Issue: NormalVector: Invalid dimension for determining normal!

Numerical methods and mathematical models of Elmer
Post Reply
norbert
Posts: 4
Joined: 25 Sep 2015, 09:49
Antispam: Yes

Issue: NormalVector: Invalid dimension for determining normal!

Post by norbert » 23 Jul 2018, 16:44

Hi,

I am trying to run a thin film (simple Couette problem) simulation in Elmer. In principle I have a plate with a fluid entering on a pressure side and the left and right faces have periodic boundary conditions. Basically this is an old case, now unfortunately the solver during preconditioning throws the errormessage "NormalVector: Invalid dimension for determining normal!". I have traced this to the function ElementDescription.F90 (lines 11345 ff) where a case structure leads to the default error because Boundary % TYPE % DIMENSION seems to be not 2 (or 0 or 1) as expected.

I have tried mesh generation through foamextend -> Elmer and Salome unv -> Elmer via ElmerGrid. Both lead to the same behaviour. Furthermore, I have tried older Elmer versions and installs and the behaviour is identical.

I have rebuild the .sif file but to no avail.

I have attached a minimal example which is showing the non working behaviour.

Would be grateful if anybody could help/ or comment on this behaviour as I do not know at which point I should continue looking for a solution.

Thanks and Regards,

Norbert
Attachments
MinimalNonWorkingExample.zip
Contains sif, mesh, startfile
(367.69 KiB) Downloaded 12 times

norbert
Posts: 4
Joined: 25 Sep 2015, 09:49
Antispam: Yes

Re: Issue: NormalVector: Invalid dimension for determining normal!

Post by norbert » 24 Jul 2018, 12:16

Hi,

in the morning I tested a very old version of ELmer (v6.2) which was packed in CAE Linux2011. This version runs the above test case without problems (requiring a change in the output file name from file to string). The Elmerversion I used peviously was 8.3.

I am still a little baffled by the behaviour and would appreciate tipps to understand how I can manage to use the current Elmer distribution.

Thanks and kind Regards,

Norbert

norbert
Posts: 4
Joined: 25 Sep 2015, 09:49
Antispam: Yes

Re: Issue: NormalVector: Invalid dimension for determining normal!

Post by norbert » 08 Oct 2018, 18:11

Hi all,

I still haven't figure out a reason for the above mentioned behaviour. Is there something that changed in the solver behaviour (treatment of normals?) during a version change? I have tried to find and study changelogs, however I cannot find a change that could cause the behaviour. Would like to set up a simulation with the newest Elmer version.

Thank you and Regards,
Norbert

raback
Site Admin
Posts: 3291
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Issue: NormalVector: Invalid dimension for determining normal!

Post by raback » 09 Oct 2018, 11:35

Hi

I had a quick look at the problem. Your problem is that you assigned "body id=1" to the boundary where you want the Reynolds equation is valid. This is problematic because the body 1 is already assigned for the true bulk body. Instead you should use an empty index i.e. "body id=2" and create a new group.

I also don't see how your cavitation is supposed to work. For the present case I took them out.

I added a better Fatal to the code in case it is used for 3D elements.

Attached is a case that produces something.

-Peter
Attachments
case_no_ela_Val_Teststruk.sif
working command file
(5.41 KiB) Downloaded 2 times

norbert
Posts: 4
Joined: 25 Sep 2015, 09:49
Antispam: Yes

Re: Issue: NormalVector: Invalid dimension for determining normal!

Post by norbert » 11 Oct 2018, 18:33

Hello Peter,

thank you very much. I really appreciate your time.

Seems like I will have to look at my sif file generation once more and review it :oops:

However, I still am baffled that the calculations run on older versions of Elmer.

Thanks,

Norbert

Post Reply