The second generation FETI methods and their application to the parallel solution of large-scale linear and geometrically non-linear structural analysis problems

The FETI algorithms are a family of numerically scalable domain decompositi on methods. They have been designed in the early 1990s for solving iterativ ely and on parallel machines, large-scale systems of equations arising from the finite element discretization of solid mechanics, structural engineeri ng, structural dynamics, and acoustic scattering problems, and for analyzin g complex structures obtained from the assembly of substructures with incom patible discrete interfaces. In this paper, we present the second generatio n of these methods that operate more efficiently on large numbers of subdom ains, offer greater robustness, better performance, and more flexibility fo r implementation on a wider variety of computational platforms. We also rep ort on the application and performance of these methods for the solution of geometrically non-linear structural analysis problems. We discuss key aspe cts of their implementation on shared and distributed memory parallel proce ssors, benchmark them against optimized direct sparse solvers, and highligh t their potential with the solution of large-scale structural mechanics pro blems with several million degrees of freedom. (C) 2000 Elsevier Science S. A. All rights reserved.