WhitneyAVSolver crash

Numerical methods and mathematical models of Elmer
Post Reply
danielgboyle
Posts: 6
Joined: 24 Oct 2011, 20:26
Antispam: Yes

WhitneyAVSolver crash

Post by danielgboyle »

I managed to create a magnetic field around a wire in elmer by using test case mgdyn_bh as an example. (just changed the coil to be air) After analyzing the same model in Ansys the field intensity is within 2% at a radial distance of .004 m from the center. Also there is good agreement with formulas. So I ran 1 magnetostatic model and the results are fairly accurate.

The problem is that I then modified this sif file so I could use it on a simple coil model which I brought in from Ansys using the Elmer supplied macro. I changed the boundary conditions, and made sure the 3 dimensional current body forces were transferred correctly. The problem is that the model crashes in mgdynamics. Here are the last few lines from the output to the screen before it crashes in mgdynamics.

DefUtils::DefaultDirichletBCs: Setting Dirichlet boundary conditions
DefUtils::DefaultDirichletBCs: Dirichlet boundary conditions set
CRS_IncompleteLU: ILU(0) (Real), Starting Factorization:
CRS_IncompleteLU: ILU(0) (Real), NOF nonzeros: 2754517
CRS_IncompleteLU: ILU(0) (Real), filling (%) : 100
CRS_IncompleteLU: ILU(0) (Real), Factorization ready at (s): 0.13
74 0.8983E-10
ComputeChange: NS (ITER=1) (NRM,RELC): ( 13.329001 2.0000000 ) :: mgdynamics

It should work but I just can't figure it out. It may be the mesh which I see has double numbers when I look at the surface numbers. If anyone could help I would appreciate it. :)
Also when it is debugged I am going to upload it with explanations so future new users of magnetostatics will have an initial model to run and compare against formulas.
Attachments
capcoil.zip
elmer mesh files, sif file and output of failed run in a zip file
(1.4 MiB) Downloaded 547 times
Juha
Site Admin
Posts: 357
Joined: 21 Aug 2009, 15:11

Re: WhitneyAVSolver crash

Post by Juha »

Hi,
the edge basis functions for second order elements have not been implemented
within ElmerSolver, sorry.There is even (a relatively new) error check implemented
in SVN head to capture this case.
Regards, Juha
danielgboyle
Posts: 6
Joined: 24 Oct 2011, 20:26
Antispam: Yes

Re: WhitneyAVSolver crash

Post by danielgboyle »

Hi Juha,

I had a lot of problems with those files above and beyond having second order elements. However I rewrote the elmer ansys macro for Ansys Workbench SImulation (which does not have areas available) and thus now I can import arbitrary geometry using 504 Whitney AV elements. Also I already know the bc's from my naming convention, and I just edit the sif file.

Thank you for replying so quickly. I am very impressed with Elmer. It is a great development that you are involved with. :D
madstamm
Posts: 37
Joined: 07 Apr 2010, 09:56

Re: WhitneyAVSolver crash

Post by madstamm »

Hi daniel,

Nice work. Good intro model too, this coil! I was working with the whitney solvers beg of this year and am just getting back into it. Hope to get my new linux box soon and post some results.

Elmer is very promising for EM simulation, we just have to use it more - and make some tutorials :)

I think one really important tutorial that doesnt exist yet is one showing an example of an entire open source flow, to model, mesh, simulate and then calculate the inductance (henries) of a iron core coil. If we could demonstrate that, it would be a huge step forward:
- Modeler (ex: blender)
- correct definition of boundary surfaces
- correct use of Export scripts
- Meshing tool (ex: engrid, netgen)
- useful grid density optimization tips
- correct import to elmer
- show how to setup solvers to calc electric current distribution
- setup solver for resulting mag field calc
- show how to calc inductance
Mike
Post Reply