Journal of Computer Applications ›› 2018, Vol. 38 ›› Issue (4): 1072-1077.DOI: 10.11772/j.issn.1001-9081.2017092384

Previous Articles     Next Articles

Parallel algorithm for explicit finite element analysis based on efficient parallel computational strategy

FU Chaojiang1,2, WANG Tianqi1, LIN Yuerong1   

  1. 1. College of Civil Engineering, Fujian University of Technology, Fuzhou Fujian 350180, China;
    2. Fujian Provincial Key Laboratory of Advanced Technology and Informatization in Civil Engineering, Fuzhou Fujian 350180, China
  • Received:2017-10-09 Revised:2017-12-17 Online:2018-04-10 Published:2018-04-09
  • Supported by:
    This work is partially supported by the National Natural Science Foundation of Chian (51378124).

基于有效并行求解策略的显式有限元分析并行算法

付朝江1,2, 王天奇1, 林悦荣1   

  1. 1. 福建工程学院 土木工程学院, 福州 350108;
    2. 福建省土木工程新技术与信息化重点实验室, 福州 350108
  • 通讯作者: 付朝江
  • 作者简介:付朝江(1966-),男,江西九江人,教授,博士,主要研究方向:并行计算、数值分析;王天奇(1994-),男,江苏南京人,硕士研究生,主要研究方向:数值分析;林悦荣(1991-),男,江西赣州人,硕士研究生,主要研究方向:数值分析。
  • 基金资助:
    国家自然科学基金资助项目(51378124)。

Abstract: Concerning the time-consuming problem of finite element analysis for solving the nonlinear dynamic problems of large-scale structure, some parallel computational strategies for implementing explicit nonlinear finite element analysis were proposed under the environment of Message Passing Interface (MPI) cluster. Based on the technique of domain decomposition with explicit message passing, using overlapped, non-overlapped domain decomposition techniques and Dynamic Task Allocation (DTA) algorithm, domain decomposition parallel algorithms for overlapped domain, non-overlapped domain, clustering for DTA, DTA and Dynamic Load Balancing (DLB) were researched by overlapping calculations and communications to improve the performance of communication between processors. A parallel finite element analysis program was developed with message passing interface as software development environment. Some numerical examples were implemented on workstation cluster to evaluate the performance of the parallel algorithm, the computation performance was also compared with the conventional Newmark algorithm. The experimental results show that the performance of the algorithm for dynamic task allocation with clustering technique is better than that of the dynamic task allocation, which is lower than that of the domain decomposition algorithm, and the dynamic load balancing algorithm is the best. For the problem with the same size, the proposed algorithms are faster and better than conventional Newmark algorithm. The proposed algorithms are efficient for parallel computing of nonlinear dynamic problems of structure.

Key words: parallel computational strategy, domain decomposition, dynamic load balancing, parallel algorithm, finite element analysis

摘要: 针对大规模结构非线性动力问题的有限元分析非常耗时,基于消息传递接口(MPI)机群环境,提出多种基于并行求解策略的显式有限元并行算法。基于显式消息传递的区域分解技术,采取重叠、非重叠区域分解技术及动态任务分配方法,通过将计算与通信重叠,优化处理器间的通信,对非重叠通信区域分解并行算法、重叠通信区域分解并行算法、群动态任务分配算法、动态任务分配算法及动态负载平衡算法进行研究。为在机群环境下实现非线性动力有限元分析,开发了基于有效并行求解策略的显式有限元并行算法。编写了基于消息传递编程模式的并行有限元程序,在工作站机群上实现了数值算例,分析了算法的性能,并与传统的Newmark算法进行了比较。算例表明:群动态任务分配算法的性能优于动态任务分配算法,低于区域分解算法的性能,动态负载平衡算法最优。对相同规模的问题提出的算法比Newmark算法快,优于Newmark算法。对结构非线性动力问题的有限元分析,所提出的并行算法是可行有效的。

关键词: 并行计算策略, 区域分解, 动态负载平衡, 并行算法, 有限元分析

CLC Number: