欢迎访问《轻工机械》稿件在线采编系统!设为首页 | 加入收藏    
信息公告:  
文章检索:
稿件处理系统
期刊信息
  • 中国标准连:ISSN1005-2895
  • 续出版物号: CN 33-1180/TH
  • 主管单位:轻工业杭州机电设计研究院有限公司
  • 主办单位:轻工业杭州机电设计研究院有限公司、中国轻工机械协会、中国轻工业机械总公司
  • 社  长:刘安江
  • 主  编:黄丽珍
  • 地  址:杭州市余杭区高教路970号西溪联合科技广场4-711
  • 电子邮件:qgjxzz@126.com
理事单位          MORE>>
陆建阳, 张伟峰.超大变形分析无网格法并行计算[J].轻工机械,2012,30(5):
超大变形分析无网格法并行计算
Parallel Computation of Meshfree Methods for Extremely Large Deformation Analysis
  
DOI:
中文关键词:  无网格法  并行计算  超大变形  重构核质点法(RKPM)
英文关键词:meshfree methods  parallel computation  extremely large deformation  reproducing kernel particle method (RKPM)
基金项目:
作者单位
陆建阳, 张伟峰  
摘要点击次数: 1520
全文下载次数: 0
中文摘要:
      鉴于无网格法相较于其他技术计算任务更重,以及求解问题本身的高度复杂和大强度计算,并行计算对此显得尤具吸引力。重构核质点法(RKPM)是一种对固体和结构进行大应变弹塑性分析的常用无网格法,考虑到它能精确建模超大变形问题而无网格扭曲现象,以及能方便地对需要精细化的区域进行简单改变质点定义即可求解,在此仅集中讨论该方法。并行程序包括网格分区预分析和并行计算,后者包括各处理器上分区间的显式信息传递。文中用基于图像的Metis程序来进行网格分区,由于其重定义技术能应用于不同几何部件的共享区域,该程序常见于基于网格的分析中。并行模拟和MPI信息传递在SGI Onyx3900超级计算机上完成。文中对不同分区的并行计算效果和性能进行了比较分析,并给出了无网格法与有限元法的对比结果。
英文摘要:
      Due to the heavier computation requirement than other competitive techniques and the essence of applications that are usually highly complex and computationally intensive, parallel computing is especially attractive for these meshfree methods. The paper focused on discussion the Reproducing Kernel Particle Method (RKPM), one of the meshfree methods for large strain elasto plastic analysis of solid and structures, in considering with its ability to accurately model extremely large deformations without mesh distortion problems, and its ease of adaptive modeling by simply changing particle definitions for desired refinement regions. The parallel procedure primarily consists of a mesh partitioning pre analysis phase and parallel computing which includes explicit message passing among partitions on individual processors; with redefinition techniques applied to the shared zones of different geometrical parts, the graph based procedure Metis, which is quite popular for mesh based analysis, is used for partitioning in this meshfree analysis. Parallel simulations have been conducted on an SGI Onyx3900 supercomputer with MPI message passing statements. The effectiveness and performance with different partitions has then been compared, and a comparison of the meshfree method with finite element methods is also presented.
查看全文  查看/发表评论  下载PDF阅读器
关闭