Effect of the Target Motion Sampling temperature treatment method on the statistics and performance

Tuomas Viitanen (Corresponding Author), Jaakko Leppänen

Research output: Contribution to journalArticleScientificpeer-review

1 Citation (Scopus)

Abstract

Target Motion Sampling (TMS) is a stochastic on-the-fly temperature treatment technique that is being developed as a part of the Monte Carlo reactor physics code Serpent. The method provides for modeling of arbitrary temperatures in continuous-energy Monte Carlo tracking routines with only one set of cross sections stored in the computer memory. Previously, only the performance of the TMS method in terms of CPU time per transported neutron has been discussed. Since the effective cross sections are not calculated at any point of a transport simulation with TMS, reaction rate estimators must be scored using sampled cross sections, which is expected to increase the variances and, consequently, to decrease the figures-of-merit. This paper examines the effects of the TMS on the statistics and performance in practical calculations involving reaction rate estimation with collision estimators. Against all expectations it turned out that the usage of sampled response values has no practical effect on the performance of reaction rate estimators when using TMS with elevated basis cross section temperatures (EBT), i.e. the usual way. With 0 Kelvin cross sections a significant increase in the variances of capture rate estimators was observed right below the energy region of unresolved resonances, but at these energies the figures-of-merit could be increased using a simple resampling technique to decrease the variances of the responses. It was, however, noticed that the usage of the TMS method increases the statistical deviances of all estimators, including the flux estimator, by tens of percents in the vicinity of very strong resonances. This effect is actually not related to the usage of sampled responses, but is instead an inherent property of the TMS tracking method and concerns both EBT and 0 K calculations.
Original languageEnglish
Pages (from-to)217-225
JournalAnnals of Nuclear Energy
Volume82
DOIs
Publication statusPublished - 2015
MoE publication typeA1 Journal article-refereed

Fingerprint

Statistics
Sampling
Reaction rates
Temperature
Program processors
Neutrons
Physics
Fluxes
Data storage equipment

Keywords

  • nuclear reactors
  • Monte Carlo code
  • collision estimators
  • simulation
  • on-the-fly temperature treatment

Cite this

@article{317e8a2afa154a5ba872e8d65ca8d0bf,
title = "Effect of the Target Motion Sampling temperature treatment method on the statistics and performance",
abstract = "Target Motion Sampling (TMS) is a stochastic on-the-fly temperature treatment technique that is being developed as a part of the Monte Carlo reactor physics code Serpent. The method provides for modeling of arbitrary temperatures in continuous-energy Monte Carlo tracking routines with only one set of cross sections stored in the computer memory. Previously, only the performance of the TMS method in terms of CPU time per transported neutron has been discussed. Since the effective cross sections are not calculated at any point of a transport simulation with TMS, reaction rate estimators must be scored using sampled cross sections, which is expected to increase the variances and, consequently, to decrease the figures-of-merit. This paper examines the effects of the TMS on the statistics and performance in practical calculations involving reaction rate estimation with collision estimators. Against all expectations it turned out that the usage of sampled response values has no practical effect on the performance of reaction rate estimators when using TMS with elevated basis cross section temperatures (EBT), i.e. the usual way. With 0 Kelvin cross sections a significant increase in the variances of capture rate estimators was observed right below the energy region of unresolved resonances, but at these energies the figures-of-merit could be increased using a simple resampling technique to decrease the variances of the responses. It was, however, noticed that the usage of the TMS method increases the statistical deviances of all estimators, including the flux estimator, by tens of percents in the vicinity of very strong resonances. This effect is actually not related to the usage of sampled responses, but is instead an inherent property of the TMS tracking method and concerns both EBT and 0 K calculations.",
keywords = "nuclear reactors, Monte Carlo code, collision estimators, simulation, on-the-fly temperature treatment",
author = "Tuomas Viitanen and Jaakko Lepp{\"a}nen",
year = "2015",
doi = "10.1016/j.anucene.2014.08.033",
language = "English",
volume = "82",
pages = "217--225",
journal = "Annals of Nuclear Energy",
issn = "0306-4549",
publisher = "Elsevier",

}

Effect of the Target Motion Sampling temperature treatment method on the statistics and performance. / Viitanen, Tuomas (Corresponding Author); Leppänen, Jaakko.

In: Annals of Nuclear Energy, Vol. 82, 2015, p. 217-225.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Effect of the Target Motion Sampling temperature treatment method on the statistics and performance

AU - Viitanen, Tuomas

AU - Leppänen, Jaakko

PY - 2015

Y1 - 2015

N2 - Target Motion Sampling (TMS) is a stochastic on-the-fly temperature treatment technique that is being developed as a part of the Monte Carlo reactor physics code Serpent. The method provides for modeling of arbitrary temperatures in continuous-energy Monte Carlo tracking routines with only one set of cross sections stored in the computer memory. Previously, only the performance of the TMS method in terms of CPU time per transported neutron has been discussed. Since the effective cross sections are not calculated at any point of a transport simulation with TMS, reaction rate estimators must be scored using sampled cross sections, which is expected to increase the variances and, consequently, to decrease the figures-of-merit. This paper examines the effects of the TMS on the statistics and performance in practical calculations involving reaction rate estimation with collision estimators. Against all expectations it turned out that the usage of sampled response values has no practical effect on the performance of reaction rate estimators when using TMS with elevated basis cross section temperatures (EBT), i.e. the usual way. With 0 Kelvin cross sections a significant increase in the variances of capture rate estimators was observed right below the energy region of unresolved resonances, but at these energies the figures-of-merit could be increased using a simple resampling technique to decrease the variances of the responses. It was, however, noticed that the usage of the TMS method increases the statistical deviances of all estimators, including the flux estimator, by tens of percents in the vicinity of very strong resonances. This effect is actually not related to the usage of sampled responses, but is instead an inherent property of the TMS tracking method and concerns both EBT and 0 K calculations.

AB - Target Motion Sampling (TMS) is a stochastic on-the-fly temperature treatment technique that is being developed as a part of the Monte Carlo reactor physics code Serpent. The method provides for modeling of arbitrary temperatures in continuous-energy Monte Carlo tracking routines with only one set of cross sections stored in the computer memory. Previously, only the performance of the TMS method in terms of CPU time per transported neutron has been discussed. Since the effective cross sections are not calculated at any point of a transport simulation with TMS, reaction rate estimators must be scored using sampled cross sections, which is expected to increase the variances and, consequently, to decrease the figures-of-merit. This paper examines the effects of the TMS on the statistics and performance in practical calculations involving reaction rate estimation with collision estimators. Against all expectations it turned out that the usage of sampled response values has no practical effect on the performance of reaction rate estimators when using TMS with elevated basis cross section temperatures (EBT), i.e. the usual way. With 0 Kelvin cross sections a significant increase in the variances of capture rate estimators was observed right below the energy region of unresolved resonances, but at these energies the figures-of-merit could be increased using a simple resampling technique to decrease the variances of the responses. It was, however, noticed that the usage of the TMS method increases the statistical deviances of all estimators, including the flux estimator, by tens of percents in the vicinity of very strong resonances. This effect is actually not related to the usage of sampled responses, but is instead an inherent property of the TMS tracking method and concerns both EBT and 0 K calculations.

KW - nuclear reactors

KW - Monte Carlo code

KW - collision estimators

KW - simulation

KW - on-the-fly temperature treatment

U2 - 10.1016/j.anucene.2014.08.033

DO - 10.1016/j.anucene.2014.08.033

M3 - Article

VL - 82

SP - 217

EP - 225

JO - Annals of Nuclear Energy

JF - Annals of Nuclear Energy

SN - 0306-4549

ER -