Mechanical and Civil Engineering Seminar

The large eddy simulation (LES) technique for turbulent flows have become a standard tool of academic research, but has yet to really make an impact on the engineering design and analysis process in more applied situations. The talk will identify a range of reasons for this situation, and will describe recent work towards solving two of the main obstacles: the need for robust and accurate wall-models in LES, and the need for a solution-driven approach to grid-adaptation in LES. The proposed approach to wall-modeling is based on the multi-scale nature of turbulent boundary layers and on the need for showing grid-convergence in numerical simulations. These considerations naturally lead to a number of criteria on wall-modeled LES, and a simple method that satisfies these criteria is presented. This is then shown to lead to excellent accuracy on a number of test cases, including supersonic boundary layers and shock/boundary-layer interactions. A simple method that enables predictions of laminar-to-turbulence transition is presented and tested on a boundary layer and a multi-element airfoil at flight Reynolds number. The talk will end by discussing a proposed approach to automatically find a near-optimal grid in LES of wall-bounded flows. A directional error indicator is introduced, which measures the level of kinetic energy at scales near the grid-spacing in a directional manner; this then enables the grid to be adapted to resolve the turbulence in a more balanced way. The method is tested on turbulent channel flows with very reasonable results.