Without more details is difficult to help you. What you describe seems more an effect of the numerical dissipation introduced by the scheme.

Hello Professor Denaro,

and first of all thank you for coming back to my question so quickly! I see - I attached a screenshot of the side section cut of my simple fluid volume with the flow direction from left to right (image 1). My goal is to make sure that a body in the flow is actually hit by flow with a turbulence intensity level of (close to) 1 percent and is keeping this level at least until the end of the body. Increasing the turbulent length scale (9cm to 30cm) at the inlet gives me higher intensity values through the channel (image 2) - but it doesn't influence the dissipation of the reintroduced turbulence levels. That is what I am looking for - influencing this turbulence level -based on modified turbulent diffusivity source - after the region. My mesh resolution in this case is 1.5 cm through the whole field.

Does that give you more of an idea and the details you where asking for?

If we're talking about numerical dissipation - do you think that the reintroduction region is interpreted as a wall? So modified diffusivity based turbulence levels do experience more numerical dissipation the further they are away from this wall? How could I achieve some kind of backscatter that adds this additional turbulent kinetic energy back to the filtered equations of the simulation? Because that is something I do not see. The resolved turbulent kinetic energy stays the same even after the second region (image 3)

Thanks a lot for your advices

Let me address a point, in terms of DES you have some switching parameter between RANS and LES, for example as described for the Ansys code http://www.afs.enea.it/project/neptunius/docs/fluent/html/th/node89.htm

Now, you are working in an external zone, far from any walls, so your formulation should switch properly towards a LES solution. The first question is about your resolved variables, do you see a 3D unsteady field developing and correlating? The images you attached does not show clearly such correlation, it appears that the flow develops in a more similar way as a RANS formulation.

What is the type of discretization you are introducing? Have you first performed a simulation in a controlled test-case used for DES?

Hello,

thank you for the clarification - I guess I understood that part right then. I am working with a commercial tool so my filtering is implicit and mesh dependent. Following the mesh density I chose, it should be fine enough to switch to LES at that point - if it's not already in LES mode because the region is technically no wall right? If the mesh wasn't fine enough to switch to LES it would stay in RANS is that correct?

I do see a field developing yes - I attached an image with the actual body in the flow field that I took out of my other calculations to get a better view on what's happening with the modified diffusivity after it got introduced to the field.. It is an iso surface of the vorticity magnitude (image 4). Is that what you meant with your question?

I havn't performed a controlled DES test case so far - my setup is already supposed to be a simplified model.. Do you suggest that to make sure my settings are actually DES settings? What could be a typical test case? Something like the a cylinder in a flow with the comparison of Karman streets in Spalart's article? Article Detached-Eddy Simulation

Dear Timotheus Haeussler

the image you posted is not fully exhaustive, it is only a qualitative representation of the fact you get a 3D and unsteady solution. But you need some more quantitative evaluation to check if the field is physically correlating or you are still in a numerical transient. From the image I see the field containing very large structures but seems to me that no smaller structures (up to the grid size) are present. That could be a signal of a field that needs to correlate again but it could be also the warning about an excess of numerical viscosity.

In conclusion, I strongly suggest to use this setting to replicate the standard case of the flow over a cylinder and compare quantitatively your solution to those reported in literature.

Hello Prof. Tapan K. Sengupta

Thank you for your suggestions and interest in giving me advices on how to solve this issue. I appreciate you stating me that it is good to bring it on. Oftentimes I am questioning my work a lot – so it is always a good feedback to hear that the questions I am asking are not stupid.

My whole mesh in the simplified case I am working on is uniform. I tried breaking down my set-up to as little different influences as possible.

My discretization is already the highest implemented in the tool I am using – I forgot to add this paragraph to my last post. Following the set-up of the case I am actually looking forward to apply my conclusions from the simplified case on, I am using a Hybrid Second-Order Upwind/Bounded-Central and a Second-Order Euler implicit temporal discretization.

Also thank you for your feedback Prof. Filippo Maria Denaro

I will try your suggestion and will set up a standard cylinder case.