I am experiencing strange behavior when varying the number of particles or threads and even get different results depending on whether shannon entropy is enabled or disabled.
The outputs I’m primarily looking at are keff and gamma and neutron dose.
Initially I was thinking it could have been system or version specific, but this same behavior has been seen on V0.14 and V0.15 on both a laptop (16 threads) and HPC (up to 112 threads).
I can’t see any clear pattern linking the inconsistency - any advice or suggestions to try would be greatly appreciated.
Photon transport is on. Source region is defined within a region, only fissionable.
Results shown in image - the dose rate of ~90rem/y is the more reasonable output, with keff ~1.06. All tests were with the same code except for the number of threads/particles/batches so I would have expected the results to only deviate a small amount. These tests were performed on V0.14.0.
Hi Scott, welcome to community.
I think there were some problem with your geometry, have you double check your geometry with geometry debug?
openmc.run(geometry_debug=True)
If you think the problem came from the photon transport feature, you could turn off this feature to see if the keff became more consistent, if not, this must be came from another problem with your model. Also I think Shannon entrophy is not the reason for this, but if you want to check again, you might want to turn off Shannon entrophy. The mission is to localized the problem.
Also, it is quite difficult for us to help since we cant imagine the case problem you encounter, maybe you want to elaborate some details on your model.
Thank you. This has resolved the issues!
I was not aware that geometry could be an issue if it ran, it is good to know that overlap is not checked for by default. The geometry plots in openmc-plotter looked fine also but that was not enough.
Thanks for your advice.
Scott
