Numerical Solution of Partial Differential Equations: Theory, Algorithms, and Their Applications


M. Buck, O. Iliev, H. Andrä. Multiscale Coarsening for Linear Elasticity by Energy Minimization. Numerical Solution of Partial Differential Equations: Theory, Algorithms, and Their Applications, Springer Proceedings in Mathematics & Statistics,  Volume 45, pp 21-44, 2013.


M. Buck, O. Iliev, H. Andrä


Linear elasticity, Domain decomposition, Robust coarse spaces, Energy-minimizing shape functions




In this work, we construct energy-minimizing coarse spaces for the finite element discretization of mixed boundary value problems for displacements in compressible linear elasticity. Motivated from the multiscale analysis of highly heterogeneous composite materials, basis functions on a triangular coarse mesh are constructed, obeying a minimal energy property subject to global pointwise constraints. These constraints allow that the coarse space exactly contains the rigid body translations, while rigid body rotations are preserved approximately. The application is twofold. Resolving the heterogeneities on the finest scale, we utilize the energy-minimizing coarse space for the construction of robust two-level overlapping domain decomposition preconditioners. Thereby, we do not assume that coefficient jumps are resolved by the coarse grid, nor do we impose assumptions on the alignment of material jumps and the coarse triangulation. Weonly assume that the size of the inclusions is small compared to the coarse mesh diameter. Ournumerical tests show uniform convergence rates independent of the contrast in the Young’s modulus within the heterogeneous material. Furthermore, we numerically observe the properties of the energy-minimizing coarse space in an upscaling framework. Therefore, we present numerical results showing the approximation errors of the energy-minimizing coarse space w.r.t. the fine-scale solution.