<div dir="ltr"><div><div><div>Hi there,<br><br></div>I've just put together something similar in 2D. The solution doesn't seem wrong at a first glance, but the Newton takes one iteration only, which is suspicious. So be careful with this.<br><br></div>Best Regards<br><br></div>Peter<br><div><div><br></div></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Oct 28, 2015 at 3:11 PM, Frederic Trillaud Pighi <span dir="ltr"><<a href="mailto:ftrillaudp@pumas.iingen.unam.mx" target="_blank">ftrillaudp@pumas.iingen.unam.mx</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Dear Guillaume and Peter,<br>
<br>
The model is 3D because I add a recurring error in the 2D case that I<br>
could not fix to answer quick Guillaume's problem. As Christophe<br>
mentioned, it is not necessary to force the boundary condition in this<br>
case. It should not make much difference as far as I understood (to be<br>
checked). If you impose a Dirichlet condition to zero, I would say ou<br>
need the shell transformation since in reality the magnetic field should<br>
be of the form I/(2*pi*R) at the boundary.<br>
The code I sent turns but there is a problem of convergence to be<br>
solved. I tried the built-in solver and ended up copying Christophe's<br>
model.<br>
<br>
For Guillaume:<br>
<span class="">>why is it not necessary to have a cut through the surrounding air<br>
region?<br>
</span>If you check the model that I sent you, you will see that the air is cut<br>
using the cohomoloy solver (there are two regions: air and<br>
superconductor, no more). In the visibility option you will see<br>
something like "Line" with a name starting with H..., it will show you<br>
the cut (see attached *.png).<br>
<br>
Best,<br>
<br>
Frederic<br>
<div><div class="h5"><br>
<br>
<br>
<br>
On Wed, 2015-10-28 at 12:42 +0000, DILASSER Guillaume wrote:<br>
> Good Afternoon,<br>
><br>
><br>
><br>
> I had the same reaction as Peter regarding the fact that Frédéric's<br>
> model is in 3D and the apparent lack of boundary conditions. Yet I<br>
> also have another question in mind : why is it not necessary to have a<br>
> cut through the surrounding air region ? It is not simply connected if<br>
> I am not mistaken. Is it because the basis function used in the air<br>
> region are BF_GradNode and not BF_Node (I don't think so, otherwise I<br>
> don't understand this example) ? Or is it because the cohomology<br>
> solver does the same job ?<br>
><br>
><br>
><br>
> Anyway, I reworked a lot what I previously sent to the mailing list to<br>
> use the same H-phi formulation as Christophe and I enclose the latest<br>
> version of my files with this email. It doesn’t work right know and<br>
> worse it makes GetDP crash (“Windows has encountered a problem and<br>
> must close”). The errors takes place during the processing, while<br>
> executing Generate[System] at the first time step in a<br>
> TimeLoopTheta[]. It seems that the program can’t generate a<br>
> ‘_BF_Entity_13’ (GroupOfNodesOf) which is associated with the<br>
> condition linked to the transport current through the tape. Some<br>
> information on my hardware that may be relevant : I run GetDP on a<br>
> Dell Workstation with Intel Xeon E5-2650 v2, 128 Go RAM and a Window<br>
> 8.1 OS.<br>
><br>
><br>
><br>
> Finally, about the usual geometry for HTS ReBCO tapes :<br>
><br>
> · The cross section is rectangular<br>
><br>
> · Width is usually 4, 6 or 12 mm, it is possible to find 2 mm-<br>
> or 8 mm-width tape but they are very rare<br>
><br>
> · The thickness is around 100 µm and decomposes as follows :<br>
> first a substrate (Hastelloy / Stainless Steel) of about 60 µm. For a<br>
> fine study of the tape it would be useful to take the magnetization of<br>
> the substrate into account (the second benchmark on this page). On top<br>
> of the substrate are about 1 µm of buffer layers that are very<br>
> resistive and about 1 µm of ReBCO superconductor. Finally the whole<br>
> structure is covered with electroplated copper, which adds something<br>
> like 20 µm on the top, bottom and each sides.<br>
><br>
> Hoping this may be useful,<br>
><br>
><br>
><br>
> Sincerely Yours,<br>
><br>
><br>
><br>
> Guillaume DILASSER<br>
><br>
> Doctorant SACM / LEAS<br>
><br>
> CEA - Centre de Saclay - Bât.123 - PC 319c<br>
><br>
> 91191 Gif sur Yvette Cedex - France -<br>
><br>
><br>
><br>
> <a href="mailto:guillaume.dilasser@cea.fr">guillaume.dilasser@cea.fr</a><br>
><br>
><br>
><br>
><br>
><br>
><br>
><br>
> -----Message d'origine-----<br>
> De : getdp [mailto:<a href="mailto:getdp-bounces@ace20.montefiore.ulg.ac.be">getdp-bounces@ace20.montefiore.ulg.ac.be</a>] De la<br>
> part de Christophe Geuzaine<br>
> Envoyé : mercredi 28 octobre 2015 10:42<br>
> À : Peter Kis <<a href="mailto:md2z34@gmail.com">md2z34@gmail.com</a>><br>
> Cc : Frederic Trillaud Pighi <<a href="mailto:ftrillaudp@pumas.iingen.unam.mx">ftrillaudp@pumas.iingen.unam.mx</a>>;<br>
> <a href="mailto:getdp@geuz.org">getdp@geuz.org</a><br>
> Objet : Re: [Getdp] Non-linear magnetodynamics with a superconducting<br>
> tape<br>
><br>
><br>
><br>
><br>
><br>
> > On 28 Oct 2015, at 08:59, Peter Kis <<a href="mailto:md2z34@gmail.com">md2z34@gmail.com</a>> wrote:<br>
><br>
> ><br>
><br>
> > Hi,<br>
><br>
> ><br>
><br>
> > I checked out Frederic's model and can't understand why it is in 3D.<br>
> Since it is a straight conductor, it should be modeled in 2D.<br>
><br>
> > In fact my question is related to this issue. Is this Form1<br>
> function<br>
><br>
> > space suitable for a 2D model? (I think Form1P isn't suitable<br>
> because<br>
><br>
> > H is in-plane vector, so it should be Form1.)<br>
><br>
><br>
><br>
> I've updated the example on the wiki to also perform 2D calculations:<br>
> give it a try and let me know. You will see that the formulation does<br>
> not need to be changed at all - only the geometry.<br>
><br>
><br>
><br>
> @Frederic & Peter: do you have "canonical" geometries for<br>
> superconducting tapes? We could either create a new parametric<br>
> geometry for those, or maybe just extend "helix.geo" to handle e.g.<br>
> rectangular (instead of circular) cross-sections.<br>
><br>
><br>
><br>
> > Moreover I don't see any boundary condition on the outermost<br>
> surfaces, where the magnetic field should be set to zero. It is also<br>
> missing form Christoph's Helix model too. Apparently it works, but<br>
> why?<br>
><br>
> ><br>
><br>
><br>
><br>
> The "helix" model imposes (weakly) a zero normal magnetic flux density<br>
> on the boundary of the air box; I think this is more appropriate than<br>
> imposing (strongly) that the magnetic field vanishes on the boundary.<br>
> To make it even better we could add a shell transformation to<br>
> infinity.<br>
><br>
><br>
><br>
> Cheers,<br>
><br>
><br>
><br>
> Christophe<br>
><br>
><br>
><br>
><br>
><br>
> > Sorry for asking a lot.<br>
><br>
> ><br>
><br>
> > Best Regards<br>
><br>
> ><br>
><br>
> > Peter<br>
><br>
> ><br>
><br>
> ><br>
><br>
> > On Sun, Oct 25, 2015 at 12:48 AM, Frederic Trillaud Pighi<br>
> <<a href="mailto:ftrillaudp@pumas.iingen.unam.mx">ftrillaudp@pumas.iingen.unam.mx</a>> wrote:<br>
><br>
> > Dear Guillaume,<br>
><br>
> ><br>
><br>
> > I have completed a run of the code I sent you (I should have<br>
> checked<br>
><br>
> > it before sending it!!). Even though it runs, It does not seem to<br>
> work<br>
><br>
> > correctly. Indeed, it does not seem to converge. I will look<br>
> through<br>
><br>
> > it in my spare time.<br>
><br>
> ><br>
><br>
> > Best,<br>
><br>
> ><br>
><br>
> > Frederic<br>
><br>
> ><br>
><br>
> ><br>
><br>
> > On Wed, 2015-10-21 at 14:14 +0000, DILASSER Guillaume wrote:<br>
><br>
> > > Good evening,<br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > > I am writing this Email to hopefully get some advice on how to<br>
><br>
> > > implement in GetDP a non-linear material law corresponding to the<br>
><br>
> > > behavior of a HTS superconductor. I have recently started to use<br>
><br>
> > > GetDP with the aim of simulating superconducting electromagnets<br>
> but<br>
><br>
> > > for now I am still learning how to use the software. I chose to<br>
> work<br>
><br>
> > > first on an example available here (HTS modelling workgroup<br>
> website,<br>
><br>
> > > it is example 1 at the top of the page) to be able to compare my<br>
><br>
> > > results to those of the community. The scenario implies a simple<br>
><br>
> > > strand of superconductor with a rectangular cross-section to<br>
> which<br>
><br>
> > > is applied an AC current, everything is then surrounded by air.<br>
> For<br>
><br>
> > > the benchmark, the aim is mostly to compute the AC losses in the<br>
><br>
> > > material during the current cycle.<br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > > Since I have much to learn, I tackled the problem step by step by<br>
><br>
> > > first considering a linear case in which I have resistive<br>
> material<br>
><br>
> > > in place of the superconductor. I enclose the files I have been<br>
><br>
> > > working on for those who want to have a look at them (sadly it is<br>
><br>
> > > mostly written in French…). The outline of what I tried is<br>
> following<br>
><br>
> > > : I considered the 2D case to resolve the problem in the<br>
><br>
> > > cross-section but did not implement yet the use of symmetries. I<br>
><br>
> > > wrote the A-Phi formulation with constant parameters nu and<br>
> sigma.<br>
><br>
> > > The resolution of the magnetodynamic problem is fairly trivial as<br>
> everything is linear.<br>
><br>
> > > I guess the solution shown in the enclosed files is about to be<br>
><br>
> > > correct yet I still have some unanswered questions :<br>
><br>
> > ><br>
><br>
> > > · Is the way I implemented the shell-to-infinite-domain<br>
><br>
> > > transformation correct ? I had some error like “the Jacobian x is<br>
><br>
> > > not in the range [a,b]” which I did not understand when messing<br>
> with<br>
><br>
> > > the VolSphShell Jacobian, is there a specific documentation on<br>
> the<br>
><br>
> > > topic ?<br>
><br>
> > ><br>
><br>
> > > · When I try to drop the thickness of the tape down to<br>
> about 1<br>
><br>
> > > µm (the width of the tape is about 4 mm), Gmsh gives an error as<br>
><br>
> > > “some points are coincident”. I tried the solutions to prevent<br>
> that,<br>
><br>
> > > namely decrease the size of the mesh elements in the conductor<br>
><br>
> > > and/or increase the mesh.RandomFactor but without success. What<br>
> did I miss ?<br>
><br>
> > ><br>
><br>
> > > · I tried to compute the AC losses in the material in the<br>
><br>
> > > post-process but unfortunately I must have done something wrong as<br>
> I<br>
><br>
> > > always find 0… Where does it come from ?<br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > > Those were the issues I still have with my first (resistive)<br>
><br>
> > > scenario but to switch now to the original example I need to add<br>
> the<br>
><br>
> > > E-J material law for the superconductor. My plan is to begin with<br>
> a<br>
><br>
> > > simple non-linear power law E = e0 * (|J|/jc)^N * (J/jc), e0 =<br>
> 10-4<br>
><br>
> > > V/m, jc =<br>
><br>
> > > 108 A/m², N about 10. For the A-Phi formulation I have to<br>
> implement<br>
><br>
> > > a<br>
><br>
> > > sigma(E) = sigma(dA/dt + gradPhi) relation in the equations but<br>
> what<br>
><br>
> > > is precisely to do remains unclear.<br>
><br>
> > ><br>
><br>
> > > · The <a href="http://helix.pro" rel="noreferrer" target="_blank">helix.pro</a> example gave me an idea of what I could do<br>
> even<br>
><br>
> > > if it is not the same formulation. However, if I am unsure on how<br>
> I<br>
><br>
> > > can linearize the expressions involving sigma. If I understood<br>
> well,<br>
><br>
> > > it is forbidden to write a term such as<br>
> [ DtDof[ sigma[Dof[{a}],{v}]<br>
><br>
> > > * Dof[{a}], {a} ].<br>
><br>
> > ><br>
><br>
> > > · Or, is this possible to use some of the Built-in<br>
> functions<br>
><br>
> > > like IterativeLoop[] with non-linear term like the one above ?<br>
><br>
> > ><br>
><br>
> > > I would really appreciate your advice on how I can (and should)<br>
><br>
> > > implement this material power-law behavior. I realize that A-Phi<br>
> was<br>
><br>
> > > probably not the best formulation to work with as the sigma<br>
> appears<br>
><br>
> > > a lot in the equation and is a function of two variables…<br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > > Faithfully yours,<br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > > Guillaume DILASSER<br>
><br>
> > ><br>
><br>
> > > Doctorant SACM / LEAS<br>
><br>
> > ><br>
><br>
> > > CEA - Centre de Saclay - Bât.123 - PC 319c<br>
><br>
> > ><br>
><br>
> > > 91191 Gif sur Yvette Cedex - France -<br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > > <a href="mailto:guillaume.dilasser@cea.fr">guillaume.dilasser@cea.fr</a><br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > ><br>
><br>
> > > _______________________________________________<br>
><br>
> > > getdp mailing list<br>
><br>
> > > <a href="mailto:getdp@geuz.org">getdp@geuz.org</a><br>
><br>
> > > <a href="http://www.geuz.org/mailman/listinfo/getdp" rel="noreferrer" target="_blank">http://www.geuz.org/mailman/listinfo/getdp</a><br>
><br>
> ><br>
><br>
> ><br>
><br>
> > _______________________________________________<br>
><br>
> > getdp mailing list<br>
><br>
> > <a href="mailto:getdp@geuz.org">getdp@geuz.org</a><br>
><br>
> > <a href="http://www.geuz.org/mailman/listinfo/getdp" rel="noreferrer" target="_blank">http://www.geuz.org/mailman/listinfo/getdp</a><br>
><br>
> ><br>
><br>
> > _______________________________________________<br>
><br>
> > getdp mailing list<br>
><br>
> > <a href="mailto:getdp@geuz.org">getdp@geuz.org</a><br>
><br>
> > <a href="http://www.geuz.org/mailman/listinfo/getdp" rel="noreferrer" target="_blank">http://www.geuz.org/mailman/listinfo/getdp</a><br>
><br>
><br>
><br>
> --<br>
><br>
> Prof. Christophe Geuzaine<br>
><br>
> University of Liege, Electrical Engineering and Computer<br>
</div></div>> Sciencehttp://<a href="http://www.montefiore.ulg.ac.be/~geuzaine" rel="noreferrer" target="_blank">www.montefiore.ulg.ac.be/~geuzaine</a><br>
<div class="HOEnZb"><div class="h5">><br>
><br>
><br>
> Tetrahedron V, July 4-5 2016: <a href="http://tetrahedron.montefiore.ulg.ac.be" rel="noreferrer" target="_blank">http://tetrahedron.montefiore.ulg.ac.be</a><br>
><br>
> Free software: <a href="http://gmsh.info" rel="noreferrer" target="_blank">http://gmsh.info</a> |<a href="http://getdp.info" rel="noreferrer" target="_blank">http://getdp.info</a> |<a href="http://onelab.info" rel="noreferrer" target="_blank">http://onelab.info</a><br>
><br>
><br>
><br>
><br>
><br>
> _______________________________________________<br>
><br>
> getdp mailing list<br>
><br>
> <a href="mailto:getdp@geuz.org">getdp@geuz.org</a><br>
><br>
> <a href="http://www.geuz.org/mailman/listinfo/getdp" rel="noreferrer" target="_blank">http://www.geuz.org/mailman/listinfo/getdp</a><br>
><br>
><br>
</div></div></blockquote></div><br></div>