[makaveli_lcf]

Hi Foamers!

I got and issue using OpenFOAM v2.x:

Courant Number jumps 4 times if compare to OpenFOAM v1.6 and 1.7.x.
I tried cases for different geometries ofcouse with the same setting between OF versions and result is always the same!
Actually this issue results in a calculation time increasing 4 times. Unappropriated!

Please, comment on that issue, what is your experience with v2.x usage?

Cheer,
Alex

[makaveli_lcf]

Reported at Mantis:

http://www.openfoam.com/mantisbt/view.php?id=424

[makaveli_lcf]

Issue closed (http://www.openfoam.com/mantisbt/view.php?id=424#c1039):

The method for calculating the Courant number has been changed since version 1.7.x to be better represent the stability requirement of collocated meshes. In general the change hase resulted in slightly lower values.

A change in the method to calculate the Courant number does not in itself change the stability of the algorithm or lead to divergence as it is not used in the algorithm; only to set the time-step. If you have a case which behaved differently for the same fixed time-step between different versions of OpenFOAM the issue does not relate to a change in the way in which the Courant number is calculated.

So then it means that OF 2.x is less stable with the same fixed time step...
Will do more checks and report regarding actual calculation time.

[alberto]

Originally Posted by makaveli_lcf

Issue closed (http://www.openfoam.com/mantisbt/view.php?id=424#c1039):



So then it means that OF 2.x is less stable with the same fixed time step...
Will do more checks and report regarding actual calculation time.

What solver? I was aware of the change in the definition of Co, which was made to remove inconsistency at boundaries, but I haven't experienced major issues with 2.1.x.

[makaveli_lcf]

Hi Alberto!

pimpleFoam. According to my test to keep the same Co number (with which solution does not diverge) actual calculation time increased 2 times (I remind with the same constant dt I've got Co_new_OF = 4*Co_old_OF).

[makaveli_lcf]

Slow down, to my opinion, is not 4 times because now turbulence is solved at last PIMPLE iteration. But of course I tried with turbulence updated each iteration: the same result, solution diverges with the dt I used before with old OF versions..

[alberto]

Originally Posted by makaveli_lcf

Slow down, to my opinion, is not 4 times because now turbulence is solved at last PIMPLE iteration. But of course I tried with turbulence updated each iteration: the same result, solution diverges with the dt I used before with old OF versions..

OK, let's try to do some troubleshooting, if you like

Do you use any particular boundary condition? For example mapped, or time-varying, or custom BC? Some of these were updated in 2.x.

[makaveli_lcf]

Nope, standard k-epsilon (also tried with realizable which I mostly use) with standard wall functions for k, epsilon, nut, fixed values for U at inlet and zero gradient at outlet, pressure 0 at outlet and zero gradient at the inlet. I think it would be better if I upload here some test case. Will do it on Monday.

[alberto]

Yes, that's a good idea. The BC's are surely not the reason.