[Cheese]

Hi. I just did a fully LES and the results are quite surprising. The turbulent kinetic energy (TKE) of my RANS-Simulations (exact same setup) are much higher. But the streamprofiles of the LES (e.g. velocity) are showing a much better mixing. The TKE-profile at the inlet also does not fit to the specified boundary condition. On the other hand the manual calculated TKE based on the velocity fluctuation does fit. Now i am wondering if ANSYS is calculating the TKE just with the results of the sub grid model. I do not find any confirmation for this, but it seems quite obvious with the given results. Does anyone know more about this? Thanks in advance.

[ghorrocks]

Can you explain your question more clearly? The TKE profile at the inlet of which simulation (LES or RANS?) does not fit the specified boundary condition? What are you using for your boundary condition? What did you expect the profile to look like and what did you get?

Now i am wondering if ANSYS is calculating the TKE just with the results of the sub grid model.

Again, please explain this question more as I have no idea of what context you are asking.

[Cheese]

Thank you for your reply. The main issue is that the TKE of the LES in the whole domain is much lower than that of my RANS-simulation. For example: At the inlet it´s about 5-20 m^2/s^2 while the values of the RANS are between 100-180 m^2/s^2.

I specified a turbulent intensity profile as boundary condition at the inlet. If i calculate the TKE manually (based on the fluctuations of the LES results or experimental data) i get the values 100-180 m^2/s^2 of my RANS-Simulation. That´s why i am thinking that the TKE calculation by ANSYS does not include the solved part of the turbulence.

Shoudn´t be the dimension of the TKE of the LES and the RANS-Simulation at least close to each other?

The mixing of the LES is even higher. Does this mean that the turbulent values should be higher as well?

Unfortunately i can not calculate the TKE manually for the whole domain since the trace of the reynolds stress tensor is missing.

If i still did not make my point clear, just ignore the upper part: My question is: Does ANSYS take the modelled and the solved part of the turbulence for the calculation of the TKE into account?

Thanks again for your answer, ghorrocks.

[ghorrocks]

No, the TKE variable in the LES model does not include the resolved turbulence. By definition it represents just the unresolved turbulent fluctuations.

While I cannot answer your other questions as it would require a detailed analysis of your situation and some research, here are some comments which may be useful:

Don't forget there are many ways of defining turbulent fluctuations from a LES flow, such as time averaging, spatial averaging and ensemble averaging. They will all give you different results. And also for time and space averaging you then have to define a time or length scale to average to - so the choice of this scale will affect the TKE.

RANS turbulence models are not perfect. The k-e turbulence model is well known for over-estimating the TKE in several situations, especially in free shear layers and around separations. So comparing to LES model to k-e is expected to lead to significant differences.

[Cheese]

Originally Posted by ghorrocks

No, the TKE variable in the LES model does not include the resolved turbulence. By definition it represents just the unresolved turbulent fluctuations.

Thank you a lot. That´s all i needed to know.

Edit: I am aware of the over-estimating problem. In my case the turbulence is under-estimated (U-turn flow) and i am trying to figure out how the statistic turbulence models can be improved in the given case.