[philippebv]

Good afternoon foamers,

Does anyone have an idea on how to correctly feed the LES part of the DES calculation when a prescribed velocity profile is used at the inlet? For example when you have experimental data or an upstream RANS simulation. I'm using profil1DfixedValue (Sig Turbomachinery Library OpenFoamTurbo) and I would like to superimpose fluctuations to that velocity profile.

Thank you for your help,

Philippe

[grtabor]

For LES, mapping the downstream values back to the inlet works very well; this was first done by Eugene de Villiers for his PhD thesis. If you couple this with a feedback algorithm you can drive the solution towards any profile you care to specify. Shameless plug follows; take a look at

Baba-Ahmadi, M.H. and Tabor, G.R. (2009): Inlet conditions for LES using Mapping and Feedback Control, Computers and Fluids, 38#6, pp 1299 - 1311

Baba-Ahmadi, M.H., and Tabor, G. (2008): Inlet Conditions for LES of gas-turbine swirl injectors, AIAA.J., 46#7, pp 1782 - 1790

I don't see any inherent reason why this couldn't be adapted to use for DES, although there might be issues concerning the near wall/RANS region.

Gavin

[philippebv]

Hi everyone,

M. Tabor,
Thank you for your answer. I hadn't seen those papers and I'm glad to have them in front of me now. The variants to the baseline technique are interesting indeed for an application where a specified profile is desired at the inlet. By any chance, do you plan on making the implementations of the "Methods" (in Baba-Ahmadi, M.H. and Tabor, G.R. 2009) available for other foamers? I should have known that you were working on the subject of inlet conditions for LES and talk to you about it in Montreal! Anyways, I hope you can provide me with your best feeling for my application :

Imagine you have a diffuser, with swirling inlet, and for which the inlet conditions are axisymmetric (u,v,w,p,nut) from an upstream RANS simulation. The velocity profile looks like :

velocityProfileInlet.jpg

As you can imagine, feeding the DES calculation with turbulent quantities issued from a RANS calculation is very questionable, although this is what I'm doing at the moment and that I'm trying to improve. What is mind bugging is the concern you pointed out with the near wall/RANS region. Even with an appropriate LES inlet, I'm not quite sure what would be the right value of nut in the RANS region at the inlet.

Coming back to the mapping, the issues I have are that :
-It is not a fully developed flow, and I'm not very comfortable with the mapping plane downstream of the inlet since I don't want to have a growing boundary layer. Maybe I could use a free slip condition in the inlet section (between inlet and mapping plane) to overcome this?

-There is no physical constant cross-section downstream of the diffuser section. The inlet is right at the beginning of the diffuser as can be seen on this picture :

diffuserSection.jpg

so it is unlikely to imagine that I could map some plane in the interior of the domain to specify the inlet condition. Or maybe using a scaling of the quantities to adapt a bigger cross-section mapping plane to the inlet plane?

The simplest way for me would have been to simply add random fluctuations to the inlet profile, but unfortunatly and as De Villiers says, this doesn't work!

Do you have answers, ideas, suggestions? Or maybe this message is too long and you haven't read it up to here

Have a good one guys,

Philippe

[grtabor]

Hi Philippe,

Our control algorithm approach works for any profile. Since the control is based on the discrepancy between the target and actual flow conditions _at the mapping plane_, then you should decide where your non-fully-developed flow profile is located, treat that as the mapping plane, and extend the domain upstream sufficiently to apply the method. The mapped section probably isn't physically realistic (we are applying all sorts of tricks to force the flow in the right direction there) but downstream of the mapped plane it should evolve in a physically realistic manner (I hope!)

The expanding nature of your diffusor might be a bit of a problem!

As for the RANS/LES interface; I've never really got my head around that for DES anyway - it seems that there is likely a jump in the conditions where the switch occurs. Does anyone blend the two together over a number of cells? One possibility at the inlet might be to do a RANS calculation across the whole inlet, and only apply the mapping/control algorithm to the core region?

In terms of release of the code; its on my list of priorities.... I'm working with another PhD student who is busy improving the code and applying it to a heat transfer case; I'll talk to him on Monday to see where he has got to.

Gavin

[philippebv]

Thank you for your explanations, it's always nice to have the input of an expert!

I'll see what I can do from there with my case and I'll let you know how I manage. Keep us informed with the release of the code.

Have a good day,

Philippe

[panda60]

Originally Posted by grtabor

For LES, mapping the downstream values back to the inlet works very well; this was first done by Eugene de Villiers for his PhD thesis. If you couple this with a feedback algorithm you can drive the solution towards any profile you care to specify. Shameless plug follows; take a look at

Baba-Ahmadi, M.H. and Tabor, G.R. (2009): Inlet conditions for LES using Mapping and Feedback Control, Computers and Fluids, 38#6, pp 1299 - 1311

Baba-Ahmadi, M.H., and Tabor, G. (2008): Inlet Conditions for LES of gas-turbine swirl injectors, AIAA.J., 46#7, pp 1782 - 1790

I don't see any inherent reason why this couldn't be adapted to use for DES, although there might be issues concerning the near wall/RANS region.

Gavin

Dear Prof.Tabor,
I want to use recycling method, if "directMappedFixedValue" is the best utility we can modify ? Could you give me some suggestions on how to modify to use this in OpenFOAM ? Thanks.

[anishtain4]

Hi Dr. Tabor,
I wonder if your boundary condition every got to be released to other researchers?

[tiam]

Originally Posted by anishtain4

Hi Dr. Tabor,
I wonder if your boundary condition every got to be released to other researchers?

I would like to second this request. In fact, the 2010 review paper contains a bunch of methods, as far as I understood, implemented in OpenFOAM. It would be great to get those codes!

Best,
Timofey