LSODE issues: particles outside wall? #332
Comments
|
Investigating a bit further, I've found that the "red" particles have a |
|
I've constructed a "cleaned" hdf5 file from the DEPO step that removes the particles with |
|
The deposition part of the code follows neutrals on straight lines till they reach a point where Ionization is calculated on a chord which is ~0.01 m outside the plasma domain. Becasue of the cylindircal gridding this can result in particle ioninzing slightly outside the VMEC This effect could be an issue if running with the If a particle does not deposit in the plasma, then we follow it to the wall or again a domain which is Could you please post your full log file up to and including the first error. It's hard to understand how the code is being run. It almost looks like you're running with |
|
Thanks @lazersos.
I don't think that's the issue. I'm not sure if I mentioned this above, but I added two additional triangles to 'catch' the beam. You can see one of them in the top Figure above, with a cluster of black points to the left of the cluster in the plasma. I don't know why the other wall panels don't seem to stop any particles. Is there some assumption of a normal direction? (Do triangles only stop the beam from one side?) I will post the DEPO run log, which proceeds without errors, and then the SLOW log. Depo run logSlowing run logNote that this is after eliminating (hopefully all!) of the particles that had |
|
I've now updated to the most recent commit, |
Having two-or-three issues with Beams3D. I'm not sure how they might all be connected, or not.
perlmutterusing version "4.07".I'm trying to do deposition and slowing runs on a W7X file. My immediate issue is that during the slowing runs I'm getting LSODE issues and finally crashing.
The LSODE issues are something to do with NaNs, shown here:
There is also
and finally traces like
I'm now looking at the .h5 file which is the output of the initial deposition step and seeing some unexpected outputs.
Unexpected result 1
Some markers are at the boundary upper and lower boundary box edges: how did they get there?
See the blue points in the image.
In the image, the beam lines are shown at right. I don't have a working wall file so I authored this 'rectangular torus' wall file and cut out three of the triangles so the beams can make it into the plasma.
I'm not sure how some of the markers at center and left got there --- did they need to pass through the walls? Did they start with a large random directional scatter?
(I have
NPOINC=3and I am plotting the second locational index.)Unexpected results 2
I would have expected that markers had positions (gyropositions?) inside the LCFS, indicated by s < 1. However, 7 particles have a neutral ionization point with 1 < S < 1.02, and many more have gyrocenter S in the same range or even up to 1.5:
First pic: markers with s > 0.
Second pic: there's light between the points and the LCFS.
Unexpected result 3
I was surprised to find that I couldn't load some Datasets of the HDF5 file in Mathematica, related to particle positions, because they had
nan. I haven't done a full investigation, but a set of 62 particles has, for its third positional index (the initial position of the gyrocenter)nanfor at leastR_lines,U_linesandvll_lines. These particles are located within the plasma itself, within the same region as the rest of the normally ionized particles.See the images here, where a few of them are shown in red:
Questions on a way forward
Here's a few things I'm considering:
.h5file from step 1 (deposition) to remove particles withnan-positions or which haves > 1. The latter may be important since, at the moment, I'm using the LCFS as the wall in Step 2 (slowing).s > 1particles from Step 1.Might one of these work? Has either been done before? Is there a better way to cull bad particles before Step2 or to make step2 more resilient?
The text was updated successfully, but these errors were encountered: