Correlation of errors in the Monte Carlo fission source and the fission matrix fundamental-mode eigenvector
2016 (English)In: Annals of Nuclear Energy, ISSN 0306-4549, E-ISSN 1873-2100, Vol. 94, 415-421 p.Article in journal (Refereed) PublishedText
Previous studies raised a question about the level of a possible correlation of errors in the cumulative Monte Carlo fission source and the fundamental-mode eigenvector of the fission matrix. A number of new methods tally the fission matrix during the actual Monte Carlo criticality calculation, and use its fundamental-mode eigenvector for various tasks. The methods assume the fission matrix eigenvector is a better representation of the fission source distribution than the actual Monte Carlo fission source, although the fission matrix and its eigenvectors do contain statistical and other errors. A recent study showed that the eigenvector could be used for an unbiased estimation of errors in the cumulative fission source if the errors in the eigenvector and the cumulative fission source were not correlated. Here we present new numerical study results that answer the question about the level of the possible error correlation. The results may be of importance to all methods that use the fission matrix. New numerical tests show that the error correlation is present at a level which strongly depends on properties of the spatial mesh used for tallying the fission matrix. The error correlation is relatively strong when the mesh is coarse, while the correlation weakens as the mesh gets finer. We suggest that the coarseness of the mesh is measured in terms of the value of the largest element in the tallied fission matrix as that way accounts for the mesh as well as system properties. In our test simulations, we observe only negligible error correlations when the value of the largest element in the fission matrix is about 0.1. Relatively strong error correlations appear when the value of the largest element in the fission matrix raises above about 0.5. We also study the effect of the error correlations on accuracy of the eigenvector-based error estimator. The numerical tests show that the eigenvector-based estimator consistently underestimates the errors in the cumulative fission source when a strong correlation is present between the errors in the fission matrix eigenvector and the cumulative fission source (i.e., when the mesh is too coarse). The error estimates are distributed around the real error value when the mesh is sufficiently fine.
Place, publisher, year, edition, pages
Elsevier, 2016. Vol. 94, 415-421 p.
Correlation, Eigenvector, Error, Fission matrix, Fission source, Monte Carlo, Correlation methods, Errors, Mesh generation, Monte Carlo methods, Correlation of errors, Criticality calculations, Error correlation, Fission matrixes, Fission source distribution, Fission sources, Strong correlation, Unbiased estimation, Eigenvalues and eigenfunctions
IdentifiersURN: urn:nbn:se:kth:diva-186898DOI: 10.1016/j.anucene.2016.04.013ISI: 000377231600046ScopusID: 2-s2.0-84964504934OAI: oai:DiVA.org:kth-186898DiVA: diva2:929300
QC 201605182016-05-182016-05-162016-07-05Bibliographically approved