element analysis is proposed. Based on the two-level partitioning and twice condensation strategies, the proposed method not only improves memory access rate through the distributed storage of a large amount of data, but also significantly improves communication rate with the three-layer parallelization of the computational procedure. Moreover, the interface equations' size is further reduced and its solution time is considerably reduced. Thus, it can improve the efficiency rates of parallel computing of large-scale problems by fully exploiting the architecture characteristics of distributed memory parallel computers. Finally, typical numerical experiments were used to validate the correctness and efficiency of the proposed method. 
Keywords:
本文献已被 CNKI 等数据库收录!
点击此处可从《力学学报》浏览原始摘要信息
点击此处可从《力学学报》下载免费的PDF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