Page 1 of 1

RotatingBCMagnetoDynamicsGeneric test failed

Posted: 10 Jan 2017, 01:03
by lazyTom
Hi all,

after having downloaded from GitHub, I've tried to run the test RotatingBCMagnetoDynamicsGeneric found in

https://github.com/ElmerCSC/elmerfem/tr ... icsGeneric

and had a fail.

I'm using Elmer on VM on a Mac. I've run all the commands required to build the mesh before starting the test. Below is the list of what I typed

Code: Select all

ElmerGrid 14 2 mortar_quarter.msh -autoclean
ElmerGrid extrude_quarter.eg
ElmerSolver
The errors returned are

Code: Select all

ERROR:: IterSolve: Failed convergence tolerances.
and

Code: Select all

WARNING:: CompareToReferenceSolution: Solver 2 FAILED:  Norm = 4.12269261E-03  RefNorm = 4.12179248E-03
CompareToReferenceSolution: Relative Error to reference norm: 2.183837E-04
I've increased Linear System Max Iterations to 700 but it didn't help. I thought I would give it a go in the forum before I continue with random tweaking :D
The complete output of ElmerSolver is attached below
ElmerSolver_Output.zip
(5.8 KiB) Downloaded 301 times
About, the output, the case_mod.sif file that you will see is simply a copy of case.sif (a precaution to avoid altering the original)

If anyone can give me hand to understand why it's not working it would be much appreciated

Regards,

Marco

Re: RotatingBCMagnetoDynamicsGeneric test failed

Posted: 10 Jan 2017, 13:55
by raback
Hi

The discrepancy to the reference solution is not too large (2e-4) considering that we are looking at the curl of the vector field. Unfortunately these problems do not converge that well and there may be some variations between different platforms and compilers. Of course ideally the tests should be more robust but unfortunately this is not always the case. So I don't think there is anything wrong with your setup.

-Peter

Re: RotatingBCMagnetoDynamicsGeneric test failed

Posted: 11 Jan 2017, 09:50
by lazyTom
Hi Peter,

thanks for clarifying. I have modified the last line of the .sif file ("Solver 2 :: Reference Norm = 4.12268808E-03") and I got a PASS :D.

I have some further questions, once you have extruded the 2D mesh (quarter_mesh.msh), and you are ready to prepare boundary conditions in the .sif file, how do you know which boundary/body number belong to a specific surface/body on the mesh?
Also, in the introductory comment of the .sif file, you say that there is a skew option for the rotor. How do you implement it? Looking at the manual RigidMeshMapper solver is not able to deform the mesh.
By the way, RigidMeshMapper works natively with degrees or radians?

Regards,

Marco

Re: RotatingBCMagnetoDynamicsGeneric test failed

Posted: 16 Jan 2017, 02:02
by lazyTom
Hi all,

Just a quick update on the skew feature, it took me some time staring at the MATC equation but eventually the penny dropped (tx(0) and tx(1) are always devious to interprete XD). By the way, it works in degrees. It's a very nice feature!

Still working on the boundaries identification question...

Cheers,

Marco