Convergence Acceleration Techniques for Coupled Adaptive Cartesian-Strand Grid Solutions

Convergence acceleration techniques are evaluated for a two-dimensional strand and Cartesian grid overset framework using a novel computational infrastructure. The method is based on a coupled multigrid algorithm tailored to the unique grid topology of near-body strand grids and off-body nested Cartesian grids. For the strand grids, which are automatically generated prismatic grids extruded from the body surface, efficient directional coarsening is achieved by agglomerating the unstructured surface and structured strand directions independently. For the nested Cartesian grids, a block-structured approach makes use of existing multi-level grid data, requiring no additional coarse mesh construction. A modular, multi-language computational infrastructure is used to couple near- and off-body solvers using overset domain connectivity algorithms that take advantage of the compact strand and Cartesian grid storage. Results are presented which show dramatic multigrid speed-up for strand and Cartesian grids independently, and together as a coupled system.