[lakeat]

Thank you very much, Markus.

What is Turbulence Special Interest Group, and where is it, may I ask?

[lakeat]

Hi Markus,

What about using a camera instead of a scanner? I think it would be fine. I use camera for many times and I see the papers look okay.

Regards,
Daniel

[braennstroem]

Hi,

@ Markus: yes, the gradient idea is not really working...

For the SST model, one simple switching condition could be y*=y*sqrt(k)/nu

volScalarField S = sqrt(2.0) * mag(symm(fvc::grad(U_)));
volScalarField Ssgs = sqrt(2*S*S);
volScalarField nuSGS=sqr(Cs_*delta()) * Ssgs;
yStar=y_*sqrt(k_)/nu();
f_=((tanh(0.01*(yStar-500.0)))+1)/2;
volScalarField nuT=(a1_*k_/max(a1_*omega_, F2()*sqrt(S2)));

nut_ = (1-f_)*nuT + (f_)* min(nuSGS,nuT);

Could this work for you?

About the SIG, you can a little bit on:
http://openfoamwiki.net/index.php/Sig_Turbulence

Regards!
Fabian

[cfdmarkus]

Hi Fabian,

I like the fact, that switching the model equations gives you well defined models for the LES and RANS part. Eventhough, the Smagorinski Sgs model is probably not the best choice. If you would just switch the length scale, as in the DES approach, it is arguable whether the RANS model performs well as a Sgs model (I guess it does not).
Having a properly defined SGS and RANS model seems very appealing to me. However, I am not really sure what the most consistent hybrid formulation would look like. For the moment, I haven't decided which method I am going to use.

Regards
Markus

[philippebv]

Hi Markus,

As a DES user, I found very useful the use of the DDES implementation. I suggest you consider it for the hybrid formulation to choose. Of course, it was well suited for my application and might not be for yours, but it definitely helps to prevent the switch from RANS to LES inside the boundary layer.

Regards,

Philippe

[lakeat]

Hi dear friends,

I made a simple DDES implementation, and not sure if it works. Do you have any idea and experience in using DDES?
Howto check it? Anyway, I am working on channelFlow to see if it works better than DES as a wall function.

The 2nd question is in S-A model:
----------<pre>
volScalarField r = min
(
nuTilda_
/(
max(Stilda, dimensionedScalar("SMALL", Stilda.dimensions(), SMALL))
*sqr(kappa_*dTilda_)
),
scalar(10.0)
); </pre>
-----------
But why r's limit is 10.0?
I have set my rd (let you understand) in DDES no larger than 1.0 (Please correct me if I was wrong)

so, any ideas?


Daniel WEI

[philippebv]

Good morning Daniel, I hope you're doing fine!

My DDES implementation does not limit the value of rd. Maybe it would be a better practice to do it but the tests I made so far look good for the values of rd and fd. The picture above shows those values in the ERCOFTAC conical diffuser (attached B.L.).



I have very little experience in channelFlow, but I guess you can get better results with DDES if you use fine grids. To check your implementation, I suggest you refine your mesh a few times to see the effect. Then you can plot rd and fd to see if they make sense. I would be very interested to see some of your results!

Keep us posted and good luck,

Philippe

[lakeat]

My openSUSE is broken, So I spent a whole afternoon reinstall the OpenSUSE-11.1. I'll get OpenFOAM work tomorrow. Because of the Spring Festival, I have been delayed for a whole month.

In Fact, "Your Image IS NOT Here".

"Refine the mesh"? Is it a good idea to do that? As far as I remember what Spalart said before, DES is not always show consistency with grid refining.

You did not limit the value of rd? Will it be larger than 1.0 some places?
So far in channelFlow, I have saw a very very thin layer near the wall, both in fd and rd fields. The changes are really "steep" near the wall, just as what has been expected. I will do postChannel these days.

It's very nice to see your progress, I am very interested too to see your results!

Happy day!

DW

[philippebv]

Dear Daniel, I hope you enjoyed the Spring Festival, even if you were delayed in your work;)

I'll try again for the image :



The way I see it, DDES is just meant to "protect" the boundary layer from LES mode. It forces the RANS formulation to stay active as long as f_d=0. So the point of implementing it is to use what Spalart calls "ambiguous spacings" in the wall parallel direction. So to answer your question, I think that yes it is a good idea to refine the mesh and see the effect because that's the purpose of DDES.

As for r_d, I'm pretty sure you don't have to limit it because of the definition of fd :

f_d = 1-tanh([8r_d]³)

which automatically limits the right hand side to 1. f_d is then between 0 and 1 as expected.

Regards,

Philippe

[lakeat]

Thank you, you're right!
Your Image IS NOT Here again
Good night, Philippe!

[philippebv]

[philippebv]

[philippebv]

Can't do it sorry. I'll send you an email

Bye

[lakeat]

Hello, fomers, where are you now, long time no discussion, How's everything？
What did you guys use DES work for?
And what's going on with maxDelta implementation?

Daniel