Elmer 8.2 / 8.3 - mgdynamics solver convergence problem

Numerical methods and mathematical models of Elmer
Post Reply
Xan
Posts: 8
Joined: 25 Aug 2015, 12:18
Antispam: Yes

Elmer 8.2 / 8.3 - mgdynamics solver convergence problem

Post by Xan »

Dear all,

I am working with the usage of mgdynamics solver to simulate magnetic field distribution and eddy currents in the element. Recently I updated Elmer from version 7.0 to 8.3, because of the improvement in calculations speed. Since then around half of my simulations (I conduct a lot of them) had failed to converge. I had tried different mesh densities and convergence parameters and nothing succeed. I also tried on ElmerFEM 8.2 for Windows with the same results.

The problem case (with .sif, mesh folder and some elmer functions) you can download here.

The case consists of two coils - exciting one powered by alternating current and receiving one, in which magnetic field is induced and results are given further analysis. Between the coils there is a block, in which I inscribe the physical parameters of some tested objects' model, but for this case I have set the parameters of the block to be constant in entire volume.

The interesting part of this problem is setting convergence parameters:
-when I set Linear System Convergence Tolerance to 1e-7 the convergence parameter for Dirichlet boundary conditions oscillates around 1e-5 for around 1300 steps and then rises to 1e40
-when I set when I set Linear System Convergence Tolerance to 1e-5 the convergence parameter for Dirichlet boundary conditions oscillates around 1e-3 for around 1300 steps and then rises to 1e40
-when I set Linear System Convergence Tolerance to 1e-3 the convergence parameter for Dirichlet boundary conditions oscillates around 1e-1 for around 1300 steps and then rises to 1e40

I will be extremely grateful for any advices and suggestions.
Paweł
Post Reply