low reynolds re turbulence boundary and yPlus values
 

Hallo all,

I'm solving with the buoyantBoussinesqSimpleFoam solver with low-re turbulence models, so I don't use any wall-functions.

So I'm using for my k field e.d.:
wall1
{
type fixedValue;
value uniform 0.00000001;
}

for the epsilon field:
wall1
{
type fixedValue;
value uniform 0.000001;
}
and for the nut field:
inlet
{
type calculated;
}

Is it right to set the nut field to calculated? The simulations works great, the residuals are ca. 10-9.
When trying to plot the yPlus values with yPlusRAS I get following message:
no nutWallFunction patches
Writing yPlus to field yPlus

Plotting them in paraview I see only yPlus = 0.
So do I need a wallfunction in my nut-field? Or did I set the case wrong?

Thanks a lot,
Thomas

Hi all,

shouldn't the yPlus value be calculated without using wallfunctions by:

y+ = (y*u_tau)/nu
with
u_tau=sqrt(Tau_wall/rho)
Tau_wall = mu du/dz at the nearest cell to the wall.
so Tau_wall = mu * u_cell/y_cell


So:
y+=sqrt (y * u_cell/nu)

So I'm not sure, but in the RAsModel.C and in the yPlusRAS.C data should be modified to calculate the yPlus-values for low-reynolds turbulence models.

Regards Thomas

I am not sure about this but interested.
First of all I think you still have to use a low re turbulence model in 1.6 for low-re calculations. Afaik it is not possible to use one of the high-reynolds models and only set nut to calculated. I guess you did this?
I think calculation of y+ is independend of high or low re model always done by the formula you have given. So this should be ok.
I don't kow if setting nut to calculated is right for low-re-turbulencemodels, maybe someone can help us about that.

Hi Thomas and bastil,

I had exactly the same issue, for my (periodic) pipe flow with RANS models. For obatining fields of u+ and y+ (especially for plotting u+ over y+ velocity profiles) I created a new utility which does the job for me.

Since its really an ugly piece of code, I would prefer to send it to you via eMail then uploading it here or posting the entire code, so if you are still interested, please tell me and provide me with your eMail adress.

Best,
Florian

Hallo basil and Florian,

firstly I'm using real low-reynolds turbulence models e.d. LaunderSharmaKE and LamBremhorstKE.

For high reynolds-number models you calculate the y+ values in an different way, because you are modelling the velocity-profil in the wall-next-cell. So you have to get the real wall-velocity-gradient, using the velocity-field values give you too low gradients.
The y-plus value is calculated in the RASModel.C with the log-law:
Yp = pow(Cmu, 0.25)
*y_[patchNo]
*sqrt(k()().boundaryField()[patchNo].patchInternalField())
/nu().boundaryField()[patchNo];

So you can only use it for high-reynolds-turbulence-models.
So Florian, it would be very kind of you to send me your code. I will send you my e-mail adress.

Regards,
Thomas

In fact, my utility is intended for the use of low-Re RANS models. I myself use the LaunderSharmaKE model.

Best,
Florian

I see. I was not aware of this.
What is the correct way to set-up a low-reynolds case in 1.6? Is it to set type calculated; for nut? Is there an example out there for this?

Regards.

Summarizing
 

Hello,
please can someone check if this is correct?

1. using a high-Re turbulence model and setting nut as nutWallFunction, the velocity profile near the wall is modeled as a high-Re model requires; i.e. the previous OF version behavior is reproduced. In this way, y+ can be obtained as output.
2. using a high-Re turbulence model and setting nut as calculated, the velocity profile near the wall is obtained from the velocity field, and the y+ is required by the calculation, thus it is not an output. In this case, the velocity gradient at the wall are too high and the boundary layer is thicker than it should be.
3. using a low-Re turbulence model requires to set nut as calculated, since this is the only possible way to solve the flow; no y+ value as output.
4. using a low-Re turbulence model and if a y+ value is necessary as output, the modification made by Florian should be used.

Cheers,
Maddalena

PS: this will answer to a similar question I posted some days ago here as well...

Maddalena,

al you write makes perfectly sence to me. However I am not sure about the exact of 1.6. Maybe one of the developers can comment on this since there are many similar threads around there at the moment.

Hello BastiL,
yes, it would be great if the developers could comment on that!
Some more observations:

• It seems to me that using the setup 1 is more difficult to keep the simulation stable (i.e. k and epsilon explode) in comparison of setup 2. But since the latter gives higher wall velocity gradient, should not to be the other way round?
• Can one conclude that setup 2 is a sort of coupling between high-Re and low-Re turbulence model, and this can be used in the proper boundaries if the turbulence field of the case requires a high-Re model in a zone and a low-re model in a different zone?

Cheers,

maddalena

Hello,

setup 1 - for my high-Re kEpsilon simulations I also use wall function for nut, I have no problems with instability and exploding k and epsilon values.... (might be a flaw somewhere else in your setup)

setup 3 - why should it be neccesary to use calculated for nut? I use zeroGradient and the simulation results are quite good (look at attached uPlus yPlus plot), I just made a test using calculated and its not changing considerably.

best,
Florian

After some review of other threads I think I got it:
1. is right. This is the traditional "Hi-Re" modelling.
2. I do not think you should do this.
3. I think we must not set nut for that case. Set k end epsilon as "zeroGradient". Look at the tutorial compressible/sonicFoam/ras/nacaAirfoil (uses LaunderSharmaKE). This is the traditional "Low-Re" modelling.
4. is right
to add 5. If you have a hybrid/all yplus model OpenFOAM 1.6 offers kind of a hybrid wall function. This seems to work quite fine at least for komega-Turbulence. You need to set nut to "nutSpalartAllmarasWallFunction" even though this name is quite confusing. It is intended as kind of a general purpose blending function... (http://www.cfd-online.com/Forums/ope...tml#post244144). Sometimes I simply wish the documentation to be better....

To come to an end I have posted a bug report to get the yPlus-calculation fixed.

Hello All,

I am trying to simulate an impingingjet using the simpleFoam in OF1.6.x. I am trying to follow the threads and tried different things. But I am still getting the same error. I am posting my error and attaching my initial bc. Your support will be greatly appreciated.

Time = 759

DILUPBiCG: Solving for Ux, Initial residual = 0.364807, Final residual = 0.0331038, No Iterations 1
DILUPBiCG: Solving for Uy, Initial residual = 0.401396, Final residual = 0.0185339, No Iterations 1
DILUPBiCG: Solving for Uz, Initial residual = 0.439448, Final residual = 0.0046591, No Iterations 1
GAMG: Solving for p, Initial residual = 0.569614, Final residual = 0.00428777, No Iterations 66
time step continuity errors : sum local = 3.14096e+42, global = 3.09629e+39, cumulative = 6.80133e+40
DILUPBiCG: Solving for epsilon, Initial residual = 1.08298e-09, Final residual = 1.08298e-09, No Iterations 0
DILUPBiCG: Solving for k, Initial residual = 0.000115294, Final residual = 2.84805e-06, No Iterations 2
ExecutionTime = 47327.3 s ClockTime = 48069 s

Time = 760

DILUPBiCG: Solving for Ux, Initial residual = 0.506818, Final residual = 0.00973123, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 0.603294, Final residual = 0.0122224, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 0.280386, Final residual = 0.00518198, No Iterations 2
GAMG: Solving for p, Initial residual = 0.0695246, Final residual = 0.000686704, No Iterations 30
time step continuity errors : sum local = 1.41992e+43, global = 4.71133e+41, cumulative = 5.39147e+41
DILUPBiCG: Solving for epsilon, Initial residual = 3.87187e-11, Final residual = 3.87187e-11, No Iterations 0
DILUPBiCG: Solving for k, Initial residual = 1, Final residual = 0.0894909, No Iterations 1
bounding k, min: -7.27376e+51 max: 5.05817e+89 average: 5.17561e+84
ExecutionTime = 47386.5 s ClockTime = 48159 s

Time = 761

DILUPBiCG: Solving for Ux, Initial residual = 0.417362, Final residual = 0.0145424, No Iterations 1
DILUPBiCG: Solving for Uy, Initial residual = 0.51114, Final residual = 0.0224704, No Iterations 1
DILUPBiCG: Solving for Uz, Initial residual = 0.376437, Final residual = 0.0288099, No Iterations 1
GAMG: Solving for p, Initial residual = 0.629327, Final residual = 0.00528483, No Iterations 8
time step continuity errors : sum local = 1.11456e+78, global = 4.87405e+75, cumulative = 4.87405e+75
DILUPBiCG: Solving for epsilon, Initial residual = 4.23728e-09, Final residual = 4.23728e-09, No Iterations 0
#0 Foam::error::printStack(Foam::Ostream&) in "/home/jish/OpenFOAM/OpenFOAM-1.6.x/lib/linux64GccDPOpt/libOpenFOAM.so"
#1 Foam::sigFpe::sigFpeHandler(int) in "/home/jish/OpenFOAM/OpenFOAM-1.6.x/lib/linux64GccDPOpt/libOpenFOAM.so"
#2 ?? in "/lib/libc.so.6"
#3 Foam::PBiCG::solve(Foam::Field<double>&, Foam::Field<double> const&, unsigned char) const in "/home/jish/OpenFOAM/OpenFOAM-1.6.x/lib/linux64GccDPOpt/libOpenFOAM.so"
#4 Foam::fvMatrix<double>::solve(Foam::dictionary const&) in "/home/jish/OpenFOAM/OpenFOAM-1.6.x/lib/linux64GccDPOpt/libfiniteVolume.so"
#5 Foam::lduMatrix::solverPerformance Foam::solve<double>(Foam::tmp<Foam::fvMatrix<doubl e> > const&) in "/home/jish/OpenFOAM/OpenFOAM-1.6.x/lib/linux64GccDPOpt/libincompressibleRASModels.so"
#6 Foam::incompressible::RASModels::kEpsilon::correct () in "/home/jish/OpenFOAM/OpenFOAM-1.6.x/lib/linux64GccDPOpt/libincompressibleRASModels.so"
#7 main in "/home/jish/OpenFOAM/OpenFOAM-1.6.x/applications/bin/linux64GccDPOpt/simpleFoam"
#8 __libc_start_main in "/lib/libc.so.6"
#9 _start at /build/buildd/eglibc-2.10.1/csu/../sysdeps/x86_64/elf/start.S:116
Floating point exception




thanks

regards
jish

Hi everybody,

i am new in openFoam and at the moment also trying to make a low-Reynolds-Model work.
I use the LamBremhorstKE with the pisoFoam solver. I think that k and epsilon explode, and i get this error message:








/*---------------------------------------------------------------------------*\
| ========= | |
| \\ / F ield | OpenFOAM: The Open Source CFD Toolbox |
| \\ / O peration | Version: 1.6 |
| \\ / A nd | Web: www.OpenFOAM.org |
| \\/ M anipulation | |
\*---------------------------------------------------------------------------*/
Build : 1.6-f802ff2d6c5a
Exec : pisoFoam
Date : Apr 12 2010
Time : 17:38:30
Host : itb15
PID : 13551
Case : /CFD/Moritz/ahmed_low_re_2
nProcs : 1
SigFpe : Enabling floating point exception trapping (FOAM_SIGFPE).

// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
Create time

Create mesh for time = 0

Reading field p

Reading field U

Reading/calculating face flux field phi

Selecting incompressible transport model Newtonian
Selecting turbulence model type RASModel
Selecting RAS turbulence model LamBremhorstKE
--> FOAM Warning :
From function GeometricField<Type, PatchField, GeoMesh>::readIfPresent()
in file /CFD/Daniel/src-openfoam/install/OpenFOAM-1.6/src/OpenFOAM/lnInclude/GeometricField.C at line 107
read option IOobject::MUST_READ suggests that a read constructor for field epsilon would be more appropriate.
^[[5;5~LamBremhorstKECoeffs
{
Cmu 0.09;
C1 1.44;
C2 1.92;
alphaEps 1.3;
}


Starting time loop

Time = 1e-07

Courant Number mean: 0 max: 0
DILUPBiCG: Solving for Ux, Initial residual = 1, Final residual = 1.31903e-16, No Iterations 1
DILUPBiCG: Solving for Uy, Initial residual = 0, Final residual = 0, No Iterations 0
DILUPBiCG: Solving for Uz, Initial residual = 0, Final residual = 0, No Iterations 0
DICPCG: Solving for p, Initial residual = 1, Final residual = 0.0978004, No Iterations 297
time step continuity errors : sum local = 5.58285e-08, global = -1.88251e-10, cumulative = -1.88251e-10
DICPCG: Solving for p, Initial residual = 0.0402446, Final residual = 9.95978e-07, No Iterations 367
time step continuity errors : sum local = 8.99511e-11, global = 1.63882e-14, cumulative = -1.88235e-10
#0 Foam::error::printStack(Foam::Ostream&) in "/usr/local/OPENFOAM/OpenFOAM-1.6/lib/linux64GccDPOpt/libOpenFOAM.so"
#1 Foam::sigFpe::sigFpeHandler(int) in "/usr/local/OPENFOAM/OpenFOAM-1.6/lib/linux64GccDPOpt/libOpenFOAM.so"
#2 ?? in "/lib64/libc.so.6"
#3 Foam::divide(Foam::Field<double>&, Foam::UList<double> const&, Foam::UList<double> const&) in "/usr/local/OPENFOAM/OpenFOAM-1.6/lib/linux64GccDPOpt/libOpenFOAM.so"
#4 void Foam::divide<Foam::fvPatchField, Foam::volMesh>(Foam::GeometricField<double, Foam::fvPatchField, Foam::volMesh>&, Foam::GeometricField<double, Foam::fvPatchField, Foam::volMesh> const&, Foam::GeometricField<double, Foam::fvPatchField, Foam::volMesh> const&) in "/usr/local/OPENFOAM/OpenFOAM-1.6/lib/linux64GccDPOpt/libincompressibleRASModels.so"
#5 Foam::tmp<Foam::GeometricField<double, Foam::fvPatchField, Foam::volMesh> > Foam::operator/<Foam::fvPatchField, Foam::volMesh>(Foam::tmp<Foam::GeometricField<doub le, Foam::fvPatchField, Foam::volMesh> > const&, Foam::GeometricField<double, Foam::fvPatchField, Foam::volMesh> const&) in "/usr/local/OPENFOAM/OpenFOAM-1.6/lib/linux64GccDPOpt/libincompressibleRASModels.so"
#6 Foam::incompressible::RASModels::LamBremhorstKE::c orrect() in "/usr/local/OPENFOAM/OpenFOAM-1.6/lib/linux64GccDPOpt/libincompressibleRASModels.so"
#7 main in "/usr/local/OPENFOAM/OpenFOAM-1.6/applications/bin/linux64GccDPOpt/pisoFoam"
#8 __libc_start_main in "/lib64/libc.so.6"
#9 __gxx_personality_v0 in "/usr/local/OPENFOAM/OpenFOAM-1.6/applications/bin/linux64GccDPOpt/pisoFoam"
Gleitkomma-Ausnahme







I also ran the model with a coarser mesh, there it started calculating but k and epsilon went to -e12 to e12 and it finally crashed.
by the way, is it allowed that those values become negative?


Does anybody know if the LamBremhorst-Model is working in the actual version?

If yes, could you give me a hint, which initial conditions you use?
As boundary cond. i use

epsilon zeroGradient
k fixed value 0.000000001
nut calculated

all surfaces are defined as patch.

it would be great if somebody could explain experiences maybe already made.

all the best,
Moritz

Florian, Thomas,

after reporting a bug the Problem has been fixed in 1.6.x. Since I sometimes still use 1.5-dev I am wondering if the problem with yPlusRAS exists there, too. As far as I see - yes. Right? If yes, Florian could you please send me your fixed version? Thanks.

Regards Bastian

Edit: To make it clear: My older post was generally right (saying that yplus should be calculated identially for all turbulence models). The fact ist that yPlusRAS calculates a value that FLUENT refers to as ySTAR:
http://www.cfd-online.com/Forums/ope...lus-ystar.html
yPLUS is defined here:
http://www.cfd-online.com/Wiki/Dimensionless_wall_distance_(y_plus) and is independant of turbulence modelling (however, you have to get the wall shear stress right)

Hi Thomas/Florian/Bastil,

I am having the same problem as Thomas had. If it is possible would you mind make the code to calculate yPlusRAS publicly available?

Thanks and regards,
Robert

Hi guys!

I am sorry but I have no access to any data at the moment (vacation!)

Thomas, didn't I send you my utility to calculate y-plus for low-Re RANS models once. If so, maybe you could send it to Robert.

Since so many prople are asking for it, I will review it and make it public after my vacation in mid of March.

Best,
Florian

Hej,

thank you florian, that would just be great.

For me it does not make any sense to set k and epsilon to zeroGradient at the wall. I recently had a look in the original paper and they use k = epsilon = 0 there, which is conform to my understanding of a low-Re-model so far.
Concerning the turbulent viscosity mut or nut, I have tested a calculated as well as a zeroGradient boundary condition. There was no difference difference in both calculations.

Just for a comparison I have tested k and epsilon with a zeroGradient bc. This results in an unphysical thickening of the boundary layer.

Sebastian

Hi,
I agree with you about the value of k and epsilon near the wall. If the reynolds is low then the value of k and espilon should be low too.
However in the sonicFoam/nacaAirfoil tutorial (compressible case) k and epsilon are have a high fixed value (several thousands).
In incompressible/boundaryFoam/boundaryLaunderSharma (incompressible case) k and epsilon have a very low value at the wall (around 1e-10 for both).

I just don't understand on what is based this difference, and why this is different for compressible and incompressible fluid.

Hi malaboss,

Do you find something wich can explain why there is this difference?

Valentin

Well,
I just watch the boundary conditions of both cases (nacaairfoil and boundaryLaunderSharma) which use LaunderSharma.
In nacaAirfoil, wall function are used while it should be low Re wall function or nothing at all in a low Re model. This could explain the high values of k at wall, and as a conclusion, Naca Airfoil would not be a good example to understand which boundary conditions we have to set for a Low Re model.
See : http://www.cfd-online.com/Wiki/Low-R...els#References for boundary conditions

Be careful though, I may be wrong since I don't know how you use a wall function at the wall for a compressible case like naca airfoil. I tried to see how wall fucntion are implemented in Launder and Sharma models, but all I found was this : http://www.cfd-online.com/Forums/ope...k-e-model.html which states that Launder and Sharma was conceived for incompressible cases. About the general implementation of wall function in compressible cases, it seems like there is no big difference with the incompressible. http://jullio.pe.kr/fluent6.1/help/html/ug/node451.htm
I did not manage to see the y+ value of the naca airfoil case as I could not run the case.

Please tell me if you find something new and more relevant !

Thank you for your help malaboss :)
I don't know if launder-sharma is correct in the compressible cases but it is in the list of tubulent models available in compressible.
I will give a try to this model in my case and will say if it's ok.

Do you know the boundary conditions for nut and alphat that I have to give at walls? I found posts wich say zeroGradient but I'm not sure about this.

If we keep on reasoning as we did just before, nut should be set to zero (fixed value) on the wall. There is no turbulence so there is no turbulence viscosity, that is, no viscosity due to turbulence.

For alpha t (again i'm not an expert for compressible flows) it represents the thermal diffusivity, proportionnal to thermal conductivity. Near the wall we have nearly no turbulence, so no convection, hence no conductivity. To me, you can fix the value to zero for alpha t.


Where did you find the Boundary conditions you are talking about, as it doesn't match my thoughts ?

I agree with you. According to me zeroGradient doesn't have any sense.

Hi, I have simulated impinging jet before and it succeed even without initialization
1) What is the turbulence model in openfoam used? Is it a Low-Re or high-Re model.
My suggestion would be Low-Re model as impinging jet involves complex physics that the standard wall function approach might not be able to predict well. From you initial condition settings (epsilonWallFunction, kqRWallFunction, nutSpalartAllmarasWallFunction you are using a high-Re model right?) I tried modelling in fluent impinging jet axisymmetric - use realisable KE and enhanced wall treatment. I attended a course by fluent that taught us that for complex physics it is better to use enhanced wall treatment. In OpenFOAM, there is no equivalent enhanced wall treatment in the Hi-Re Realisable ke model so another way to get around this is to use Low-Re model such as laundersharmaKE which contains damping functions to resolve boundary layer flow. Some people suggest using nut_wall: nutSpalartAllmarasWallFunction or nutSpalartAllmarasStandardWallFunction to make a Hi-Re model looks like low-Re model. I do not think this is right way to do. I have read the thread somewhere on this matter. So strongly recommend to try LaunderSharmaKE.

2) How does your geometry looks like? can you draw out the problem?

3) What are your fvSchemse and fvSolution? Sometimes these are the things that cause continuity to blow up.

4) For nut, is there any reason you use nutSpalartAllmarasStandardWallfunction over nutkWallFunction if using Hi-Re model? Any advantage?