The Full Multigrid Method Applied to Turbulent Flow in Ventilated Enclosures Using Structured and Unstructured Grids

Abstract: Computational fluid dynamics require large computer resources and it is thus of paramount interest to have as efficient an algorithm as possible. This means that both an accurate discretization and a fast solver are required.

In this work, we have used the full multigrid concept, which is a general and optimal solver. It has been applied to subsonic flows in ventilated enclosures, and as these flows are usually turbulent, we have used several different turbulence models, such as the k-.epsilon. model, the k-.omega. model and large eddy simulation methods. We have investigated the effects of such models on the performance of the multigrid method and present some guidelines to improve both the speed and the robustness of the multigrid method in combination with turbulence models.

We have employed and developed this multigrid technique for structured (one-block( grids and locally refined grids, as well as for completely unstructured grids. Both the local mesh refinement method and the adaptivity showed to be very efficient and substantial reductions in both RAM memory usage and CPU-time usage were observed. All of these discretizations are at least second-order accurate, which has been confirmed on laminar benchmark problems. The performance of the multigrid has in these model cases been found to be high, i.e. convergence within work comparable with 100 residual evaluations. However, more important is that the same performance was also most often shown for turbulent flows on both structured and unstructured grids.