[Getdp] Various questions about superconductors, cohomology & procedural constraints in GetDP

[Christophe Geuzaine]

Hi Guillaume,

I didn't have time to have a detailed look at your files, but the NaN you get when running them as-is is due to an erroneous value of "NbTurns" (it should be 1 instead of 2). Changing this the model converges nicely. Note that there also seem to be mistakes in your use of ScalingFactor, as changing it leads to an erroneous geometry.

The conducting matrix in helix.pro makes the problem a bit "easier" to solve - you add more dissipation. But removing it should also clearly work: in my initial tests I had just to impose stronger tolerances on the nonlinear residual, and sometimes reduce the time step.

Without the conducting matrix, what is the "physical" way to impose the current? In each "filament"?

Christophe

> On 07 Dec 2015, at 15:51, DILASSER Guillaume <Guillaume.DILASSER at cea.fr> wrote:
> 
> Hello everyone,
> 
> I am writing to (hopefully) have some response to the questions I have been asking myself about getDP and the simulations of superconducting materials. 
> 
> A) In every simulation of superconducting behavior I make, I am more or less obliged to add some resistive (ohmic) elements in order to get results. Otherwise, the computation wil sooner or later end with an external PETSc error "singular matrix". The model I am currently experimenting on is derived from Christophe's helix.pro (the non linear material law is the same), in which there is also always a resistive matrix that you can't remove. Could anyone with numerical know-how explain to me why this error is happening ? Would there be another way to consider only the superconductor ?
> 
> B) In the enclosed project file I tried to automate the implementation of current constraints yet I am not sure I made it right. The system I want to work on is a multi-layered superconducting coil made of REBCO tape, since I want to be able to isolate what is happening in each separate turn, I need to proceed to a lot of cuts (one between each layer) and generate the corresponding cohomology constraints. Could anyone tell me if what I implemented is ok(-ish) ?
> 
> C) Obviously, my current implementation is not working. With more than one turn I get NaN non-linear residue, but even with just one turn I get wrong results (even is I made the same computation in another project file earlier and the results were fine). It seem that (see enclosed picture) the cohomology cut hitting the superconducting surface is the cause. One can see that I have a weird (and surely wrong) behavior at the intersection. It is almost as if (despite the current constraint I try to enforce) the total current flowing through the conductor was set to zero... Can anyone help me to understand what is happening there ?
> 
> I would sincerely appreciate your help on those questions. Faithfully yours,
> 
> Guillaume Dilasser
> <Bobinette_supra_2.geo><Bobinette_supra_2.par><Bobinette_supra_2.pro><GDilasserMesh&Cut.jpg><GDilasserStrangeBfield.jpg><GDilasserJdensity.jpg>_______________________________________________
> getdp mailing list
> getdp at geuz.org
> http://www.geuz.org/mailman/listinfo/getdp

-- 
Prof. Christophe Geuzaine
University of Liege, Electrical Engineering and Computer Science 
http://www.montefiore.ulg.ac.be/~geuzaine

Tetrahedron V, July 4-5 2016: http://tetrahedron.montefiore.ulg.ac.be
Free software: http://gmsh.info | http://getdp.info | http://onelab.info