CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Convergence problems (http://www.cfd-online.com/Forums/cfx/19416-convergence-problems.html)

Olivier Macchion January 10, 2003 09:30

Convergence problems
 
Hello All !

I'm trying to compute the flow inside a bent pipe using the SST model in CFX 5.5.1. I managed to compute it without problems for a coarse mesh and the first prism at 0.0001 m fron the surface, but when I look at the y+ values they are around 100 whereas they should be between 2 and 5 to get an accurate solution. So I decided to reduce the size of the first prism and right now it is 0.5e-05. The solution converges well, but when it is near the MAX residual 1e-04, it begins to oscillate. And then it stops saying that the solution has converged. I have looked at the mesh again and it seens ok. Is there something to do in order to get a cleaner convergence ?

Thank you very much.

Regards,

Olivier

Chris January 11, 2003 14:11

Re: Convergence problems
 
Hello Olivier,

oscillating residuals can have different reasons:

-where are the maximal residuals?

-check your grid quality in CFX-post (are the worst residuals in the region where the worst cells are?)

-use a different timestep

If that is all does not give you a hint, the most probable thing is, that the flow shows some instationary effects. Change to transient simulation with >=3 inner loops and you will be able to reduce the residuals further than with stationary effects.

Greeetings,

Chris

Olivier Macchion January 21, 2003 10:17

Re: Convergence problems
 
Thank you very much Chris. Just another question : how can you know where the maximal residuals are ?

Thank you very much.

Olivier

Chris January 21, 2003 14:12

Re: Convergence problems
 
Hello Olivier,

there are two possibilities:

- The location of the maximum values of the residuals is always included in the ".out"-file that is also displayed in the solver-monitor-window. You can view the location by creatin a point at the specified coordinates with CFX-post.

- Set the expert variable "output eq residuals" to "T". Then you will be able to view the residuals in CFX-Post just like any other variable.

Greetings,

Chris

Olivier Macchion January 22, 2003 01:57

Re: Convergence problems
 
Thank you Chris !

Greetings

Olivier


All times are GMT -4. The time now is 22:31.