Comparison between substructure method and domain decomposition method
Advantages and drawbacks of SSM [SubStructure Method (direct scheme)] in contrast with DDM [Domain Decomposition Method (iterative scheme)] is investigated. In higher-order nonlinear problem, several iterative methods show slow convergence or are hard to converge. In such case, the direct scheme will be inevitable. In this paper, direct scheme applied to substructure method which is suitable for parallel computer has been examined. In the previous year, a research-program was built to investigate parallel efficiency of both direct and iterative schemes by Cray-T3D. The program has been enhanced for this research by adding hierarchical substructure method as well as nonlinear capabilities and has been tuned up for VPP300 supercomputer. Using some test problems with over 1,000,000 DOF (degrees of freedom), are examined characteristics of substructure method and domain decomposition methods. Consequently, it has been shown that substructure method has superiority in some specific problems and requires much more memories comparing with the other in general. Domain decomposition method shows slow convergence in some problems, but it is superior in most cases to substructure method. It is shown that hierarchical substructure method has high efficiency in computational time, too.
Unable to display preview. Download preview PDF.
- 1.Yagawa, G., Soneda, N., Yoshimura, S.: A large scale finite element analysis using domain decomposition method on a parallel computer. Comput. Struct. 38, No.3 (1991) 269–281Google Scholar
- 2.Farhat, C., Roux, F.X.: A method of finite element tearing and interconnecting and its parallel solution algorithm. Int. j. numer. Methods. Eng. 32 (1991) 1205–1227Google Scholar
- 3.Saxena, M., Perucchio, R.: Parallel fem algorithms based on recursive spatial decomposition — ii. automatic analysis via hierarchical substructuring. Comput. Struct. 47, No.1 (1993) 143–154Google Scholar
- 4.Yagawa, G., Yoshimura, S., Soneda, N.: A parallel finite element method with a supercomputer network. Comput. Struct. 47,No.3 (1993) 407–418Google Scholar
- 5.Elwi, A.E., Murray, D.W.: Skyline algorithms for multilevel substructure analysis. Int. J. numer. Methods. Eng. 21 (1985) 465–479Google Scholar
- 6.Miyoshi, T.: Supercomputing in Computational Solid Mechanics. JSME A,57,541 (1991–9) 16–21Google Scholar
- 7.Edelman, A., Schreiber, R.: Tutorial Workshop on Parallel Processing. Tokyo and Osaka, Japan (January,1996)Google Scholar