A localized orthogonal decomposition method for heterogeneous Stokes problems

From MaRDI portal





In this paper, a multiscale method is proposed for heterogeneous Stokes problems. It is based on localized orthogonal decomposition (LOD) by decomposing the solution space into a fine-scale space and its orthogonal complement. Because the functions average out on coarse scales, choosing the fine-scale space, a finite-dimensional mesh-based complement space is obtained that is adapted to the problem, which has uniform approximation properties under minimal structural assumptions on the coefficients. The basis functions decay exponentially whose computation can thus be localized to subdomains. For Stokes problems, the divergence-free constraint poses a major challenge to LOD-type methods. To overcome this problem, the Stokes problem is reformulated, such that the divergence is piecewise constant with respect to some coarse mesh, as the solution space for the velocity. The divergence-free velocity is then recovered using a piecewise constant Lagrange multiplier defined on the same mesh.\N\NAn a priori error analysis is performed and optimal orders of convergence are obtained as the mesh size is decreased, provided that the support of the basis functions is allowed to increase logarithmically with the desired accuracy. The implementation of the proposed multiscale method is discussed and numerical experiments are presented that confirm the theoretical results.



Cites work









This page was built for publication: A localized orthogonal decomposition method for heterogeneous Stokes problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6939700)