code_saturne User's Forum

David,
 
Still I could not fix the problem. Please suggest me something. The case file and listing file are attached here.
 
Thanks
Mita

It seems you're trying to use the steady algorithm. We sometimes have some difficulty to make it converge with buyoancy-driven flows (it's the case here if I'm right...).
Could you try the "standard" algorithm, i.e. unsteady algorithm with a constant and uniform time-step ? Make sure to have a low-enough time-step to ensure a good convergence (i.e. CFL number below 1 or 5 at least). And you should set the listing frequency output to 1 so that we have more information on the simulation.

Hi David,
 
Per your suggestion, I tried to use standard algorithm, i.e. unsteady algorithm with a constant and uniform time step (1e-05) but still solution is diverging.  Initially Courant number was small but as time goes, Courant number goes high. Here how did you calculate CFL condition/Courant number? Is it calculated as follows: CFL = (dt * Umax)/dx where dt = time step, Umax = maximum velocity and dx = element size. I tried with some more smaller time steps but it did not work yet. If you can suggest me some more trick to tackle this problem, I would appreciate your help. Is there anything else that can help me?
 
Thanks
Mita

Hi Mita,

The Courant number (and others) is computed in the dttvar subroutine. This is basically what you wrote, except that it is a bit more complex so as to take into account all kind of cells.

As for your convergence issue, I've seen Alexandre anwered to your other thread. I suggest we continue in the new one, if you don't mind.

David

Hello David,
 
Thank you for your reply. It is okay, we will continue the convergence issue in the new thread.
 
Regards
Mita