AMG for problems of elasticity (Q1083916)

The papers deals with the use of the algebraic multigrid method (AMG) for solving scalar partial differential equations. The algebraic multigrid method is designed to utilize the principles of the geometrically oriented multigrid method in order to obtain fast solutions to matrix equations. It is an efficient iterative method which can be used as a ''black box'' solver for systems of equations with emphasis on those arising in elasticity. The basic philosophy of AMG, in contrast to usual multigrid methods, is to fix the relaxation method and choose the coarse grid and the interpolation in such a way that the error resulting from relaxation lies in the range of the interpolation operator. A coarse-grid problem can then be defined which eliminates that error. A model problem used for test was the 2-D plane-stress problem. Four different types of discretizations (finite differences on a non-staggered or staggered grid, finite elements with bilinear test functions or with linear elements) and five different boundary conditions were used. The results show that AMG is an efficient solver for some problems in elasticity, in spite of a poor enough convergence.