11.2.3. Grid Resolution SRS Models

Grid Resolution SRS models are discussed in the following sections:

11.2.3.1. Wall Boundary Layers

For wall boundary layers, it is important to distinguish if they are computed in RANS or wall-resolved or Wall-Modeled LES (WMLES) mode. Only the IDDES and the SBES model can be run in WMLES mode reliably.

In the case of RANS mode, the requirements are the same as for any RANS model. For a wall-resolved LES, it is typically recommended to use a mesh with a grid spacing scaling with where x is the streamwise, y the wall normal and z the spanwise direction (for example, channel flow). However, in complex applications, the distinction between streamwise and spanwise direction is not feasible and then would be required. This scaling demonstrates the strong Reynolds number dependency of the LES approach for wall bounded flows.

For the IDDES and SBES models in WMLES mode, the above requirements can be relaxed. The grid spacing no longer scale with the wall friction, but with the boundary layer thickness . It is recommended that you use . The wall normal resolution should be like for a finely resolved RANS simulation, meaning a near wall resolution of and around 30 nodes inside the boundary layer.

11.2.3.2. Free Shear Flows

For free shear flows, it is difficult to provide general recommendations, as there are many different flow scenarios. The current recommendation is therefore based on the most common (and most frequent) free shear flow – a turbulent mixing layer. It will not necessarily apply to other free shear flows like jets and you are advised to perform tests as to the optimal resolution of your specific flow.

For free shear flows and SRS models, one should aim for uniform isotropic cells (all edges have similar length). The shear layer should be covered by ~10-20 cells. As the layer thickness is often hard to estimate, one can also look at the ratio

(11–3)

Where is the maximum cell edge length. It is available in ANSYS CFD-Post for all turbulence models. This estimate should be based on the RANS solution and not on the SAS / DES / SDES / SBES solution for the given flow. The above requirement for the mixing layer translates into:

(11–4)

In these estimates, 10 cells across the layer (or ) is truly a lower limit. The higher values of 20 cells and are much safer.


Release 18.2 - © ANSYS, Inc. All rights reserved.