CFD Online Discussion Forums

CFD Online Discussion Forums (
-   OpenFOAM Running, Solving & CFD (
-   -   how can I make sure rhoSimplecFoam NEVER crashes? (

mihaipruna September 21, 2012 17:10

how can I make sure rhoSimplecFoam NEVER crashes?
Basically I want it to automatically cut off big numbers, replace divisions by 0 with something else, etc.
I tried relaxation factors and cellLimited gradients.
Is there a way to ensure you get a solution (even inaccurate) , always?

yhaomin2007 December 25, 2012 23:55

Hi, Mihai

Have you solved your problem, I think I have met the same problem like you.
I am running a simulation with a large area change, which makes the simulation very easy to crash.

Thank you

mihaipruna December 27, 2012 10:12

I had to streamline the geometry further. I was not able to get it to converge on my original model. WIth that, and residuals set to 10E-3, I was able to get it to converge. I had to set residuals this high because divergence occurred after a certain number of iterations. The residuals dropped to a certain level and then they started increasing (p mainly), and then the density would eventually reach either very large or small values, then it would crash.

So, in essence, I have not found a simple solution or a setting that would always ensure convergence just yet.

yhaomin2007 December 27, 2012 10:31

Is your flow a pressure driven flow or velocity driven flow? I think you can try time varying condition to get approach the condition you want. And I also suggest you try to decrease the relaxation factor to ensure stable.

mihaipruna December 27, 2012 12:10

I decided to go with pressure driven flow (total pressure).
In order to prevent my simulation from crashing right away , I had , for the inlet:

type totalPressure;
gamma 1.400000;
rho rho;
p0 uniform 29400.000000;
value uniform 26500.000000;

the 26500 was the static pressure and that was set to fixed at the outlet. I was not able to get the flow to a Mach number above 0.4 without it eventually crashing. Relaxation factors helped a bit.

So, again, it was a combination of factors and tweaks of residuals, relaxation factors, geometry and boundary conditions that yielded a converged solution.

All times are GMT -4. The time now is 11:40.