Comparison of global sensitivity analysis methods : application to fuel behavior modeling

Timo Ikonen

Research output: Contribution to journalArticleScientificpeer-review

18 Citations (Scopus)

Abstract

Fuel performance codes have two characteristics that make their sensitivity analysis challenging: large uncertainties in input parameters and complex, non-linear and non-additive structure of the models. The complex structure of the code leads to interactions between inputs that show as cross terms in the sensitivity analysis. Due to the large uncertainties of the inputs these interactions are significant, sometimes even dominating the sensitivity analysis. For the same reason, standard linearization techniques do not usually perform well in the analysis of fuel performance codes. More sophisticated methods are typically needed in the analysis. To this end, we compare the performance of several sensitivity analysis methods in the analysis of a steady state FRAPCON simulation. The comparison of importance rankings obtained with the various methods shows that even the simplest methods can be sufficient for the analysis of fuel maximum temperature. However, the analysis of the gap conductance requires more powerful methods that take into account the interactions of the inputs. In some cases, moment-independent methods are needed. We also investigate the computational cost of the various methods and present recommendations as to which methods to use in the analysis.
Original languageEnglish
Pages (from-to)72-80
JournalNuclear Engineering and Design
Volume297
DOIs
Publication statusPublished - 2016
MoE publication typeA1 Journal article-refereed

Fingerprint

sensitivity analysis
Sensitivity analysis
modeling
ranking
interactions
linearization
Linearization
recommendations
costs
moments
comparison
method
analysis
Costs
simulation
Temperature
temperature
Uncertainty
cost
code

Keywords

  • complex structure
  • computational costs
  • fuel performance
  • gap conductances
  • global sensitivity analysis
  • maximum temperature
  • moment independents
  • standard linearization technique

Cite this

@article{5bda191d6fa643ae9244b7ceed5e357b,
title = "Comparison of global sensitivity analysis methods : application to fuel behavior modeling",
abstract = "Fuel performance codes have two characteristics that make their sensitivity analysis challenging: large uncertainties in input parameters and complex, non-linear and non-additive structure of the models. The complex structure of the code leads to interactions between inputs that show as cross terms in the sensitivity analysis. Due to the large uncertainties of the inputs these interactions are significant, sometimes even dominating the sensitivity analysis. For the same reason, standard linearization techniques do not usually perform well in the analysis of fuel performance codes. More sophisticated methods are typically needed in the analysis. To this end, we compare the performance of several sensitivity analysis methods in the analysis of a steady state FRAPCON simulation. The comparison of importance rankings obtained with the various methods shows that even the simplest methods can be sufficient for the analysis of fuel maximum temperature. However, the analysis of the gap conductance requires more powerful methods that take into account the interactions of the inputs. In some cases, moment-independent methods are needed. We also investigate the computational cost of the various methods and present recommendations as to which methods to use in the analysis.",
keywords = "complex structure, computational costs, fuel performance, gap conductances, global sensitivity analysis, maximum temperature, moment independents, standard linearization technique",
author = "Timo Ikonen",
year = "2016",
doi = "10.1016/j.nucengdes.2015.11.025",
language = "English",
volume = "297",
pages = "72--80",
journal = "Nuclear Engineering and Design",
issn = "0029-5493",
publisher = "Elsevier",

}

Comparison of global sensitivity analysis methods : application to fuel behavior modeling. / Ikonen, Timo.

In: Nuclear Engineering and Design, Vol. 297, 2016, p. 72-80.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Comparison of global sensitivity analysis methods : application to fuel behavior modeling

AU - Ikonen, Timo

PY - 2016

Y1 - 2016

N2 - Fuel performance codes have two characteristics that make their sensitivity analysis challenging: large uncertainties in input parameters and complex, non-linear and non-additive structure of the models. The complex structure of the code leads to interactions between inputs that show as cross terms in the sensitivity analysis. Due to the large uncertainties of the inputs these interactions are significant, sometimes even dominating the sensitivity analysis. For the same reason, standard linearization techniques do not usually perform well in the analysis of fuel performance codes. More sophisticated methods are typically needed in the analysis. To this end, we compare the performance of several sensitivity analysis methods in the analysis of a steady state FRAPCON simulation. The comparison of importance rankings obtained with the various methods shows that even the simplest methods can be sufficient for the analysis of fuel maximum temperature. However, the analysis of the gap conductance requires more powerful methods that take into account the interactions of the inputs. In some cases, moment-independent methods are needed. We also investigate the computational cost of the various methods and present recommendations as to which methods to use in the analysis.

AB - Fuel performance codes have two characteristics that make their sensitivity analysis challenging: large uncertainties in input parameters and complex, non-linear and non-additive structure of the models. The complex structure of the code leads to interactions between inputs that show as cross terms in the sensitivity analysis. Due to the large uncertainties of the inputs these interactions are significant, sometimes even dominating the sensitivity analysis. For the same reason, standard linearization techniques do not usually perform well in the analysis of fuel performance codes. More sophisticated methods are typically needed in the analysis. To this end, we compare the performance of several sensitivity analysis methods in the analysis of a steady state FRAPCON simulation. The comparison of importance rankings obtained with the various methods shows that even the simplest methods can be sufficient for the analysis of fuel maximum temperature. However, the analysis of the gap conductance requires more powerful methods that take into account the interactions of the inputs. In some cases, moment-independent methods are needed. We also investigate the computational cost of the various methods and present recommendations as to which methods to use in the analysis.

KW - complex structure

KW - computational costs

KW - fuel performance

KW - gap conductances

KW - global sensitivity analysis

KW - maximum temperature

KW - moment independents

KW - standard linearization technique

U2 - 10.1016/j.nucengdes.2015.11.025

DO - 10.1016/j.nucengdes.2015.11.025

M3 - Article

VL - 297

SP - 72

EP - 80

JO - Nuclear Engineering and Design

JF - Nuclear Engineering and Design

SN - 0029-5493

ER -