Problem with border conditions within WhitneyAVSolver

[szewro]

Hi Peter!

I am trying to model the behaviour of our non-magnetic chamber in Earth magnetic field. I started from the simplest system - permalloy ball in uniform magnetic field.

Uniform magnetic field is given by the boundary conditions of air cylinder:

Boundary Condition 1
Target Boundaries (3) = 1 2 3
Magnetic Flux Density 1 = Real 0
Magnetic Flux Density 2 = Real 0.1
Magnetic Flux Density 3 = Real 0
End


I generated the mesh with Netgen 6.2.
However, when I try to solve the system I have got segmentation error. Please help me to overcome this problem - the modelling of non-magnetic chamber is very important for our bioengineering research. Is this error caused by Netgen mesh or boundary conditions?

Her please find the link to our code:
simple ball example

Thank you in advance for your kind help!

With best regards!

Roman Szewczyk


ps. And there is the error description.
I am using Linux Mint with the newest Elmer from ppa server.

WhitneyAVSolver:: Boundary tree edges: 366 of total: 1095

Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

Backtrace for this error:
#0 0x7fced66f22da in ???
#1 0x7fced66f1503 in ???
#2 0x7fced6324f1f in ???
#3 0x7fceb1cd262b in dirichletafromb
at /build/elmerfem-csc-eg-nMxCWP/elmerfem-csc-eg-8.4-0ppa0/fem/src/modules/MagnetoDynamics/WhitneyAVSolver.F90:2490
#4 0x7fceb1cd4db7 in dosolve
at /build/elmerfem-csc-eg-nMxCWP/elmerfem-csc-eg-8.4-0ppa0/fem/src/modules/MagnetoDynamics/WhitneyAVSolver.F90:788
#5 0x7fceb1cda1d2 in whitneyavsolver_
at /build/elmerfem-csc-eg-nMxCWP/elmerfem-csc-eg-8.4-0ppa0/fem/src/modules/MagnetoDynamics/WhitneyAVSolver.F90:458
#6 0x7fced6cc69c3 in __mainutils_MOD_singlesolver
at /build/elmerfem-csc-eg-nMxCWP/elmerfem-csc-eg-8.4-0ppa0/fem/src/MainUtils.F90:5129
#7 0x7fced6cd9fc8 in __mainutils_MOD_solveractivate
at /build/elmerfem-csc-eg-nMxCWP/elmerfem-csc-eg-8.4-0ppa0/fem/src/MainUtils.F90:5365
#8 0x7fced6cdc7cf in __mainutils_MOD_solveequations
at /build/elmerfem-csc-eg-nMxCWP/elmerfem-csc-eg-8.4-0ppa0/fem/src/MainUtils.F90:2933
#9 0x7fced6ea4afc in execsimulation
at /build/elmerfem-csc-eg-nMxCWP/elmerfem-csc-eg-8.4-0ppa0/fem/src/ElmerSolver.F90:2339
#10 0x7fced6eaa63f in elmersolver_
at /build/elmerfem-csc-eg-nMxCWP/elmerfem-csc-eg-8.4-0ppa0/fem/src/ElmerSolver.F90:601
#11 0x55ea08f9817b in solver
at /build/elmerfem-csc-eg-nMxCWP/elmerfem-csc-eg-8.4-0ppa0/fem/src/Solver.F90:69
#12 0x55ea08f97ece in main
at /build/elmerfem-csc-eg-nMxCWP/elmerfem-csc-eg-8.4-0ppa0/fem/src/Solver.F90:34
Segmentation fault (core dumped)

[raback]

Hi Roman,

Yes, there seems to be a problem. It is related to elimination of edge dofs in conjunction with Dirichlet conditions. It only happens if there are more then one iteration. Hence, for your test problem you could reduce the number of nonlinear iterations to one.

I have a tentative fix, but will test & review it before committing.

EDIT: Just committed the fix to devel.

-Peter

[szewro]

Dear Peter!

Thank you very much for your kind help. It works!

With best regards,

Roman