Hi Everyone,
I want to ask regarding batches when running OpenMC. As users we need to specify how many batches of particles to simulate and how many particles are in each batch. But I still do not understand the difference between the active and inactive. Alsa I want to know the reasons that make OpenMC result more accurate, is it because of the number of batches or the number of particles or both? I really appreciate for any explanation.
Thanks
Kristina
Hi Kristina,
In a k-eigenvalue problem, the distribution of source neutrons is not known at the beginning of the simulation, so the first thing the code needs to do is iterate on the fission source until it “converges”. At that point, it is ok to start tallies since they will no longer be affected by the initial guess of the source distribution, which is specified by the user. The batches during which the fission source is still converging are referred to as “inactive” because no tallies are scored during those batches. Once source convergence is reached, tallies are turned on (hence “active”).
Regarding your second question, the statistical uncertainty will decrease both with the number of particles per batch as well as the number of active batches. To be more precise, the variance of the mean for a tally estimator is inversely proportional to the product of the number of active batches and the number of particles per batch.
Hope this helps!
Best,
Paul
Hi,
Is there any way to understand/estimate the number of inactive batches that is needed to converge the fission source??or it's users choice??
Thanks,
Imtiaj
Hi Imtiaj,
Make trial runs and see when the oscillation in keff is somewhat stabilized. Tallying keff (active batches) may begin after that many number of batches (inactive). If it is a 1m by 1m reactor, inactive batches=50, particles per batch=50000, may be sufficient. I think it depends on the model size and smoothness of geometry.
Also, initial source guess should contain coordinates from most regions in problem geometry for faster source convergence or lower number of inactive batches.
Refer LA-UR-09-3136 for more clarity.
-Vedha
Thank u veda, i got it.
Imtiaj
For a k-eigenvalue calculation, you need to ensure that both keff and the fission source distribution have converged before you begin active batches. While checking that keff converges is a necessary condition, it is often not sufficient to ensure that the fission source distribution has also converged. In fact, it has been proven that keff converges faster than the source distribution, and so other means of assessing whether the source has converged are needed. One commonly-used method is to calculate the so-called “Shannon entropy” of the source distribution and use that as a proxy for the convergence of the source distribution. Please have a look at this notebook which demonstrates how it can be used in OpenMC. This technical report from Forrest Brown at LANL also has some good technical details on the background that you may wish to review.
Best regards,
Paul
Thanks a lot. It was really helpful.