A New CE TSUNAMI-3D Capability for Calculating Undersampling Metrics and Biases

Journal Article · Wed Jun 15 04:00:00 EDT 2016 · Transactions of the American Nuclear Society

OSTI ID:22991925

Perfetti, Christopher; Rearden, Bradley ^[1]

---

1. Oak Ridge National Laboratory, P.O. Box 2008, Bldg. 5700, Oak Ridge, TN 37831-6170 (United States)

This study investigates the viability of several statistical metrics for predicting the undersampling biases in continuous-energy (CE) Monte Carlo transport simulation tally estimates. Undersampling is a phenomenon in which a Monte Carlo simulation fails to use enough particles per generation to adequately sample fission points in all regions of a problem, resulting in biases in tally estimates that are much larger than the statistical uncertainties, potentially leading to erroneous conclusions regarding system performance and safety. Previous work found that biases due to undersampling could be as large as several hundred per cent mille (pcm) for eigenvalue estimates and up to tens or hundreds of percent for flux tally estimates. This paper documents the encapsulation of undersampling metrics within CE TSUNAMI-3D in the SCALE code system and examines the correlation between the various undersampling metric scores and the magnitude of various undersampling biases. Of the metrics examined, the RHW metric, the Tally Entropy metric, and the uncertainty in the scores for these metrics produce scores that correlated well with the fraction of the undersampling biases in various eigenvalue, flux, and energy-integrated reaction rate tallies in an infinitely reflected model of a PWR fuel assembly. The average number of nonzero scores per generation encountered by a tally was also found to show promise at predicting the magnitude of undersampling biases, and it also found use in detecting and filtering extremely undersampled tallies in the application of the other metrics. Future work will strengthen these conclusions by improving the degree of convergence for the examined tally and undersampling bias estimates, apply these metrics to a more broad variety of systems, and investigate variance reduction approaches for mitigating undersampling in Monte Carlo tally estimates. (authors)

• OSTI ID: 22991925
• Journal Information: Transactions of the American Nuclear Society, Journal Name: Transactions of the American Nuclear Society Journal Issue: 1 Vol. 114; ISSN 0003-018X
• Country of Publication: United States
• Language: English

Related Subjects

42 ENGINEERING
97 MATHEMATICS AND COMPUTING
EIGENVALUES
ENCAPSULATION
FILTERS
FISSION
FUEL ASSEMBLIES
MONTE CARLO METHOD
PWR TYPE REACTORS
REACTION KINETICS
SAFETY