论文信息 - A Domain Decomposition Method for First-Order PDEs

A Domain Decomposition Method for First-Order PDEs

In this report a nonoverlapping domain decomposition method to solve first-order, time-dependent partial differential equations is developed. The time discretization used is implicit, which gives a large system of equations to solve for each time step. Preconditioners with a fast inversion based on a fast modified sine transform are defined. Theoretical analysis of the method is presented, indicating that the ratio in the grid might be crucial for the convergence. Finally numerical results from a parallel implementation on an Intel Paragon are presented, showing very nice properties. Especially a nonuniform decomposition of the domain leads to very good results.

Lina Hemmingsson

[1] H. Elman. Iterative methods for large, sparse, nonsymmetric systems of linear equations , 1982 .

[2] Maksymilian Dryja,et al. A capacitance matrix method for Dirichlet problem on polygon region , 1982 .

[3] Y. Saad. Krylov subspace methods for solving large unsymmetric linear systems , 1981 .

[4] O. Axelsson,et al. A restarted version of a generalized preconditioned conjugate gradient method , 1988 .

[5] Bertil Gustafsson,et al. Navier-Stokes equations for almost incompressible flow , 1991 .

[6] William Gropp,et al. A comparison of domain decomposition techniques for elliptic partial differential equations and their parallel implementation , 1985, PP.