
[Sponsors] 
Problems with boundary conditions for a lowRekOmegaSST turbulence model 

LinkBack  Thread Tools  Display Modes 
August 27, 2008, 10:31 
Hi,
I have extended the kOmeg

#1 
Member
Markus Weinmann
Join Date: Mar 2009
Location: Stuttgart, Germany
Posts: 77
Rep Power: 9 
Hi,
I have extended the kOmegaSST turbulence model in OF1.5 by a low Reynolds number correction. The model extension was straightforward, however, I am having problems specifying appropriate wall boundary condition for "omega". In the low Reynolds number version the wall boundary condition for omega is a function of the laminar viscosity and the distance to the first grid point. As a consequence, I cannot use the "standard" zeroGradient and fixedValue boundary conditions for omega at the wall. My first idea was to create a new boundary condition which computes and specifies omega at the wall. I already had a look at the default fixedValue and some other derived boundary conditions. Since I am still a beginner in OpenFoam, my knowledge is not sufficient to adapt any of the existing boundary condition for my purposes. Could anybody please give me some guidlines with which of the existing boundary condition I should start with and what are the important lines that need to be changed. Thanks Markus 

August 27, 2008, 11:33 
Hi Markus
Search the forum

#2 
Senior Member
Niels Gjoel Jacobsen
Join Date: Mar 2009
Location: Deltares, Delft, The Netherlands
Posts: 1,713
Rep Power: 27 
Hi Markus
Search the forum for a thread called "Calculation of Pressure Loss". In there is a discussion of boundary conditions for omega over rough boundaries. Even though Wilcox states that the boundary condition should work on smooth boundaries I have not had any good experience with it. In the same thread there is also a implementation of the Wilcox rough boundary. This you could use to get inspired. Best regards, Niels
__________________
Please note that I do not use the Friendfeature, so do not be offended, if I do not accept a request. 

August 27, 2008, 15:47 
Hi Niels
Thanks a lot  I w

#3 
Member
Markus Weinmann
Join Date: Mar 2009
Location: Stuttgart, Germany
Posts: 77
Rep Power: 9 
Hi Niels
Thanks a lot  I will have a look at it. Markus 

August 28, 2008, 07:14 
Hi again,
I have now solved

#4 
Member
Markus Weinmann
Join Date: Mar 2009
Location: Stuttgart, Germany
Posts: 77
Rep Power: 9 
Hi again,
I have now solved the problem of specifying the boundary conditions! (using: omega_boundaryField()[patchi]== X and omega_.correctBoundaryConditions()). However, I discovered some unusual bahaviour which I cannot explain: In order to check that I am using the correct values to compute the boundary conditions, I have printed the values for the wall distance and the laminar viscosity on the wallpatch and in the cells attached to the wallpatch. During this check, I did not compute new boundary conditions for omega at the wall. As a result, I get incorrect values for the wall distance and the lam. viscosity in the first grid cell in exactly two cells, all other wallcells have the correct value for the wall distance and lam. viscosity. Also, in the first grid cell, nu() has changed, even though, I am using incompressibleNewtoniansimpleFoam. The values of nu() in the same two cells as before are again completely unreasonable. When I include the computation of wall boundary conditions for omega, the situation becomes even worse. Now, only the values at wallpatch are correct and the values in the cells are all incorrect. Does anybody know what is happening here. Markus WITHOUT omega_boundaryField()[patchi]== X and omega_.correctBoundaryConditions() included: wallpatch dy: 1e15  wallcell dy: 1.15459e312  wallpatch nu: 1.461e05  wallcell nu: 1.15459e312 wallpatch dy: 1e15  wallcell dy: 1.15459e312  wallpatch nu: 1.461e05  wallcell nu: 1.15459e312 wallpatch dy: 1e15  wallcell dy: 2.5e05  wallpatch nu: 1.461e05  wallcell nu: 2.5e05 wallpatch dy: 1e15  wallcell dy: 2.5e05  wallpatch nu: 1.461e05  wallcell nu: 2.5e05 ..... WITH omega_boundaryField()[patchi]== X and and omega_.correctBoundaryConditions()included: wallpatch dy: 1e15  wallcell dy: 1.15459e312  wallpatch nu: 1.461e05  wallcell nu: 1.15459e312 wallpatch dy: 1e15  wallcell dy: 4.79861e317  wallpatch nu: 1.461e05  wallcell nu: 4.79861e317 wallpatch dy: 1e15  wallcell dy: 6.25e10  wallpatch nu: 1.461e05  wallcell nu: 6.25e10 wallpatch dy: 1e15  wallcell dy: 6.25e10  wallpatch nu: 1.461e05  wallcell nu: 6.25e10 .... 

August 28, 2008, 14:53 
The dy values are not set it s

#5 
Super Moderator
Mattijs Janssens
Join Date: Mar 2009
Posts: 1,416
Rep Power: 17 
The dy values are not set it seems. Compile the code with DFULLDEBUG g O0 and run through valgrind.


September 1, 2008, 03:37 
Ok thanks again  I will run t

#6 
Member
Markus Weinmann
Join Date: Mar 2009
Location: Stuttgart, Germany
Posts: 77
Rep Power: 9 
Ok thanks again  I will run through valgrind and see what happens.


September 3, 2008, 10:40 
Hi
would you mind to explai

#7 
Senior Member

Hi
would you mind to explain different of the boundary layer calculating method for OF and CFX of Mentor`s.and what about the Markus` low Reynolds number correction ?is there any reference paper of the SST in OF and Markus` low Reynolds number correction ? you see i am calculate the torque for the blade of pump impeller,but with the flow separation the standard wall function of loglaw is not suitable for this calculation as i have tried,and made the analysis by RE method. i found a huge error of this modeling method not numerical works. would you mind to give me some good idea? wayne 

September 3, 2008, 12:01 
Hi
I don't know anything ab

#8 
Member
Markus Weinmann
Join Date: Mar 2009
Location: Stuttgart, Germany
Posts: 77
Rep Power: 9 
Hi
I don't know anything about the SST model in CFX. The SST model in OF uses a wall function approach. In my opinion, wall function may work ok for very simple flows. If you want to simulate more complex flows (e.g. flow separation under adverse pressure gradient) I would always fully resolve the boundary layer up to the wall. The SST model in its original formulation is derived for high Reynolds numbers. This however does not mean you have to use wall functions. You still can fully resolve the boundary layer and with that avoid the assumptions inherent in the wall function approach. Since the original SST model is derived for high Re, you cannot expect to get accurate profiles for the turbulent quantities very close to the wall where the flow has a low local Re. If you want "more" accurate profiles for the turbulence quantities close to the wall you need low Re corrections. This only corrects turbulence quantities, the mean velocity profile for high Re flows will be largly unaffected since the turbulent viscosity is small where the low Re corrections becomes active. Maybe that helps a little bit. Markus 

January 22, 2009, 06:21 
Hi Markus,
I'm very intereste

#9 
Senior Member

Hi Markus,
I'm very interested in your approach for LowRe version of komega. I'm calculating the turbulence statistics in the bl of a thick flat plate ( thickness = 3% of the chord), so with y+ 30 I cant resolve allmost nothing of the bl. Can you tell me what to do to modify the actual SST for lowRe correction? Thank you, Ivan 

January 22, 2009, 09:44 
Hi Ivan,
you have to create

#10 
Member
Markus Weinmann
Join Date: Mar 2009
Location: Stuttgart, Germany
Posts: 77
Rep Power: 9 
Hi Ivan,
you have to create a copy of the kOmegaSST.C and reomve everthing that is related to the wallfunctions. In addition you have to take care of the wall boundary condition for omega. This gives you a model which allows integration to the wall. For additional lowRe corrections you can consult the Fluent manual: http://202.185.100.7/homepage/fluent...ug/node432.htm The lowRe corrections are not really necessary unless you want to improve the turbulence quantities close to the wall by a little bit. Regards Markus 

January 22, 2009, 09:48 
Hi Markus,
I'm very intereste

#11 
Senior Member

Hi Markus,
I'm very interested in your approach for LowRe version of komega. I'm calculating the turbulence statistics in the bl of a thick flat plate ( thickness = 3% of the chord), so with y+ 30 I cant resolve allmost nothing of the bl. Can you tell me what to do to modify the actual SST for lowRe correction? Thank you, Ivan 

January 22, 2009, 10:58 
Thank you Markus,
so, I have

#12 
Senior Member

Thank you Markus,
so, I have to cancel any inclusion of "kOmegaWallFunctionsI.H", and correct the wall value of _omega as you do, with a new boundary condition for omega. Can you show an example of this boundary condition? Thank you, Ivan P.S. actually I work with an y+ of about 1.5, in your opinion is good for the model without the wall functions? 

January 22, 2009, 13:27 
you can find the solution in t

#13 
Member
Markus Weinmann
Join Date: Mar 2009
Location: Stuttgart, Germany
Posts: 77
Rep Power: 9 
you can find the solution in the thread called "Calculation of Pressure Loss". In there is a discussion of boundary conditions for omega over rough boundaries.
I tend to use y+<1. But I suggest that you do a sensitivity study for y+. Regards Markus 

February 8, 2009, 01:24 
Hi Markus
would mind tell m

#14 
Senior Member

Hi Markus
would mind tell me what is meaning of : omega_.correctBoundaryConditions() anyway,if i don`t want to use the wall function,how can i give the wall boundary for k and omega ? zeroGradient ? thanks yours wayne 

July 7, 2011, 03:03 

#15 
Senior Member
Florian Krause
Join Date: Mar 2009
Location: Munich
Posts: 103
Rep Power: 9 
Deal all,
the thread is already quite old, so there are maybe others (than Markus) who successfully implemented the lowReynolds version of the komega SST model... I am about to implement the lowReynolds version following (file is too large, but should have free access if you google it): Florian Menter, Jorge Carregal Ferreira, Thomas Eesch, and Brad Konno (2003), The SST Turbulence Model with Improved Wall Treatment for Heat Transfer Predictions in Gas Turbines Everything is clear, except one thing: In the standard (highRe) formulation betaStar is constant with a value of 0.09. In the lowRe formulation betaStar1 is corrected (eqn. 20) and blended with betaStar2 (which is constant) to calculate betaStar (eqn. 10). Now, the blending function F1 (eqn. 3) itself uses betaStar within arg1 (eqn. 5). So, basically, either I overread/misunderstand something trivial here or betaStar is blended by F1 using the belnded betaStar?! How did you guys overcome this? Did you just use the blended betaStar from the previous iteration for the blending function? As always, any help is appreciated. Thanks, Florian 

July 7, 2011, 10:55 

#16 
Senior Member
Florian Krause
Join Date: Mar 2009
Location: Munich
Posts: 103
Rep Power: 9 
I have apparently solved it by myself...
Best, Florian 

November 14, 2011, 05:44 

#17 
New Member
Nikola Mirkov
Join Date: Mar 2009
Location: Belgrade, Serbia
Posts: 18
Rep Power: 9 
Hi! Are you still working on this? I need someone to discuss these things trough. I'm reading the same paper you cited at the momment.
Nikola 

Thread Tools  
Display Modes  


Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Problems initializing a turbulence model  srinath  OpenFOAM Running, Solving & CFD  3  November 27, 2008 11:06 
Problems considering boundary conditions  Morten Andersen  CFX  1  January 18, 2007 10:04 
Convergence problems with SSG turbulence model  Ram  CFX  1  June 19, 2006 10:08 
Problems with particles & boundary conditions  Simon Steinmeyer  FLUENT  0  January 11, 2005 05:22 
Turbulence : kepsilon boundary conditions  Mohamed Sofiane KHELLADI  Main CFD Forum  1  April 25, 2000 20:10 