Investigation of Scale Adaptive Simulation (SAS) Turbulence Modelling for CFD-Applications
Fluid dynamics simulations generally require large computational recourses in form of computer power and time. There are different methods for simulating fluid flows that are more or less demanding, but also more or less accurate. Two well known computational methods are the Reynolds Averaged Navier-Stokes (RANS) and Large Eddy Simulation (LES). RANS computes the timeaveraged flow properties, while LES resolve the large structures (eddies) of the flow directly and model the small ones. Hybrid models are combinations of these two models which have been developed to improve the RANS solutions and shorten the simulation time compared to LES computations. One such model is the Scale Adaptive Simulation (SAS) model which uses the RANS model in steady flow regions, such as close to walls, and a LES like model in unsteady regions with large fluctuations.
This study was done for evaluating the SAS model compared to Unsteady RANS (URANS) and LES and their performance compared to measurements from an engineering point of view. This was done by running simulations on two different test cases, one external and one internal flow situation. The first one was flow around a wall-mounted cylinder and the second one was flow through an aorta with a coarctation in the descending aorta. The first test case was used to thoroughly evaluate the SAS model by running many simulations with URANS, SAS and LES with different element types, element sizes and flow parameters. The element types that have been analyzed are; tetrahedral, hexahedral and polyhedral. The results were compared with experiments done by Sumner et al. [7, 8, 9, 10]. The second test case was used for evaluating the SAS model even further on another flow situation. For this test case, only two SAS simulations were performed on two different grids; a structured hexahedral and an unstructured polyhedral. These results were compared with Magnetic Resonance Imaging (MRI) measurements obtained from Linköping University.
No conclusion of which one of the simulated cases gives the best overall agreement with experimental results could be concluded from the obtained results. The best prediction of the drag coefficient for the cylinder was obtained for the coarsest polyhedral mesh that was run with LES, with the disagreement 0.4 percent. The best prediction of the Strouhal number was obtained for a URANS simulation performed on the coarsest mesh with an improved grid close to the cylinder surface, generating less than one, with a disagreement of 3 percent compared to measurements. For the meshes used, it was found that the polyhedral mesh gave the best overall results and the tetrahedral mesh gave the worst results for the cylinder case. For the aorta case the SAS model produced velocity components that had acceptable agreement with the MRI-measurements, but gave very poor results for the turbulent kinetic energy. The main conclusion of this thesis was that the SAS model performed better than URANS, but took longer time to compute simulations than LES, which was the model that generated the best overall results.
AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)