Premium
Iterative strategies for solving systems of linear, algebraic equations arising in 3D BE–FE analyses of tunnel drivings
Author(s) -
Payer H.J.,
Mang H. A.
Publication year - 1997
Publication title -
numerical linear algebra with applications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.02
H-Index - 53
eISSN - 1099-1506
pISSN - 1070-5325
DOI - 10.1002/(sici)1099-1506(199705/06)4:3<239::aid-nla107>3.0.co;2-3
Subject(s) - finite element method , algebraic equation , discretization , iterative method , convergence (economics) , system of linear equations , linear system , mathematics , algebraic number , field (mathematics) , basis (linear algebra) , linear equation , boundary element method , algorithm , nonlinear system , mathematical analysis , geometry , physics , quantum mechanics , pure mathematics , economics , thermodynamics , economic growth
Simulations of excavations of tunnels can be performed by means of a coupling strategy using the boundary‐element‐method (BEM) and the finite‐element‐method (FEM). The BEM is employed for the discretization of the far‐field of the tunnel, whereas the FEM is used for the interior of the tunnel and its vicinity. This strategy is based on the assumption that the influence of material non‐linearities in the far‐field can be neglected. The coefficient matrices of the resulting systems of linear, algebraic equations consist of a symmetric, banded, sparse part resulting from the FE subdomain and an unsymmetric, fully populated part obtained from the BE subdomain. Substantial reductions of the CPU time and the disk space required for the solution of the systems of linear, algebraic equations can be achieved by employing iterative solvers with preconditioning and using the computer hardware intelligently. In this paper, two strategies based on such solvers are proposed and explained in detail. The convergence behaviour of the implemented iterative solvers with regards to pure BE calculations is studied by means os solving selected numerical examples. The advantages of the two proposed methods relative to the original, direct solution strategy are demonstrated on the basis of coupled simulations. Additionally, some comments concerning the use of iterative solvers on a vector computer are given. © 1997 John Wiley & Sons, Ltd.