[rigm]

Dear Edge Users,

I was wondering whether someone is willing to share her/his experience (or whether publications are already available) on the following topics:

- Assessment of the SGS models available in LES (or DES) simulations of flows with non-negligible compressibility effects

- Assessment of the impact of explicit filtering (including deciding not to use explicit filtering) of the resolved variables

- Eventual attempts to carry out LES simulations wihtout an SGS filter (mimicking ILES or MILES) - i.e. disregarding the fact that second-order formal accuracy is likely to excessively damp the smaller resolved scales of motion

Many thanks in advance,
marcello

[Peng]

Hi Marcello,

there have been many applications of LES (or DES) for compressible flows, but not much on the assessment of SGS modelling of compressibility effects. When an SGS model calibrated in incompressible flows is applied to LES of compressible flows, many have just simply written the modelling formulation in a "compressible form", and it usually works....If the compressibility effects are verified being important, one could try to take a modeling formulation in analogy to RANS modeling for compressible turbulent flows.

You may find something useful in the book by Garnier, Sagaut and Adams (LES for Compressible Flows) and some other literature referred to in this book.

For engineering applications, implicit filtering has been commonly used in solvers using finite volume method (like in Edge). In order to reduce the effect of truncation error (stemmed from the numerical scheme used) on the SGS modelling, explicit filtering has been used by some in LES of fundamental flows (e.g. decaying isotropic, homogeneous turbulence, channel flow, mixing layer....). In this case, the filtering operation and the differentiation should commute. It would be rather complicated for unstructured solvers. I am not sure that the SFS (sub-filtered scale) modelling would help much to improve the accuracy of simulation.

A couple of PhD students at KTH are actually using Edge doing "LES" without involving any SGS model, namely, similar to ILES. This would require the numerical scheme plays a role in the energy transfer between resolved and SGS turbulence, usually, by the incurred numerical dissipation.

SHP

[rigm]

Dear Shia-Hui,

Many thanks for your reply. I know very well the book you mention.
I also know that explicit filtering is not common practice and for that reasons sgs modeling is not always physically meaningful, as you say.

It would be interesting to know what strategy the PhD students are applying to obtain physically consistent numerical dissipation out of a second order scheme.
Maybe you are in contact with them.

Another interesting question concerning LES and Edge would concern the existence of "aeroelastic" simulations where the motion of the boundaries and the energy-carrying turbulent structures have similar frequency content.

Kind regards,
marcello

[Peng]

Hi Marcello,

you have raised some fundamental (and interesting) questions concerning not only Edge but also other similar solvers in general.

I don't have much details about how the KTH researchers have enabled their ILES-type calculations using Edge. The results are rather encouraging, however. Plausibly, the numerical dissipation could have been manipulated into a form similar to the Smagorinsky model for energy dissipation. Better you get direct contact with Mr. Olle Bodin (bodin@mech.kth.se) at KTH for detailed discussion.

As regards the "aero-elastic simulations", I don't quite understand about your question(?). If a wall-resolved LES is performed, nonetheless, the time step used should be restricted not only by the CFL number, and it must also be small enough to resolve the evolution of large-scale, energetic turbulent structures in time (namely, the time step should be smaller than the turbulent time scales to resolve).

Unless one has very powerful computing resources and targeting fundamental aerodynamic flows, nevertheless, I would not encourage Edge users to perform full LES. Instead, one could consider to use hybrid RANS-LES methods. These modelling approaches have been proved able to resolve surface pressure fluctuations reasonably well on wall boundaries.

SHP.

[rigm]

Dear Shia-Hui,

Many thanks for the answer and the contact. I expressed myself badly. I was simply referring to the case where the largest eddies and the structure (i.e. the boundary) move with similar frequencies.

Another related subject is the Riemann solver used in the "Roe" scheme. Would you know if anyone has ever tried any alternative version?

Best regards,
marcello