What is best practice for pf_passive/wall discretization?

[jwai-cfs]

For background, we are using an IDS-description of the SPARC device which is then imported to the MEQ framework to do various equilibrium and axisymmetric control simulations.

1. We use the IDS to describe large macro features of the passive structures, and when it is imported into MEQ we use a discretization algorithm to break this into finer meshes. What is best practice for then saving the results of a simulation back into the IDS? It seems like information is lost if we just save the currents associated with macro elements, and lose some of the fine distribution detail. Is the idea to also save the discretization mesh back into the ids within pf_passive.loop.#.element ?

2. We use both the wall and pf_passive IDS's to describe the vacuum conductors. (So for axisymmetric simulation with MEQ, reading and writing to wall, pf_passive, pf_active, pf_plasma, and equilibrium). Is there any best practice or recommendation on which IDS's to use for axisymmetric sims? There are some redundancies where I could see using either pf_passive or wall or both, and same with pf_plasma / equilibrium.

[imbeauf]

1. You should use multiple loops to store the current distribution in the passive structures. In recent JT-60SA simulations, 72 loops were used. What would be the required number of loops if you really need to store your full passive structure mesh ?
2. The recommendation is to put all passive structures in pf_passive (even if they correspond to wall or vessel structures), so that the equilibrium solver can use only this IDS as input for the passive structures. If you are doing fixed equilibrium calculation, the equilibrium IDS can be used as an input for specifying the plasma boundary, otherwise it's purely an output of the equilibrium solver. Does this answer your question ?

[jwai-cfs]

Thanks, this is helpful!

1. The full passive structure mesh is 1000+ elements usually, so it is too much to specify by hand. I'm sure it could be less though we have not studied what is the minimum number of elements that gives realistic results. Okay, so based on your answer, maybe a good approach is to define our pf_passive.loops into a reasonable number like 50-100, and then these can be discretized further by codes that consume the IDS but we will save just the full loop currents.

2. Yes this answers it! We can consolidate it all to pf_passive.

[francescocarpanese]

Is there any case where using wall.descriptio_2d.vessel provides an advantage over pf_passive. If not, would it make sense to remove wall.descriptio_2d.vesse from the structure?

[imbeauf]

The vessel substructure is intended for generic machine description, so it must stay. The way users model the vessel for specific electromagnetic simulation may vary and is done in the dedicated pf_passive IDS. But you are right that having a j_phi node in the vessel structure is sort of duplicating the electromagnetic simulation purpose in the WALL IDS, so I would propose to remove j_phi (mark it as obsolescent since WALL is an active IDS).

[francescocarpanese]

Ok great. Having all the current data in pf_passive , while still keeping geometrical description in wall, will avoid confusion for the user.

In this direction I think it could help reviewing the uniformity of the geometrical descriptions between wall and pf_passive whenever possibile. For example,
loop(i1)/element(i2)/geometry/annulus has strong similarity to description_2d(i1)/vessel/unit(i2)/annular/thickness(:). This would be helpful for people developing magnetic equilibrium code of different types, as it reduces the number of interface needed to parse all different ids possible geometry discretisation.

[imbeauf]

I have created the PR #32
Your point of unifying the geometry descritpions in various places of the DD is more general, so I suggest we wait to address this is a more general way.

[francescocarpanese]

Thanks @imbeauf