Handling coupling to MOOSE

[Waqar-ukaea]

It might be a good idea to think of alternative ways to couple AEGIS outputs to MOOSE.

Previously I managed to couple AEGIS heatfluxes to MOOSE via writing them out in a csv format (x,y,z, heatflux) and using that to create a constant heat source function applied as a Neumann BC in a heat transfer solve in MOOSE. This works but I'm sure that there are better ways to do this.

Limitations of my current approach. (Will update as I think of any others):

• Relies on MOOSE's NearestReporterCoordinatesFunction which assigns a heatflux value to the nearest (spatially) node in the exodus mesh. I just don't think this is the best method, results in interpolation between nodes across elements and doesn't take into account the fact that I can now launch multiple particles per element.
• If I ever wanted to create a time dependent heatflux function in MOOSE which relied on multiple AEGIS "shots" with varying eqdsks/equilibrium definitions then I think this file based approach won't really suffice.
• With MOAB, I have the ability to attach metadata to elements in a mesh so I feel like there must be a way I can just attach this heatflux data to the exodus mesh that is read into MOOSE.

[Waqar-ukaea]

Had some more thoughts about this:

When working with the nearest reporter coordinates function it probably makes sense to simply launch a particle from each node in the surface mesh, that way the mapping of AEGIS heatflux to MOOSE function is 1:1. It will still suffer from the same interpolation error across elements (between nodes) but least this way we avoid any errors associated with assigning a heatflux that isn't as (spatially) close to a node as it could be.

I'm not entirely sure what the ideal situation looks like - I'll have to have more of a think about it. But I do think AEGIS calculating heatflux per element and MOOSE using those per element values will likely be more desirable than the current nodal based approach.

There is also possibly a route via platypus with MFEM coefficients. The way I previously handled coupling to MOOSE involving a MOOSE functor object to define the heatflux source is apparently very similar to the way MFEM coefficients work. I'm not sure whether or not I will be able to leverage them but MFEM also natively supports Raviart-Thomas elements which may be more suitable for representing a heatflux incident on the surface.