CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Low-Re turbulence Model without wall-functions (http://www.cfd-online.com/Forums/fluent/100169-low-re-turbulence-model-without-wall-functions.html)

bruce April 21, 2012 11:56

Low-Re turbulence Model without wall-functions
 
Dear All,

I am using Realizable k-epsilon low Re. turbulence model for a simple test case with y+~1-3. By default, the standard wall functions are selected in GUI. I want to disable all the wall functions because my grid resolves the boundary layer. I have tried fluent commands to disable by selection, but it seems it could not work as it is getting selected again automatically.

Although, it is better to include wall functions even for Low-Re simulations, I need to disable it for validation purpose.

Does any body help?

Thank You

LuckyTran April 21, 2012 17:14

Quote:

Originally Posted by bruce (Post 356022)
Dear All,

I am using Realizable k-epsilon low Re. turbulence model for a simple test case with y+~1-3. By default, the standard wall functions are selected in GUI. I want to disable all the wall functions because my grid resolves the boundary layer. I have tried fluent commands to disable by selection, but it seems it could not work as it is getting selected again automatically.

Although, it is better to include wall functions even for Low-Re simulations, I need to disable it for validation purpose.

Does any body help?

Thank You

hi bruce, if your grid is sufficiently fine then the standard wall function approach should only be resorting to the low-Re treatment. Doesn't that already do what you need it to do? Fluent uses y* ~ 11.225 as a switch between low-Re (linear) and high-Re (log) approximation. If your y* is about the same as y* but since your y+ is <3, y* is also approximately ~<3. Also, the enhanced wall function approach uses blending between the two.

You cannot run a simulation without some type of wall function, how else do you apply your wall boundary conditions?

bruce April 23, 2012 12:01

Quote:

Originally Posted by LuckyTran (Post 356102)
hi bruce, if your grid is sufficiently fine then the standard wall function approach should only be resorting to the low-Re treatment. Doesn't that already do what you need it to do? Fluent uses y* ~ 11.225 as a switch between low-Re (linear) and high-Re (log) approximation. If your y* is about the same as y* but since your y+ is <3, y* is also approximately ~<3. Also, the enhanced wall function approach uses blending between the two.

hello Lucky, this is fine.

Quote:

Originally Posted by LuckyTran (Post 356102)
You cannot run a simulation without some type of wall function, how else do you apply your wall boundary conditions?

What i was looking was a user specified boundary conditions (for all field variables) at the wall instead evaluating from wall functions. In this case, we do not need any wall functions.is that possible?

Thanks

LuckyTran April 23, 2012 12:44

Quote:

Originally Posted by bruce (Post 356425)
hello Lucky, this is fine.

What i was looking was a user specified boundary conditions (for all field variables) at the wall instead evaluating from wall functions. In this case, we do not need any wall functions.is that possible?

Thanks

Bruce, a user specified wall boundary condition is in some sense still a wall function. What I mean is that you cannot specify boundary conditions without specify variables at (at least) some cells near the wall, thus making all methods of applying wall boundary conditions essentially wall functions (in some sense). Regardless of how well your grid resolves the boundary layer you still need to apply the boundary conditions this may be a wall or symmetry or other but it is still a boundary condition and has really not much to do with how well resolved your grid is. The boundaries do not care if there is even a grid at all.

You just need to distinguish what types of variables you like to apply at cells near the wall and only change the ones that Fluent does not do exactly the way you want it to do. So ask yourself, exactly what type of wall functions are you trying to implement? And does Fluent already do that or no?

bruce April 26, 2012 14:37

Dear Lucky

now i understand better. I have one more question.

As said in the user guide of fluent, is epsilon really computed at the near wall cells using a algebraic relation? I meant the equation no. (10.8-10) in the below link.

http://jullio.pe.kr/fluent6.1/help/html/ug/node451.htm

Because, when i calculate the value of epsilon near the wall i get different results,

for example

Cmu = 0.09
kP = 0.005
kappa = 0.41
yp = 1.92e-05

epsilon = 7.38

But, i found fluent reports it as 0.88. I wonder where does this difference comes from.

Do you have any idea for this?

Thank You

LuckyTran April 26, 2012 15:00

Quote:

Originally Posted by bruce (Post 357276)
Dear Lucky

now i understand better. I have one more question.

As said in the user guide of fluent, is epsilon really computed at the near wall cells using a algebraic relation? I meant the equation no. (10.8-10) in the below link.

http://jullio.pe.kr/fluent6.1/help/html/ug/node451.htm

Because, when i calculate the value of epsilon near the wall i get different results,

for example

Cmu = 0.09
kP = 0.005
kappa = 0.41
yp = 1.92e-05

epsilon = 7.38

But, i found fluent reports it as 0.88. I wonder where does this difference comes from.

Do you have any idea for this?

Thank You

Are you using standard wall functions or enhanced wall functions? If enhanced, did you take the blending/damping function into account?

That formula you referenced is correct for the low-Re approach (two-layer approach or standard wall function approach). For enhanced, there is a damping term.

Most importantly, in the realizable-k-e model, Cmu is no longer a constant. Cmu is constant only for the standard k-e model. Did you take the variation of Cmu into account? I noticed you took Cmu as 0.09, this is the constant Cmu from the standard k-e approach.

bruce April 26, 2012 15:12

Dear Lucky,

Sorry i forgot to mention that i used standard wall function. Since i had validation problems with enhanced wall function, i change it to standard k-epsilon model with standard wall treatment. And i switched off energy and make it constant density.

And as i calculated before using the relation for epsilon, i do not know why fluent result is 0.88 instead of 7.38 (at one particular cell). Do fluent use any other relation to calculate epsilon near wall? Do you see any problem?

Thank You

LuckyTran April 26, 2012 15:57

Quote:

Originally Posted by bruce (Post 357280)
Dear Lucky,

Sorry i forgot to mention that i used standard wall function. Since i had validation problems with enhanced wall function, i change it to standard k-epsilon model with standard wall treatment. And i switched off energy and make it constant density.

And as i calculated before using the relation for epsilon, i do not know why fluent result is 0.88 instead of 7.38 (at one particular cell). Do fluent use any other relation to calculate epsilon near wall? Do you see any problem?

Thank You

I see. In that case it is odd indeed. In that case Cmu of 0.09 is correct. yp is the distance to the nearest wall (m) correct?

Did you also check the calculation of turbulent kinetic energy? It probably is wrong also if dissipation is wrong. What is fluent reporting as the turbulent kinetic energy? While you are at it, also compare the production term, the production term should be approximately the same as the dissipation.

Additionally, there were no convergence problems correct?

There are many other ways to solve for dissipation near the walls but I had thought that those were the equations used by Fluent. I will look into it more.

bruce April 26, 2012 16:07

Yes, yP is the cell center to boundary distance as by definition (half of the cell height in my case).

The turbulent kinetic energy calculation is correct indeed. turbulent kinetic energy = Cmu * k^2 / epsilon. When i calculate manually that i get correct value as done by fluent. I will check the production term soon.

And, since it is a flow over a flat plate, the convergence is very fast (1e-07 in 1500 iterations with second order upwind).

Indeed it looks odd.

Thanks

LuckyTran April 26, 2012 16:10

Quote:

Originally Posted by bruce (Post 357286)
Yes, yP is the cell center to boundary distance as by definition (half of the cell height in my case).

The turbulent kinetic energy calculation is correct indeed. turbulent kinetic energy = Cmu * k^2 / epsilon. When i calculate manually that i get correct value as done by fluent. I will check the production term soon.

And, since it is a flow over a flat plate, the convergence is very fast (1e-07 in 1500 iterations with second order upwind).

Indeed it looks odd.

Thanks

You can perhaps monitor the dissipation vs iterations just to double check the convergence but I doubt that is the case. Convergence of k and epsilon are usually very fast relative to momentum equation, that is why their under-relaxation factors are set so high.

bruce April 26, 2012 16:12

forgot to mention that epsilon was taken as 0.88 what was used by fluent.

bruce April 27, 2012 18:04

Hi Lucky,

I was very unlucky. The epsilon relation was indeed correct. The real problem was i always used paraview to visual the fluent results, for some reason data translation went wrong:confused: (may be badly interpolated) So i had used wrong kinetic energy to calculate epsilon. Now it is fine.

Thank You

LuckyTran April 27, 2012 21:04

Quote:

Originally Posted by bruce (Post 357577)
Hi Lucky,

I was very unlucky. The epsilon relation was indeed correct. The real problem was i always used paraview to visual the fluent results, for some reason data translation went wrong:confused: (may be badly interpolated) So i had used wrong kinetic energy to calculate epsilon. Now it is fine.

Thank You

hmm, must be very unlucky indeed. yes paraview is a decent software for visualization. must have made a mistake somewhere during the translation. but all is well!


All times are GMT -4. The time now is 15:37.