parallel AVWhitney: DefaultDirichletBCs IterSolve Failed
Re: parallel AVWhitney: DefaultDirichletBCs IterSolve Failed
Sure! Here it is.
- Attachments
-
- ElmerBug_IterSolve.tar.gz
- (460 KiB) Downloaded 433 times
-
- Site Admin
- Posts: 4832
- Joined: 22 Aug 2009, 11:57
- Antispam: Yes
- Location: Espoo, Finland
- Contact:
Re: parallel AVWhitney: DefaultDirichletBCs IterSolve Failed
Hi tetraeder
Unfortunately your problem is a different one than that of Eelis. It was difficult to locate. Finally I traced it to line ~320 of DiffuseConvectiveAnisotropic.src . There the effective specific heat is computed from the enthalpy using c_p=dE/dT (this is the spatial 2 method). Unfortunately if dT==0 this fails. You had some of these funny elements quite randomly. Now this should not matter much since if you have all nodes at same temperature you cannot have heat fluxes either and hence no phase change.
Fix is in the svn (there is some problem with the migration to the new svn in sf.net. We are still using the old one - it works but there is just no www-interface to it).
Still don't know what the problem in Eelis' case is. Eelis, maybe you could use dropbox to share the case with us if it is rather big.
-Peter
Unfortunately your problem is a different one than that of Eelis. It was difficult to locate. Finally I traced it to line ~320 of DiffuseConvectiveAnisotropic.src . There the effective specific heat is computed from the enthalpy using c_p=dE/dT (this is the spatial 2 method). Unfortunately if dT==0 this fails. You had some of these funny elements quite randomly. Now this should not matter much since if you have all nodes at same temperature you cannot have heat fluxes either and hence no phase change.
Fix is in the svn (there is some problem with the migration to the new svn in sf.net. We are still using the old one - it works but there is just no www-interface to it).
Still don't know what the problem in Eelis' case is. Eelis, maybe you could use dropbox to share the case with us if it is rather big.
-Peter
Re: parallel AVWhitney: DefaultDirichletBCs IterSolve Failed
Hi Peter,
thanks for the good observation. I think your idea to tighten the convergence tolerance could work. Now I'm testing it with 1e-8 tolerance. Let's hope for the best. Will take few hours to calculate. I will report. I will try to share the case for tracking the bug number 1.
Cheers,
Eelis
thanks for the good observation. I think your idea to tighten the convergence tolerance could work. Now I'm testing it with 1e-8 tolerance. Let's hope for the best. Will take few hours to calculate. I will report. I will try to share the case for tracking the bug number 1.
Cheers,
Eelis
Re: parallel AVWhitney: DefaultDirichletBCs IterSolve Failed
Hi Peter,
it seems to work now. Thanks for the help!
-Eelis
it seems to work now. Thanks for the help!
-Eelis
Re: parallel AVWhitney: DefaultDirichletBCs IterSolve Failed
So I tried this with tighter linear convergence and turns out it is not bug either. Works well.Takala wrote: Bug 2) As mentioned I cannot run different time steps.
I try to report the bug 1 more in detail when I have time. I think that really is a bug.
Cheers,
Eelis