AxialGeoMultiscale Tutorial #308
Conversation
|
|
partitioned-pipe-multiscale/fluid3d-openfoam/system/controlDict
Outdated
Show resolved
Hide resolved
|
|
||
| simpleCoeffs | ||
| { | ||
| n (2 1 1); |
There was a problem hiding this comment.
With this, we are splitting per x-direction, meaning that only one rank will be participating in the coupling (assuming that the axis of the pipe is x).
It would be interesting (not necessarily in this PR) to check if the mapping also works in parallel when the interface is split between MPI ranks. I expect surprises.
partitioned-pipe-multiscale/fluid3d-openfoam/system/preciceDict
Outdated
Show resolved
Hide resolved
| </mesh> | ||
|
|
||
| <participant name="Fluid1D"> | ||
| <use-mesh name="Fluid1D-Mesh" provide="yes" /> |
There was a problem hiding this comment.
Note that we will soon need to convert this file (and some of the code) to preCICE v3: https://precice.org/couple-your-code-porting-v2-3.html
But let's only do it after merging the preCICE branch to develop, otherwise it will be a bit chaotic to test.
| readData | ||
| ( | ||
| Velocity | ||
| ); | ||
|
|
||
| writeData | ||
| ( | ||
| Pressure | ||
| ); |
There was a problem hiding this comment.
In his recent thesis, @thesamriel developed some new boundary conditions that would be interesting to try out here. No need to do it now, unless we really have issues on the interface. But you may want to have a look: https://mediatum.ub.tum.de/node?id=1696254&change_language=en
| <m2n:sockets from="Fluid1D" to="Fluid3D" exchange-directory=".." /> | ||
|
|
||
| <coupling-scheme:serial-implicit> | ||
| <time-window-size value="0.01" /> |
There was a problem hiding this comment.
Do we really need to couple so often? How important is this choice?
I assume that OpenFOAM still adapts its time step, but still, it currently looks like this is a requirement.
|
|
||
| <coupling-scheme:serial-implicit> | ||
| <time-window-size value="0.01" /> | ||
| <max-time value="20.0" /> |
There was a problem hiding this comment.
Do we need to run the case for so long to see clear results? For tutorials, I would prefer the case finishing in < 2 min runtime. I have not yet tried to run the case, though.
| @@ -0,0 +1,5 @@ | |||
| TimeWindow TotalIterations Iterations Convergence | |||
| vertex_ids = interface.set_mesh_vertex(mesh_id, positions) | ||
| precice_dt = interface.initialize() | ||
|
|
||
| # problem definition |
There was a problem hiding this comment.
Could you write a few sentences on what exactly you are trying to solve here?
I have zero experience with nutils, so it could be useful if at least @IshaanDesai could also do a quick check. I think you have discussed it already.
| ucons = diricons | ||
| stringintegral = '(p - ' + str(numpy.rint(p_read)) + ')^2 dS' | ||
| psqr = domain.boundary['right'].integral(stringintegral @ ns, degree=2) | ||
| pcons = solver.optimize('p', psqr, droptol=1e-15) |
There was a problem hiding this comment.
I barely understand what is going on here, but a tolerance of 1e-15 sounds too strict to me.
The goal of this PR is to have a tutorial case for the newly developed AxialGeoMultiscale mapping feature.
The tutorial consists of a 1D and a 3D pipe that exchange velocity and pressure values.
This PR is still a draft, because the tutorial case is not yet satisfactory - TO DO: