Evaluating the Performance of Wall-Modelled Large-Eddy Simulation on Unstructured Grids

Abstract

Large Eddy Simulation (LES) has become a promising CFD method for flows in which turbulence plays a dominant role such as many complex engineering systems. To contribute to further advances, in this work, the performance of Wall-Modelled LES on unstructured grids is evaluated. Periodically arranged hills geometry is a frequent experimental and numerical test case because of the possibility of studying important fluid phenomena. The importance of study of this flow arise from separation and reattachment points, and hence the whole flow is sensitive to the separation process. In this work, WMLES is used to investigate the properties of a separated flow in a periodic hill channel flow. The principal idea is to evaluate the sensitivity of the predicting accuracy to grid resolutions. To do that, three mesh types including triangle, polygon, and square prism meshes are employed in the study. Grid resolution has also been considered by using four grid sizes - different cell-to-cell distance - for each type of mesh. Twelve grids are used for the simulations and the results are judged by existing reference data. The results are included for skin friction and pressure coefficients, mean and vertical velocity, and kinetic energy profiles. The results illustrate that WMLES predict the flow features accurately. The statistical data elicited from the study illustrate a noticeable influence of grid topology on the results and prove that meshing strategy plays a key role in accurate prediction. Also, results illustrated a noticeable distinction in sensitivity between separation and reattachment points. Regarding this, the reattachment point is highly sensitive to the grid size.