[philippebv]

Hi,

I know I brought this up before, but I would like to control the position of the switch between the RANS and the LES formulations with the SpalartAllmaras model. It is hard to achieve when delta=cubeRootVol. Since I'm not an expert at programming, could someone help me change the line:

delta_.internalField() = deltaCoeff_*pow(mesh().V(), 1.0/3.0);

in order to set delta=max(deltaX, deltaY, deltaZ) in the file cubeRootVolDelta.C. I'm working with hexa meshes only. Maybe I could use the variable "directions" as shown in the same file:

const Vector<label>& directions = mesh().directions();
label nD = (directions.nComponents + cmptSum(directions))/2;

But I don't know how to write the whole thing. Any solution, idea or comment will be apreciated.

Thanks,

Philippe

[lakeat]

Dear all,

Several questions again.

After reading Eugene's thesis, I feel deeply impressed by DES. My major is bridge engg., and when wind blows on the bridge girder sections, the girder will vibrate with buffeting, vortex-excited-oscillation and even fluttering, it's obvious that my main concern is Fluid-Structure Interaction (FSI), as I have been convinced now, DES works very well for large Re and flow with strong separation, that's exactly what I need, so the question remains:

1. Is DES able to apply to FSI (I name it as a NP-hard problem), I mean, Would DES be capable to predict flutter phenomenon? Any ideas? Is there any thing special that I need to pay attention?
I remember someone told me that even URANS is able to predict the drag-reduction, I am not saying that flutter is related with drag-reduction, but It would be impossible for DES to predict flutter phenomenon if it failed to predict drag-reduction...

2. For a large-span bridge girder section, the Re would reach to 10e6~10e7, it's quite high, but what's more, the flow will separate immediately at the head of girder wind fairing, Adverse Pressure Grdient is therefore strong since the separation is fierce, so, I need be careful about the APG as I use LES, Please correct me if I was wrong.

3. Are there any DES benchmarks that I can follow? Since no tutorial now for S-A model, the first problem is how can I mesh well!!?? Is there a good article on it?

4. It's obvious the unstructured grid is inevitable for my bridge case, which software do you guys use to mesh? Could you give my some advice, gmsh?

5. What is the procedure for DES? Is it in this way?
a) URANS with coarse mesh first;
b) map the field to fine mesh for DES

6. How much more time will DS costs than One-Equation model, 20%? 30%? I did not notice the lagaragian DS model in OpenFOAM, are there anyone working on it?

7. Please, give me some hints on how to draw a spectral analysis E(k)~k?



It would be very nice to hear your ideas, thanks!

Regards, \Daniel

[seang]

nice article on meshing,

Young-Person's Guide to Detached-Eddy Simulation Grid

by Philippe R. Spalart

NASA/CR-2001-211032

... title along the lines of Hitch-hiker's Guide to the Galaxy ...

[lakeat]

Hi Seang!

Thanks for your info. I have downloaded them.
I am reading some papers about DES. I feel Spalart's writing style is hard to grasp.

To my knowledge now, I found meshing in DES is ad hoc with some concern for the wall-parallel meshing......Comparing to Full-LES, in DES I should use a large spacing in the streamwise and spanwise direction. And the spacing must be of the order of BL-thickness/C_DES? So in order to use DES in my the channelFlow case, I need to remesh it, CMIIAM.

Another thought, is it possible to use Dynamic SGS models in DES?

Thank you!


\Daniel

[johndeas]

Hi,

I was looking at a way to implement heat transfer in LES. I have seen this done for boyant flow, but not for incompressible flow.

A quick look at src » turbulenceModels » LES » compressible » GenEddyVisc » GenEddyVisc.H

Indicate that

00118 virtual tmp<volscalarfield> alphaEff() const
00119 {
00120 return tmp<volscalarfield>
00121 (
00122 new volScalarField("alphaEff", muSgs_ + alpha())
00123 );
00124 }

I do not understand this operation. Since alphaEff=alphaTurb+alpha, I do not understand why muSgs_ is put directly, and not modified using Pr_turb = muSgs / alphaTurb.

Initially, I was also looking for the value of Pr_turb used in the solver.

Can somebody help clarify those two points ?

[johndeas]

The way it is written, it seem to imply that Pr_turb = 1, which puzzles me.

[villet]

Hi John,

yes, that's correct. Reynolds analogy is presumed in turbulent Prandtl number. In RANS model you can adjust turbulent Prandtl number by "alphah" coefficient.

[ivan_cozza]

Hi guys!
I'm intrested in setup a LES (or maybe DES) calculation over a flat plate, zero incidence, Re = 5.0e+5, Mach = 0.11, so incompressible.
I'm intrested in extract u' v' w' to use it as source terms in an aeroacoustic calculation. My intrest is for a quasi 2D problem, as my CAA code is 2D.
What kind of mesh resolution (y+, x+, z+) and z extension I need?
What should be a good model for this case?
I was thinking on Spalart Allmaras model, it could be a good idea?
Thanks a lot for the help!

[nnigro]

Hi

I would like to confirm the relation between Smagorinsky constant Cs and Ck and Ce coeffs used in this context.
Moreover, in the case of the dynamic approach which role plays the ce coefficient ?

Thanks a lot

Norberto Nigro

[syoon]

I'm simulating flows around a bluff body with oodles.
I simulated from 0 (s) to 0.25 (s) with delta_t=1e-5 without any problem and it seemed flows became statistically meaningful (or converged).

So I want to start averaging from 0.25 to 0.5 (s) without taking into account data before 0.25 (s). Would anyone let me know how I can do that?

[philippebv]

Hi,

In the folder 0.25 , just delete nAveragingSteps.raw (in folder Uniform) and then restart the calculation from that timestep. The averaging should start over.

Good luck,

Philippe

[syoon]

Thank you, Philippe.

In my 0.25/uniform folder, I can see only two files (fieldAveragingProperties and time) and I can't see nAveragingSteps.raw.

If I restart from that timestep, it seems that all the data before 0.25 is accounted for in averaging.

Sungho

[philippebv]

Hi Sungho,

The method I described was for OF 1.4.1 dev. I think that for version 1.5, you have to edit the controlDict during the simulation, where the fieldAveraging is defined as a function. Refer to the file OpenFOAM-1.5/tutorials/oodles/pitzDaily/system/controlDict.

I hope this helps,

Philippe

[villet]

Hi Sungho,

delete "fieldAveragingProperties" file. The number of time averaging steps is kept there.

Ville

[syoon]

Deleting'fieldAveragingProperties' file did the job.
Thank you Ville and Philippe.

Sungho

[lakeat]

Dear Eugene, and Dear All,

Concerning the second eigenvalue, I want to start a new thread, but I I'm not sure whether the problem is big enough to deserve that, so I ask you here.

I saw in your thesis a figure:
Figure 2.4: Velocity magnitude representing lifted low speed streaks (dark blue) at and isosurfaces of streamwise vortices, for .

Could you tell me how can I get in FOAM?

Many thanks in advance!

Daniel

[lakeat]

Sorry for disturbance, I've got the solution now. Lambda2 utility.

[lakeat]

I am using oodles with S-A model to do a flat-plate in a wind tunnel simulation, and the simulation blows many times, so i guess my boundary setting is wrong. Here is the picture of how I set them. Are these settings correct?
----------------------------------------------------------------------
To show in a picture:plot.JPG
----------------------------------------------------------------------
upAndDown:
Slip Wall
U slip
p slip
nuSgs zeroGradient
nuTilda slip

frontAndBack:
cyclic
U cyclic
p cyclic
nuSgs cyclic
nuTilda cyclic

plate:
no-slip Wall
U fixedValue, (0 0 0)
p zeroGradient
nuSgs zeroGradient
nuTilda fixedValue, 0

inlet:
patch
U fixedValue, (1 0 0)
p zeroGradient
nuSgs zeroGradient
nuTilda fixedValue, 5E-5

outlet:
patch
U convectiveOutlet
p fixedValue, 0
nuSgs zeroGradient
nuTilda inletOutlet
----------------------------------------------------------------------

Thanks for any help!

Daniel

[lakeat]

Hi foamers,

I don't want to start a new thread, so I post it here, I am using DES.
My simulation blows again, could anyone give me some hints, thanks.

Second, I feel puzzled about what is boundary layer on earth?
In a flat plate, according to the formula,
,

The bl thickness can be easily plotted:fplate_cr.jpg

However, what's about an inclined flat plate, if . How could it be defined?

I found in the implemented DES, is quite narrowed to the thin region around the flat plate:dTildaAbovedW.jpg.
where the wall distance is:dW.jpg
which means RANS is activated only in a very small thin area.


CubeRootVol method is used and the mesh I am using is a double-sided stretching grids along the plate, like this: (I see a great difference here in using DES between maxVolLength and CubeRootVol in defining the RANS region.)
mesh.jpg

So I am hard pressed between the two, for one reason, a double-sided stretching grids should be set to achieve the accurate simulation at the trailing edge, for the other, in DES idea, the activated RANS region should grow larger as the B.L thickens towards the trailing edge. BUT in z direction, the grid is set to be uniform. which means a coarser grids should be set towards the trailing edge.
So what's your idea, what could be the better way in achieving this goal, either by meshing or anything else?

More on, please see my Umag:
Umag.jpg
Note the region before the plate, it's obviously not correct what might go wrong?

Any help would be highly appreciated.

Daniel

[syoon]

I’m simulating a backward facing step using oodles.
I’m comparing different inlet boundary conditions.
To begin with, as shown in the figure (b), I used no fluctuation and white noise at the inlet (x=-3h).
The mean velocity profile was obtained from experimental data.

Then, in order to generate inlet turbulence properly, I’m using ‘directMappedPatch’.
I read Eugene’s thesis regarding this boundary condition and looked at ‘pitzDailyDirectMapped’ in the tutorial directory.
If my understanding is correct, directMappedPatch works as shown in the figure (b).
The old inlet (x=-3h) works as a mapping patch and a new inlet (x=-7h) acquire data from the old inlet usring directMappedPatch.
Is my understanding right?



My question is, then, how should I set up ‘system/changeDictionrayDict’?
Say, my h = 1. Then do I need to set up in system/changeDictionaryDict as follows?

dictionaryReplacement
{
boundary
{
unlet
{
Type directMappedPatch;
Offset (4 0 0);
}
}
}

Of should I use offset (-3 0 0) insteady of offset(4 0 0) ?
In other words, is the 'offset' relative to the new inlet or x=0?

Please advise me.
Sungho