Bedřich Sousedík

Multispace and multilevel BDDC

The Balancing Domain Decomposition by Constraints (BDDC) method is the most advanced method from the Balancing family of iterative substructuring methods for the solution of large systems of linear algebraic equations arising from discretization of elliptic boundary value problems. In the case of many substructures, solving the coarse problem exactly becomes a bottleneck. Since the coarse problem in BDDC has the same structure as the original problem, it is straightforward to apply the BDDC method recursively to solve the coarse problem only approximately. In this paper, we formulate a new family of abstract Multispace BDDC methods and give condition number bounds from the abstract additive Schwarz preconditioning theory. The Multilevel BDDC is then treated as a special case of the Multispace BDDC and abstract multilevel condition number bounds are given. The abstract bounds yield polylogarithmic condition number bounds for an arbitrary fixed number of levels and scalar elliptic problems discretized by finite elements in two and three spatial dimensions. Numerical experiments confirm the theory.

Adaptive BDDC in three dimensions

AbstractThe adaptive BDDC method is extended to the selection of face constraints in three dimensions. A new implementation of the BDDC method is presented based on a global formulation without an explicit coarse problem, with massive parallelism provided by a multifrontal solver. Constraints are implemented by a projection and sparsity of the projected operator is preserved by a generalized change of variables. The effectiveness of the method is illustrated on several engineering problems.

Adaptive-Multilevel BDDC and its parallel implementation

We combine the adaptive and multilevel approaches to the BDDC and formulate a method which allows an adaptive selection of constraints on each decomposition level. We also present a strategy for the solution of local eigenvalue problems in the adaptive algorithm using the LOBPCG method with a preconditioner based on standard components of the BDDC. The effectiveness of the method is illustrated on several engineering problems. It appears that the Adaptive-Multilevel BDDC algorithm is able to effectively detect troublesome parts on each decomposition level and improve convergence of the method. The developed open-source parallel implementation shows a good scalability as well as applicability to very large problems and core counts.

BDDC and FETI-DP under minimalist assumptions

SummaryThe FETI-DP, BDDC and P-FETI-DP preconditioners are derived in a particulary simple abstract form. It is shown that their properties can be obtained from only a very small set of algebraic assumptions. The presentation is purely algebraic and it does not use any particular definition of method components, such as substructures and coarse degrees of freedom. It is then shown that P-FETI-DP and BDDC are in fact the same. The FETI-DP and the BDDC preconditioned operators are of the same algebraic form, and the standard condition number bound carries over to arbitrary abstract operators of this form. The equality of eigenvalues of BDDC and FETI-DP also holds in the minimalist abstract setting. The abstract framework is explained on a standard substructuring example.

Application of the parallel BDDC preconditioner to the Stokes flow

Abstract A parallel implementation of the Balancing Domain Decomposition by Constraints (BDDC) method is described. It is based on formulation of BDDC with global matrices without explicit coarse problem. The implementation is based on the MUMPS parallel solver for computing the approximate inverse used for preconditioning. It is successfully applied to several problems of Stokes flow discretized by Taylor–Hood finite elements and BDDC is shown to be a promising method also for this class of problems.

Hierarchical Schur complement preconditioner for the stochastic Galerkin finite element methods Dedicated to Professor Ivo Marek on the occasion of his 80th birthday.

Use of the stochastic Galerkin finite element methods leads to large systems of linear equations obtained by the discretization of tensor product solution spaces along their spatial and stochastic dimensions. These systems are typically solved iteratively by a Krylov subspace method. We propose a preconditioner which takes an advantage of the recursive hierarchy in the structure of the global matrices. In particular, the matrices posses a recursive hierarchical two-by-two structure, with one of the submatrices block diagonal. Each one of the diagonal blocks in this submatrix is closely related to the deterministic mean-value problem, and the action of its inverse is in the implementation approximated by inner loops of Krylov iterations. Thus our hierarchical Schur complement preconditioner combines, on each level in the approximation of the hierarchical structure of the global matrix, the idea of Schur complement with loops for a number of mutually independent inner Krylov iterations, and several matrix-vector multiplications for the off-diagonal blocks. Neither the global matrix, nor the matrix of the preconditioner need to be formed explicitly. The ingredients include only the number of stiffness matrices from the truncated Karhunen-Lo\`{e}ve expansion and a good preconditioned for the mean-value deterministic problem. We provide a condition number bound for a model elliptic problem and the performance of the method is illustrated by numerical experiments.

Adaptive Coarse Space Selection in the BDDC and the FETI-DP Iterative Substructuring Methods: Optimal Face Degrees of Freedom

We propose adaptive selection of the coarse space of the BDDC and FETI-DP iterative substructuring methods by adding coarse degrees of freedom (dofs) on faces between substructures constructed using eigenvectors associated with the faces. Provably the minimal number of coarse dofs on the faces is added to decrease the condition number estimate under a target value specified a priori. It is assumed that corner dofs are already sufficient to prevent relative rigid body motions of any two substructures with a common face. It is shown numerically on a 2D elasticity problem that the condition number estimate based on faces is quite indicative of the actual condition number and that the method can select adaptively a hard part of the problem and concentrate computational work there to achieve the target value for the condition number and good convergence of the iterations, at a modest cost.

On Multilevel BDDC

1 Department of Mathematical Sciences, University of Colorado at Denver and Health Sciences Center, Campus Box 170, Denver, CO 80217, USA, jmandel@math.cudenver.edu,sousedik@math.cudenver.edu 2 Department of Mathematics, Faculty of Civil Engineering, Czech Technical University in Prague, Thakurova 7, 166 36 Prague 6, Czech Republic 3 Structural Dynamics Research Department, Sandia National Laboratories, Mail Stop 0847, Albuquerque NM 87185-0847, USA, crdohrm@sandia.gov

Simulation of the 2009 harmanli fire (bulgaria)

We use a coupled atmosphere-fire model to simulate a fire that occurred on August 14---17, 2009, in the Harmanli region, Bulgaria. Data was obtained from GIS and satellites imagery, and from standard atmospheric data sources. Fuel data was classified in the 13 Anderson categories. For correct fire behavior, the spatial resolution of the models needed to be fine enough to resolve the essential micrometeorological effects. The simulation results are compared to available incident data. The code runs faster than real time on a cluster. The model is available from openwfm.org and it extends WRF-Fire from WRF 3.3 release.

A posteriori error estimates applied to flow in a channel with corners

The paper consists of three parts. In the first part, we investigate a posteriori error estimates for the Stokes and Navier-Stokes equations on two-dimensional polygonal domains. Special attention is paid to the sources of the constants in the estimates, as these play a crucial role in practical applications to adaptive refinements, as we also show. In the second part, we deal with the problem of determining accurately the constants that appear in the estimates. We present a technique for calculating the constant with high accuracy. In the third part, we apply the a posteriori error estimates with the constants found numerically to the technique of adaptive mesh refinement-we solve an incompressible flow problem in a domain with corners that cause singularities in the solution.