CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Phoenics (http://www.cfd-online.com/Forums/phoenics/)
-   -   convergence parameters (http://www.cfd-online.com/Forums/phoenics/51952-convergence-parameters.html)

amv September 14, 2005 05:35

convergence parameters
 
Hi to all, I was till recent time user of fluent and now turned to phoenics. in fluent setting under-relaxation parameters for all equations , in the range of 0 to 1 resulted in convergence of the respective equations. the under-relaxation factor was the most influencing parameter for convergence of an equation. In Phoenics, in the main menu --> Numerics panel, I find a lot of parameters in the relaxation-control and iteration control panel but i am not able to set them.

Can you suggest the possible values for simple flow situations like Lid driven cavity, channel flow etc. if i use auto convergence control, the % error doesnt reduce much for velocities though it reduces for pressure and temperature ! How to under-relax all the equations to ensure convergence. Thanks.

Mike September 14, 2005 10:56

Re: convergence parameters
 
You wrote: "I find a lot of parameters in the relaxation-control and iteration control panel but i am not able to set them. "

Specifically, what happens when you try to set them, and under what conditions do you try to set them? Do they revert back to the original default settings? Or do you mean the changes have no effect on the values of the relaxation parameters reported in the RESULT file. You should certainly be able to change the values when automatic convergence control is deactivated.

Note that positive values of the relaxation parameters signify inertial relaxation, ie false time-step relaxation. Negative values signify linear relaxation. For P1, the pressure-correction equation, only linear relaxation has any meaning.


amv September 19, 2005 03:04

Re: convergence parameters
 
hi, if i set -ve values (which, i think, means under-relaxation factor)for velocities instead of using +ve values suggest by default settings. i.e. i de-activate auto convergence and then try to change +1 to -0.25 for velocities. But the solution immediately diverges. i thought that under-relaxation is a better method than using +ve values (which is false time step I guess). Does it mean that False time step approach is more powerful than under-relaxation? In fluent i used to get converged results with under-relaxation technique nicely but this is not happening with pehionics.

please if somebody can throw any light on this issue.

Mike September 19, 2005 06:36

Re: convergence parameters
 
I would suggest that inertial under-relaxation has more physical basis than linear under-relaxation, because the relaxation time scale (false-time-step) can be selected by using the local physical time scale as basis.

For the lid-driven cavity flow you could try a false time step of (cavity width)/(lid velocity), or if that doesn't converge a more conservative value of (cavity width)/(lid velocity)/(number of cells in cavity width).

amv September 21, 2005 06:49

Re: convergence parameters
 
My only problem is how to set the set of parameters in the relaxation and iteration control panel !

ok i tried that technique mike. thanks for ur suggestion. this funda i found in manual also. the best result i am able to obtain is by using 1) deactivate the auto convg control 2) cal dtfals for velocities with the technique u told 3) dont change anything else in the relaxation and iteration control panel

BUT, if u have any better suggestion , please let me know.

I want the following things 1) % error for velocities should fall atleast 1e-2. mostly it becomes stable at 1e0 2) since i am user of fluent, instead of % error , i believe in residual. how to get the o/p of equation residual. i used to get equation residual of 1e-7 generally in fluent and it is possible in phoenics?

Thanks for ur patience and quick replies.


All times are GMT -4. The time now is 05:30.