Mechanics of 21 st Century-ICTAM 04 Proceedings
Zhenhan Yao,Haitao Wang,Pengbo Wang
2010-01-01
Abstract:As well known, the boundary element methods (BEM) has remarkable advantages of dimension reduction and higher accuracy, but the conventional BEM is not efficient for large scale problems. For the conventional BEM O(N) operations and O(N) memory are required, where N is the number of DOF. Fast multipole (FM) methods presented by Rokhlin, Greengard and other mathematicians have reduced the operations and memory requirement tremendously, finally reduced to O(N). The FM BEM becomes an attractive way to solve complex practical scientific and engineering problems with BEM. There are nearly 500 papers so far have been cited by SCI in the field of fast multipole BE. In solid mechanics, especially if the local effects are important in the analysis, the higher accuracy is required, and some further investigations are still required for their application in different large scale problems. In this presentation, some investigations of authors’ group on FM BEM in solid mechanics are presented, including: a new version algorithm of O(N) FM BEM for 2D elastostatics and its application to the simulation of 2D elastic solid with large number of randomly distributed inclusions; application of FM BEM to the simulation of 2D elastic solid containing large number of randomly distributed cracks. Three years ago in authors’ group, the largest scale of practical problem solved by conventional BEM was 54,000 DOF, on a parallel PC cluster consisted of 8 PCs. The scale of numerical examples by FM BEM presented in this presentation, already reaches 544,000 DOF and 1,500,000 DOF, calculated on one PC.