CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Siemens (https://www.cfd-online.com/Forums/siemens/)
-   -   Convergence in transient case (https://www.cfd-online.com/Forums/siemens/55828-convergence-transient-case.html)

Hamodi January 31, 2007 07:17

Convergence in transient case
 
I wonder how to judje convergence when running a transient case. Because the monitoring output is global rate of change so there is no normalised residuals.

Thanks

/Hamodi

Ben January 31, 2007 18:09

Re: Convergence in transient case
 
If you are using PISO look at the corrector count it should be nice and low, also look out for warnings. If using simple transient use the residuals.

Hamodi January 31, 2007 19:09

Re: Convergence in transient case
 
T'm using, PISO the only one in v3.26, and I've 3 corrector steps. However the number of sweeps is about 33 in the last corrector step. Is there any way to decrease the number of corrector steps and to decrease the elapsed time at each time step. The maximum CFL-number is around 2.4 and the time step is 5e-5. Thanks for your help!

/Hamodi

Dave February 1, 2007 05:10

Re: Convergence in transient case
 
As a general rule of thumb, 3 or 4 pressure corrector stages is about right (when things go wrong you'll be hitting the 20 limiter). For the calculation to run stable, the mean Courant number should be around 1 (max around 10), so it sounds like you can safely increase your timestep size.

To get a better idea of how it's converging, if you go to "Output Controls > Monitor Numeric Behavior" and switch on "Print Inner Iteration Residuals" it will print the residual history for each pressure corrector in the .info file (look for EQ; PCOR, then a column of numbers). This will tell you if it's converging monotonically or not. If it's not converging well, try under relaxing the pressure correction (defaults to 1 in the Solution Method panel).

I didn't get what you want to do with decreasing the elapsed time at each time step (that's just the time step size?), or why you want to decrease the number of corrector stages. If you want to decrease your converge criteria per time step, then play with the "Reduction in Residual for Correct Stages" in the Solution Method panel.


All times are GMT -4. The time now is 11:00.