Loading…
On data assimilation with Monte-Carlo-calculated and statistically uncertain sensitivity coefficients
•Sensitivity coefficients from Monte Carlo neutron transport codes have statistical uncertainties that may affect adjustments of nuclear data with integral experiments.•xGLLS is proposed to account for the sensitivity uncertainties in the adjustment theory.•xGLLS constrains adjustments to nuclear da...
Saved in:
Published in: | Annals of nuclear energy 2020-01, Vol.135, p.106951, Article 106951 |
---|---|
Main Authors: | , , , , |
Format: | Article |
Language: | English |
Subjects: | |
Citations: | Items that this one cites Items that cite this one |
Online Access: | Get full text |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | •Sensitivity coefficients from Monte Carlo neutron transport codes have statistical uncertainties that may affect adjustments of nuclear data with integral experiments.•xGLLS is proposed to account for the sensitivity uncertainties in the adjustment theory.•xGLLS constrains adjustments to nuclear data and calculated values for large sensitivity uncertainty.•For practical applications, sensitivity uncertainties do not have a large influence.•A convergence criterion is proposed to limit the number of neutron histories in sensitivity calculations.
Sensitivity coefficients from Monte Carlo neutron transport codes have uncertainties that can affect nuclear data adjustments with integral experiments. This paper presents an extended version of Generalized Linear Least Squares (GLLS), called xGLLS, that accounts for these uncertainties. With very large sensitivity uncertainties, xGLLS constrains the nuclear data adjustments so that the posterior biases and uncertainties are larger than with GLLS. However, for the range of sensitivity uncertainties realistically encountered, xGLLS does not produce adjustments different from GLLS. This indicates that sensitivity uncertainties are not important compared to experimental, modeling, methodological, and nuclear data uncertainties. To balance a simulation’s accuracy with its computational cost, we recommend stopping a simulation once the uncertainty of a calculated integral parameter, caused by modeling and methodologies and by the sensitivities, is an order of magnitude smaller than that caused by nuclear data. |
---|---|
ISSN: | 0306-4549 1873-2100 |
DOI: | 10.1016/j.anucene.2019.106951 |