UFR 3-34 Best Practice Advice
Underlying Flow Regime 3-34
Best Practice Advice
Key physical features of the UFR in question are: separation of the turbulent boundary layer from a smooth surface driven by adverse pressure gradient, a rapid development of three-dimensional turbulent structures in the separated shear layer, its reattachment to the plane wall, and further relaxation of the reattached turbulent boundary layer to the “normal” state.
Physical Modelling Issues
Based on the conclusions formulated above the following advices for the computations of the considered UFR may be given.
Use turbulence-resolving approaches (hybrid RANS-LES or global WMLES and WRLES), since none of the currently available RANS turbulence models, either linear eddy-viscosity or RSM, ensures capturing of the challenging physical features of the UFR indicated above. However a success of the scale-resolving approaches is also not guaranteed. In order to reach it:
- Impose as realistic as possible turbulent content at the inflow of LES zone within zonal RANS-LES approaches.
- Ensure a rapid development of 3D turbulent structures in the separated shear layer within nonzonal hybrid approaches (this demands employing of special grey-area mitigation tools).
- Use sufficiently wide domains for simulation of nominally 2D geometries with periodic boundary conditions in the uniform direction.
Whether all the above recommendations are really implemented in a simulation should be checked by:
- Visualizing the unsteady solutions (see, e.g., Figs. 13, 14) and getting a visual impression on the length of the RANS-to-LES transition in the zonal RANS LES simulations, on 3-dimensionality of the early region of the separated shear layer in the non-zonal hybrid simulations, and on sufficiency of the domain width for representation of large-scale structures showing up in a flow field.
- For the considered UFR characterized by high Reynolds numbers, no transition modelling is required.
- For the considered test case, using hybrid methods with wall modelling is highly recommended, since they ensure a huge saving of computer resources compared to the WRLES and are not inferior to the latter in terms of accuracy, at least as far as the mean flow characteristics are concerned. However, this recommendation is not “general” since for other, more complex, flows belonging to the same UFR, WMLES may not be accurate enough (see, e.g., ).
- Effect of the specific choice of a wall model for LES has not been investigated, but the choice of the IDDES model may be safely recommended.
- In the global WMLES/WRLES, in the focus region of non-zonal hybrid RANS-LES simulations including a separation bubble and an initial part of the reattached turbulent boundary layer, and in the part of the attached boundary layer prior to separation in the LES zone of the zonal approaches, use numerics with as low as possible numerical dissipation, particularly, pure or close to pure central difference schemes for convective fluxes with not less than 2nd order of accuracy. Acceptability of the level of numerical dissipation should be assessed by examining snapshots of, e.g., vorticity: the size of the smallest resolved eddies should be comparable with the local grid spacing.
- In the inviscid (Euler) and departure regions of the computational domains use schemes with numerical dissipation sufficient to prevent grid oscillations or “wiggles” in these regions.
- Use a minimum second order accurate temporal integration scheme and time step sufficient to capture the motion of the turbulent eddies resolved by the grid. This corresponds to the approximate guideline for the convective CFL = of the order of 1.0
In terms of grids, although systematic grid-sensitivity studies for the UFR in question have not been performed, indirect evidence allows the following recommendations:
- In the attached boundary layer treated by WMLES use the first normal to the wall grid step in the wall units, , equal to 1 or 2 at most. Any forms of wall functions, which in principle allow much larger near wall steps, should be used with caution. Use grid-steps in the streamwise and lateral directions not larger than 1/10th (preferably 1/20th) of the local boundary layer thickness.
- In the WRLES, grid steps, Δx+ , Δz+ , Δy+ , in the near wall part of the boundary layer (up to about y+ ≈ 200) should be about 25, 12 and 1 respectively.
- In the Focus Region of the simulations, use nearly isotropic grids.
- Outside the Focus Region, gradually expand the grid cells towards the outer boundaries, avoiding sudden jumps.
Time sample / statistical processing
- Ensure that the simulation has bridged the initial transient and entered a statistically steady state before starting the collection of statistics. This should be checked by visual inspection of monitor unsteady signals and by computing and analyzing the “running average” of some key flow characteristics. As a guideline, roughly 10 convective time units of initial transient can be expected.
- Compute sufficiently long time samples for getting reliable statistical quantities. As a tentative guideline, a minimum time-sample of 10-20 convective units is recommended. An objective evaluation of sufficiency of the time sample can be done by directly checking the effect of a considerable increase of the time sample.
Comparability of CFD predictions and experiment
- Ideally, reproducing in CFD the full experimental geometry (i.e. including the “floor”, “ceiling” and side walls of the experimental test-section) would be desirable, which, unfortunately, is usually prohibitive in terms of computational cost and a lack of the needed information. In simulations of the considered 2DWMH test case an adjustment of the upper wall approximately accounting for the blockage effect of the end plates is obligatory.
- In compressible simulations of low Mach number flows, in order to reproduce the separated flow region correctly, the value of the Mach number in the simulations should be specified equal to that in the experiment.
Recommendations for Future Work
- In order to avoid inevitable uncertainties associated with representation in CFD of the “quasi-2D” or “nominally 2D” experimental setups, a new experimental campaign with 3D but easily reproducible in CFD geometry and boundary conditions would be of significant value for further improving CFD approaches as applied to the considered UFR.
Contributed by: E. Guseva, M. Strelets — Peter the Great St. Petersburg Polytechnic University (SPbPU)
© copyright ERCOFTAC 2023