CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   convergence RMS (http://www.cfd-online.com/Forums/cfx/23123-convergence-rms.html)

anna September 29, 2006 10:08

convergence RMS
 
Hello,

I would like to ask you if you can advise me something. My RMS is set to 1.0E-4. The solver is running without problem and finishing successfully the 100 iterations defined at the beginning. But the RMS of 1E-4 is not reached. There is a little pick at the very beginning by 1.E-2, goes down to 1E-3 and keeps almost constant 1e-3 till the end of the iterations. Do you have any ideas?

Joe September 29, 2006 10:35

Re: convergence RMS
 
Read "Judging convergence" in the manual.

TobiasZ October 2, 2006 03:02

Re: convergence RMS
 
Hi Anna, Reduce your physical time step by for example 5 and try again and see if the RMS gets lower. Depending on your simulation model the total number of required iterations might need more iterations. If for example the fluid flow is very slow iterations might need many iterations for convergence to obtain the required convergence criterion. Another reason for the high RMS could also be the quality and size of your mesh. Regards. Tobias

TB October 3, 2006 03:40

Re: convergence RMS
 
What turbulence model are you using? This could also be due to the inherent unsteady phenomenon in your flow. Run the transient simulation and check it out.

anna October 3, 2006 03:46

Re: convergence RMS
 
Hi, I am using SST model.

TB October 3, 2006 04:26

Re: convergence RMS
 
Have you got the same problem using standard ke model?

anna October 3, 2006 04:35

Re: convergence RMS
 
Yes, just at the beginning there is a small jump in the residuals plot and even after 500 iterations the residuals stay almost stable. But lateron by comparing the results in the output file, Global imbalance, MAX, RMS residuals everything seem to be withing normal limits. My mesh is not perfect, because I made it with ANSYS and I imported it in CFX and because of the boundary layer region I made the volumes next to it with hex mesh. So, to be more clear I have a multi-layers coil placed in a tube with running water. Next to the coil surface the Yplus for example is about 7 and next to the tube surface the Yplus is approx. 18 but eventhough I think the SSt model is the most appropriate one.

What is your opinion? Thanks a lot. Anna

TB October 3, 2006 04:50

Re: convergence RMS
 
SST model is very sensitive to the mesh quality, esp if the flow inside the domain is complex. I suggest you to include the following things in your checklist:

1.mesh (I'll try hexa if tet doesn't work properly) 2.unsteady flow effect (I found that SST model sometimes captures the unsteady phenomenon that ke model fails to show in the solution, this will cause convergence problem)

You need to solve the convergence problem before you can say anything about the accuracy.


All times are GMT -4. The time now is 08:10.