Different solutions among different computers
I am running a supersonic case with OF (sonicFoam) and i would like to share this annoying experience in order to read your advices.
I have 3 different computers with 3 different version of OF. One has Ubuntu with OF2.0.0; Other has Ubuntu with OF2.0.1 and the other one has Suse with OF2.1.0.
When I run exactly the same case in them, they reach the same solution for UBUNTU computers, but completely different with SUSE one; they have almost the same numbers at the very begining of the simulation, but so different a few iterations later that UBUNTU seems to get a solution and in SUSE the courant number rises so much that diverges.
I have attached the logs of the simulation of both of them if you want to take a look.
Shouldn't get all of them the same (although not exactly) solution? Any comment? Remember that i run exactly the same case: same BCs, same schemes, same initial conditions, etc.
Thanks in advance.
Test ubuntu with OF-2.1, or SuSE with OF-2.0.1
Thank you for your answer.
As you told me I have tested ubuntu with OF-2.1 and it reaches the same solution that suse OF-2.1. (thing that i should have tested before post the thread) Thank you for the observation.
But, it makes me think about this: if OpenFOAM release notes for the new version say nothing about changes in this solver, can anybody tell me why different versions reach different solutions for the same case? Because it is not very usual that the old version finds a solution while the new one diverges.
If it's exactly the same case, and the non-diverging solution makes sense, I would post a report along with the case on the official bug tracker.
Greetings to all!
Before reporting, I would advise to check with OpenFOAM 2.1.1 and 2.1.x as well.
Additionally, it would be helpful to have access to a simple example case that would replicate the problem, specially when you do the bug report!
Last but not least: although solvers themselves don't change, the boundary conditions and other internal algorithms keep evolving and the cases usually need to be adjusted to the new ways of defining the boundary conditions, fvSchemes, fvSolutions and so on. Therefore, it's advisable to compare the reference tutorials between versions, for those you've based your cases on.
As a good example of code evolution, check this bug report: http://www.openfoam.org/mantisbt/view.php?id=627
edit: I forgot to mention that numerical sensitivity can also change between versions!
|All times are GMT -4. The time now is 12:28.|