CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Pre-Processing (http://www.cfd-online.com/Forums/openfoam-pre-processing/)
-   -   defining of turbulence coefficient boundary condition on the wall (http://www.cfd-online.com/Forums/openfoam-pre-processing/117896-defining-turbulence-coefficient-boundary-condition-wall.html)

 immortality May 17, 2013 00:44

defining of turbulence coefficient boundary condition on the wall

hi
How can set k-omega or epsilon on the wall?
How can find out what values should be put on the wall?
I have seen some inlet flow BC explanation and formula in the net but how is it defined on the wall?

 immortality June 21, 2013 13:49

Hi
I answer my question now for future users:
we have two models:
low-Re and high-Re
in low-Re we use k:1e-12(or a very low value),for omega and epsilon two opinions there are:
a)very low values like k so that wall functions be disabled near the wall
or
b)very high values as Wilcox book says the approximated formula(doesn't need to be very accurate,for example 10000 for omega and 100000 for epsilon)...
continued...
if anyone like this tell me to continue the subject.

 immortality June 25, 2013 18:24

hi Elizabeth
If you have any question or just want to know more about this subject don't hesitate to tell me.maybe I know !

 immortality June 26, 2013 12:21

Quote:
 I would like to discuss Ehsan, but still have difficulties to express the questions I have. At the moment I mix up some topics.
Hi Lisa
ask what you want here.I told before don't hesitate.
why you have mix up with questions as you told in private?ask here all things.
which information do you need?

 aylalisa June 26, 2013 14:53

low-Re models

Hi Ehsan,

experimental setup: flow channel
fluid: water
L= 3,1m (0m-1m) (x-axis)
L = 1,0m (developing part of canal)
L = 1,5m (heating from bottom)
L = = 0,6m (final part of canal)
W=0,250m (y-axis)
H=0,02m (z-axis)
fluid = water
Environmental temperature differs from heated canal bottom temperature very less (~ 5K)
Fluid flow velocity is very small: 0,009m/s

To save cells, I've done numerical model setup without developing part of canal. Instead I've used the fully developed flow profile from other simulation (same geometrical setup). Apart from that I've defined a symmetry plane in the middle of the canal: y=0,125
Current setup: 4e06 cells

I hope to see buoyancy effects that might entail turbulence.

1) Which solver to use?
2) Which turbulence model to use?

I don't want to do a DNS since I would need a very fine mesh and I think I would not be allowed to work with a symmetry plane in that case --> over again more cells.

Makes it sense to use a low-Reynolds model in that case?
- LaunderSharmaKE
- LamBremhorstKE
- or better k--SST-model?

Is k--SST a low Re-model as well?

Does it mean, in case I use one of these models, that I solve the near wall region but still have to implement corresponding BCs?

Do low-Re models use damping functions whereas k--SST does not? Where is the difference?
Are these models so called hybrid wall functions?

Do your suggested boundary conditions go along with, for example, LaunderSharmaKE?

k, , k- : very small values
= nutLowReWallFunction

Quote:
 Originally Posted by immortality (Post 435329) Hi I answer my question now for future users: we have two models: low-Re and high-Re in low-Re we use k:1e-12(or a very low value),for omega and epsilon two opinions there are: a)very low values like k so that wall functions be disabled near the wall or b)very high values as Wilcox book says the approximated formula(doesn't need to be very accurate,for example 10000 for omega and 100000 for epsilon)... continued... if anyone like this tell me to continue the subject.
Solver-problem:
I am still not sure which solver to use.
- buoyantBoussinesqPimpleFoam
- buoyantBoussinesqSimpleFoam

Did Ivan decide for piso because of that air bubble he wanted to see?
Sorry for that stupid question, but I would like to understand.

Please correct me if I am wrong:
If I am not interested in the development of the flow structure I should possibly use bBSimpleFoam.
Since I map a laminar flow profile from another simulation without heated bottom plate as a boundary condition in this case I could use bBPimpleFoam to watch the development of the flow structure (--> influence of buoyancy effects on laminar flow in the beginning phase)
If am only interested in the final flow structure, in case a stationary flow state exists, simple should do. --> ???

Ferziger/Peric: 9.4 RANS-Models/ 9.4.1 (P. 345)
http://i.imgur.com/TABhtTF.png

I found these graphs in almost every CFD book, but did not completely understand.
My simple idea is to use a steady-state solver for a system whose state variables behave like left pic. shows, whereas for a transient problem (air bubble?, development of convection rolls?), right pic. describes the situation. I would decide for pimple in such a case.

Lisa

 nimasam June 26, 2013 16:19

Quote:
 I found these graphs in almost every CFD book, but did not completely understand. My simple idea is to use a steady-state solver for a system whose state variables behave like left pic. shows, whereas for a transient problem (air bubble?, development of convection rolls?), right pic. describes the situation. I would decide for pimple in such a case.
you are right ;), if the mean is steady, you can use steady state solver (simple in OpenFOA), but when the mean is transient you should use transient solver (piso, or pimple in OpenFOAM)
also buoyantBoussinesqPimpleFoam can be a promising choice because you can see both the transient state and final steady-state ;) but it may takes longer than using an steady state solver
im not expert in turbulence, so no judge , but alittle description of each turbulence model can be find in CFD book such as versteeg which may help you

 immortality June 26, 2013 16:47

Quote:
 I don't want to do a DNS since I would need a very fine mesh and I think I would not be allowed to work with a symmetry plane in that case --> over again more cells. Makes it sense to use a low-Reynolds model in that case? - LaunderSharmaKE - LamBremhorstKE - or better k--SST-model? Is k--SST a low Re-model as well? Does it mean, in case I use one of these models, that I solve the near wall region but still have to implement corresponding BCs? Do low-Re models use damping functions whereas k--SST does not? Where is the difference? Are these models so called hybrid wall functions? Do your suggested boundary conditions go along with, for example, LaunderSharmaKE? k, , k- : very small values = nutLowReWallFunction
when you told you are writing many questions I didn't imagine be so much text! its better to separate your questions and propound them in different forums.
thanks to Nima that helped with the flow part!
yes you can use each one of models you mentioned.k-OmegaSST can be used in low-Re simulations in incompressible cases(that seems you have)but not in compressible ones.
in incompressible all three models use damping functions AFAIK.and if you want to model low-Re set BC's low enough(near zero,it should be zero in fact) although I think in incompressible you better use:k:1e-12 or -18 or any other low value but very high value for omega and epsilon like 100000 and 1000000(see the approximate formulas in Wilcox book for better guess depending on your case).test them.and if the values be in correct range its not so important their values.solver replaces more correct values later AFAIK.
just test on a simple case at first.
and the BC's are necessary for numerical scheme be started in a good way.

 immortality June 27, 2013 07:45

1 Attachment(s)
well I write some lines,may be useful for someone!
in low-Re simulations as you know y+ should be less than 1 so that laminar sub-lyer(viscous layer) in turbulent boundary layer can be good resolved and be exactly into account(the term is calculated exactly from the nearest(first) cell to the wall)
then in low-Re model a fine mesh with y+<1 should be considered and its valuable to noted that it shouldn't goes too little too so the range (.2<=y+<=1) seems OK.
no need to say that you should first run the case and after that calculate y+ and if not in the range modify the mesh and repeat the run again till reach the y+ range.
IMPORTANT:
yPlusRAS tool in OF uses wall functions,so isn't suitable for low-Re model(it calculates in fact y* not y+,wrong name),you can find appropriate tools for more accurate y+ by searching(or a good modified for 2.2.0 version I have attached)
before all things you should be certain that low-Re models are good for you,because it consumes a lot of time!
Only use loe-Re models when you want to model EXACTLY near wall phenomena,mean when you have large adverse gradient pressure and separatin regions and want to know where separation has been started on the wall PRECISELY.
well I think thats all for low_Re
I will continue with high-Re maybe next.

 aylalisa July 4, 2013 09:27

lot of time - Low Re Models

Hallo Ehsan,

what do you understand by 'consumes a lot of time'. Could you give an example?
Maybe some simple geometry, # of cells --> time? (in case we use a low Re model)

Lisa

 immortality July 4, 2013 09:44

Hi Eli
I tried to do low-Re in a geometry like shock tube but it took a lot of time in compare to high-Re and I stopped that!because I found out that I didn't need that much accuracy near the walls.

 aylalisa July 4, 2013 10:18

I try desperately to simulate convection rolls in the canal (--> small T changes in the fluid). The canals height is very small compared to its lengths and width. Therefore I've decided to finally work with 'high' mesh resolution to be able using one of the low Re Models (LaunderSharmaKE). The simulation converged after 2776 iterations! But the results are still totally wrong :mad:.
I've used the utility you've sent me to check y+ values. y+ values are beyond a reasonable value and I guess that the boundary conditions for and k aren't good as well.

Till now I can not yet provide useful experiences :(.

 aylalisa July 4, 2013 11:15

plusPostRANS

Made a stupid mistake. Instead of 'plusPostRANS' I've started 'YplusRAS'. Now I've tried to use 'plusPostRANS', but the compiler reported an error message:

Code:

plusPostRANS.C:96:40: error: ‘class Foam::tmp<Foam::GeometricField<double, Foam::fvPatchField, Foam::volMesh> >’ has no member named ‘boundaryField’ plusPostRANS.C:99:36: error: ‘class Foam::tmp<Foam::GeometricField<double, Foam::fvPatchField, Foam::volMesh> >’ has no member named ‘boundaryField’ make: *** [Make/linux64GccDPOpt/plusPostRANS.o] Fehler 1
:confused:

 immortality July 4, 2013 11:21

whats 'plusPostRANS' at all?!:D

 aylalisa July 4, 2013 11:34

oha

:o, not the one you've sent me :confused:

 aylalisa July 4, 2013 11:39

:eek::):p

--> yPlus220
The other one (plusPostRANS) is from another one.

 immortality July 4, 2013 11:47

could you share the new utility?what does it do?

 aylalisa July 4, 2013 12:09

I've found that one in the forum:

http://www.cfd-online.com/Forums/ope...-testcase.html

compilation of 'yPlus220' :

Code:

yPlus.C:36:63: fatal error: nutWallFunction/nutWallFunctionFvPatchScalarField.H: Datei oder Verzeichnis nicht gefunden compilation terminated.
The smiley I need is not available.....should better start to learn C++. Thought I could skip it.

 All times are GMT -4. The time now is 01:26.