Relative permeability tensor in WhitneyAVSolver

[szewro]

Dear all,

I have problem with relative permeability tensor in WhitneyAVSolver.

When I define anisotropic material as:

Code: Select all

Material 2
    Name = "ferro"
    Electric Conductivity = 1.0e7
    Relative Permeability(3,3) = Real \
                50000.1 0 0 \
                0 20.0 0 \
                0 0 20.0
End

it causes NaN error in the solver.

Form

Code: Select all

Material 2
    Name = "ferro"
    Electric Conductivity = 1.0e7
    Relative Permeability = Real \
                50000.1 0 0 \
                0 20.0 0 \
                0 0 20.0
End

causes it solve, however, there is no anisotropy in the model (it seems that relative permeability is just equal 50000)

Do you know, if it possible to model anisotropic relative permeability wit ELMER FEM?

In the attachment please find simple example with anisotropic plate in uniform magnetic field.

With best regards,

Roman Szewczyk

[mzenker]

Hmmm, looking into the models manual it seems that only anisotropic conductivity can be used.

To be sure, one would have to look into the code...

Matthias

[szewro]

Hi!

Thank you for your answer!

I am not enough experienced in Fortran. It is difficult for me. Do you have an idea, where I should look for it?

With best regards,

Roman

[szewro]

There is also example with anisotropic reluctance: mgdyn_anisotropic_rel
It is promising, but I can't build it.

[mzenker]

Hi,

that one is simple.
Just type

elmerf90 reluctivity.F90 -o reluctivity.dll

on the command line (I assume you are under Windows?).
Then type

elmersolver

and the case will run.

HTH,
Matthias

[szewro]

Hi Matthias!

thank you for your help!

I work under Linux Mint. I tried to build it, and it works!
Previously, I have got an error, probably I used wrong command.

This time I tried:
elmerf90 reluctivity.F90 -o reluctivity.so

There is declaration in the code, which I don't understand:

Code: Select all

  REAL(KIND=dp) :: X(*)
  REAL(KIND=dp), POINTER CONTIG :: Y(:,:)

It is also SUBROUTINE reluct(Model, n, X, Y), not a function. However, it is not a problem.
I will do exactly the same and we will see the result. I will keep you informed.

Thank you very much again for your kind help!

With best regards,

Roman

[szewro]

Hi!

unfortunately it is not working.

I prepared file permefile.F90:

Code: Select all

SUBROUTINE permf(Model, n, X, Y)
!-------------------------------------------------------------------------------
  USE DefUtils
  IMPLICIT NONE
  TYPE(Model_t) :: Model
  INTEGER :: n
  REAL(KIND=dp) :: X(*)
  REAL(KIND=dp), POINTER CONTIG :: Y(:,:)
!-------------------------------------------------------------------------------
  
  Y = 0._dp
  Y(1,1) = 50000.1_dp
  Y(2,2) = 1000.1_dp
  Y(3,3) = 20.1_dp
!-------------------------------------------------------------------------------
END SUBROUTINE permf
!-------------------------------------------------------------------------------

I built it with:
elmerf90 permefile.F90 -o permefile.so

and I defined material in .sif as:

Code: Select all

Material 2
	Name = "ferro"
	Electric Conductivity = 1.0e7 ! 104.058
  	Relative permeability(3,3) = variable coordinate
    		real procedure "permefile" "permf"
End

Unfortunately, I again got NaN error from the solver:

Code: Select all

OptimizeBandwidth: ---------------------------------------------------------
OptimizeBandwidth: Computing matrix structure for: mgdynamics...done.
OptimizeBandwidth: Half bandwidth without optimization: 20687
OptimizeBandwidth:
OptimizeBandwidth: Bandwidth Optimization ...done.
OptimizeBandwidth: Half bandwidth after optimization: 2878
OptimizeBandwidth: ---------------------------------------------------------
DefUtils::DefaultDirichletBCs: Setting Dirichlet boundary conditions
DefUtils::DefaultDirichletBCs: Dirichlet boundary conditions set
SolveSystem: Solution trivially zero!
MGDynAssembly: Elapsed REAL time:     2.1585 (s)
DefUtils::DefaultDirichletBCs: Setting Dirichlet boundary conditions
DefUtils::DefaultDirichletBCs: Dirichlet boundary conditions set
WhitneyAVSolver::  Boundary tree edges: 94 of total: 279
       1        NaN
ERROR:: IterSolve: Numerical Error: System diverged over maximum tolerance.
Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG IEEE_DIVIDE_BY_ZERO

I still have no idea how to overcome this problem...

With best regards,

Roman

[mzenker]

Hi,

I have no experience with the WhitneyAV solver, and not enough time to look deeper into the case right now, sorry.
Someone else with more knowledge has to take over here...

Matthias

[szewro]

Hi Matthias!

I have made some experiments during the night.
I used working example: mgdyn_anisotropic_rel , I substituted .F90 file by permfile.F90:

Code: Select all

!-------------------------------------------------------------------------------
SUBROUTINE permf(Model, n, X, Y)
!-------------------------------------------------------------------------------
  USE DefUtils
  IMPLICIT NONE
  TYPE(Model_t) :: Model
  INTEGER :: n
  REAL(KIND=dp) :: X(*)
  REAL(KIND=dp), POINTER CONTIG :: Y(:,:)
!-------------------------------------------------------------------------------
  
  Y = 0._dp
  Y(1,1) = 50000.1_dp
  Y(2,2) = 2.1_dp
  Y(3,3) = 2.1_dp
!-------------------------------------------------------------------------------
END SUBROUTINE permf
!-------------------------------------------------------------------------------

and I made simple magnet-anisotropic plate test case in NETGEN:

Code: Select all

#
## aniso test
#
#  Netgen 6.2 required 
#

algebraic3d

solid core = cylinder  (0, 0, -1.5; 0, 0, 1.5; 3)
	     and plane (0, 0, -1; 0, 0, -1)
	     and plane (0, 0, 0; 0, 0, 1) -maxh = 0.3;

solid magnet = cylinder  (0, 0, 0.1; 0, 0, 3.9; 1)
	       and plane (0, 0, 0.2; 0, 0, -1)
	       and plane (0, 0, 3.5; 0, 0, 1) -maxh = 0.3;

solid range = sphere (0, 0, 0; 25)
              and not magnet
	      and not core;

tlo core  -col=[0,0,1];
tlo magnet -col=[0,1,0];
tlo range -col=[1,0,0] -transparent;

aniso_test.sif is in the attachment.

Please see the results:

aniso_test.jpg

(176.75 KiB) Not downloaded yet

It works, solves, but unfortunately, there is NO ANISOTROPY.

I agree, we need help. Fortran sources are to sophisticated for me. I give up for this moment.

Thank you for your help and discussion!

With best regards,

Roman

[mzenker]

Hi,

you should read the section 18.3 "writing a user function" in the Elmer Solver Manual. I think the problem is that you have declared a subroutine without return value(s).

HTH,
Matthias