Investigation of Wilks’ method for uncertainty quantification of trace model for BWR spray cooling

Travis Mui, Tomasz Kozlowski

Research output: Contribution to conferencePaper

Abstract

Wilks’ formula has been popularly used to quantify the minimum amount of computational work required to meaningfully assess a model’s uncertainty, due to its nonparametric statistical nature that does not require knowledge of the distribution of the qualifying parameters of interest, nor does it limit the amount of considered input uncertain parameters in the simulation model. This approach is favorable due to considerable computational expense of typical nuclear safety simulations, providing a quantifiable number of code executions that can statistically verify a desired level of safety. However, there are various existing definitions and usages of Wilks’ theorem in such scenarios, which this present study aims to investigate and quantify for a real thermal-hydraulics experiment used for reactor safety licensing. In this work, the U.S. NRC TRACE thermal-hydraulics code was chosen to simulate the separate-effect spray cooling tests for licensing BWR SVEA-64 fuel performed by ASEA-ATOM. The computational model was evaluated by performing forward uncertainty quantification (UQ) using Dakota as the analysis tool and code driver on 31 identified sensitive parameters. Using this validated model, sets of 1000 directly-sampled TRACE models for two input parameter probability distributions were collected in order to assess the applicability of Wilks’ theorem within a realistic nuclear safety analysis scenario, and compared to previous results of two other input parameter probability distributions. The obtained results will compare various Wilks-defined ‘sample sizes’ according to one-sided confidence intervals for the 1st, 2nd and 3rd-order statistics, along with the two-sided confidence interval for the 1st-order statistics. The comparison will quantify that Wilks’ method is valid at the 95%/95% tolerance/confidence level as determined by the U.S. NRC for reactor safety licensing modeling.

Original languageEnglish (US)
StatePublished - Jan 1 2016
Event17th International Topical Meeting on Nuclear Reactor Thermal Hydraulics, NURETH 2017 - Xi'an, Shaanxi, China
Duration: Sep 3 2017Sep 8 2017

Other

Other17th International Topical Meeting on Nuclear Reactor Thermal Hydraulics, NURETH 2017
CountryChina
CityXi'an, Shaanxi
Period9/3/179/8/17

Fingerprint

sprayers
licensing
Cooling
cooling
reactor safety
confidence
safety
hydraulics
Probability distributions
theorems
Hydraulics
Statistics
statistics
intervals
simulation
Uncertainty
Experiments
Hot Temperature

Keywords

  • BWR spray cooling
  • TRACE
  • Uncertainty quantification
  • Wilks method

ASJC Scopus subject areas

  • Nuclear Energy and Engineering
  • Instrumentation

Cite this

Mui, T., & Kozlowski, T. (2016). Investigation of Wilks’ method for uncertainty quantification of trace model for BWR spray cooling. Paper presented at 17th International Topical Meeting on Nuclear Reactor Thermal Hydraulics, NURETH 2017, Xi'an, Shaanxi, China.

Investigation of Wilks’ method for uncertainty quantification of trace model for BWR spray cooling. / Mui, Travis; Kozlowski, Tomasz.

2016. Paper presented at 17th International Topical Meeting on Nuclear Reactor Thermal Hydraulics, NURETH 2017, Xi'an, Shaanxi, China.

Research output: Contribution to conferencePaper

Mui, T & Kozlowski, T 2016, 'Investigation of Wilks’ method for uncertainty quantification of trace model for BWR spray cooling' Paper presented at 17th International Topical Meeting on Nuclear Reactor Thermal Hydraulics, NURETH 2017, Xi'an, Shaanxi, China, 9/3/17 - 9/8/17, .
Mui T, Kozlowski T. Investigation of Wilks’ method for uncertainty quantification of trace model for BWR spray cooling. 2016. Paper presented at 17th International Topical Meeting on Nuclear Reactor Thermal Hydraulics, NURETH 2017, Xi'an, Shaanxi, China.
Mui, Travis ; Kozlowski, Tomasz. / Investigation of Wilks’ method for uncertainty quantification of trace model for BWR spray cooling. Paper presented at 17th International Topical Meeting on Nuclear Reactor Thermal Hydraulics, NURETH 2017, Xi'an, Shaanxi, China.
@conference{a5c500b680b94d819760d7600182ca29,
title = "Investigation of Wilks’ method for uncertainty quantification of trace model for BWR spray cooling",
abstract = "Wilks’ formula has been popularly used to quantify the minimum amount of computational work required to meaningfully assess a model’s uncertainty, due to its nonparametric statistical nature that does not require knowledge of the distribution of the qualifying parameters of interest, nor does it limit the amount of considered input uncertain parameters in the simulation model. This approach is favorable due to considerable computational expense of typical nuclear safety simulations, providing a quantifiable number of code executions that can statistically verify a desired level of safety. However, there are various existing definitions and usages of Wilks’ theorem in such scenarios, which this present study aims to investigate and quantify for a real thermal-hydraulics experiment used for reactor safety licensing. In this work, the U.S. NRC TRACE thermal-hydraulics code was chosen to simulate the separate-effect spray cooling tests for licensing BWR SVEA-64 fuel performed by ASEA-ATOM. The computational model was evaluated by performing forward uncertainty quantification (UQ) using Dakota as the analysis tool and code driver on 31 identified sensitive parameters. Using this validated model, sets of 1000 directly-sampled TRACE models for two input parameter probability distributions were collected in order to assess the applicability of Wilks’ theorem within a realistic nuclear safety analysis scenario, and compared to previous results of two other input parameter probability distributions. The obtained results will compare various Wilks-defined ‘sample sizes’ according to one-sided confidence intervals for the 1st, 2nd and 3rd-order statistics, along with the two-sided confidence interval for the 1st-order statistics. The comparison will quantify that Wilks’ method is valid at the 95{\%}/95{\%} tolerance/confidence level as determined by the U.S. NRC for reactor safety licensing modeling.",
keywords = "BWR spray cooling, TRACE, Uncertainty quantification, Wilks method",
author = "Travis Mui and Tomasz Kozlowski",
year = "2016",
month = "1",
day = "1",
language = "English (US)",
note = "17th International Topical Meeting on Nuclear Reactor Thermal Hydraulics, NURETH 2017 ; Conference date: 03-09-2017 Through 08-09-2017",

}

TY - CONF

T1 - Investigation of Wilks’ method for uncertainty quantification of trace model for BWR spray cooling

AU - Mui, Travis

AU - Kozlowski, Tomasz

PY - 2016/1/1

Y1 - 2016/1/1

N2 - Wilks’ formula has been popularly used to quantify the minimum amount of computational work required to meaningfully assess a model’s uncertainty, due to its nonparametric statistical nature that does not require knowledge of the distribution of the qualifying parameters of interest, nor does it limit the amount of considered input uncertain parameters in the simulation model. This approach is favorable due to considerable computational expense of typical nuclear safety simulations, providing a quantifiable number of code executions that can statistically verify a desired level of safety. However, there are various existing definitions and usages of Wilks’ theorem in such scenarios, which this present study aims to investigate and quantify for a real thermal-hydraulics experiment used for reactor safety licensing. In this work, the U.S. NRC TRACE thermal-hydraulics code was chosen to simulate the separate-effect spray cooling tests for licensing BWR SVEA-64 fuel performed by ASEA-ATOM. The computational model was evaluated by performing forward uncertainty quantification (UQ) using Dakota as the analysis tool and code driver on 31 identified sensitive parameters. Using this validated model, sets of 1000 directly-sampled TRACE models for two input parameter probability distributions were collected in order to assess the applicability of Wilks’ theorem within a realistic nuclear safety analysis scenario, and compared to previous results of two other input parameter probability distributions. The obtained results will compare various Wilks-defined ‘sample sizes’ according to one-sided confidence intervals for the 1st, 2nd and 3rd-order statistics, along with the two-sided confidence interval for the 1st-order statistics. The comparison will quantify that Wilks’ method is valid at the 95%/95% tolerance/confidence level as determined by the U.S. NRC for reactor safety licensing modeling.

AB - Wilks’ formula has been popularly used to quantify the minimum amount of computational work required to meaningfully assess a model’s uncertainty, due to its nonparametric statistical nature that does not require knowledge of the distribution of the qualifying parameters of interest, nor does it limit the amount of considered input uncertain parameters in the simulation model. This approach is favorable due to considerable computational expense of typical nuclear safety simulations, providing a quantifiable number of code executions that can statistically verify a desired level of safety. However, there are various existing definitions and usages of Wilks’ theorem in such scenarios, which this present study aims to investigate and quantify for a real thermal-hydraulics experiment used for reactor safety licensing. In this work, the U.S. NRC TRACE thermal-hydraulics code was chosen to simulate the separate-effect spray cooling tests for licensing BWR SVEA-64 fuel performed by ASEA-ATOM. The computational model was evaluated by performing forward uncertainty quantification (UQ) using Dakota as the analysis tool and code driver on 31 identified sensitive parameters. Using this validated model, sets of 1000 directly-sampled TRACE models for two input parameter probability distributions were collected in order to assess the applicability of Wilks’ theorem within a realistic nuclear safety analysis scenario, and compared to previous results of two other input parameter probability distributions. The obtained results will compare various Wilks-defined ‘sample sizes’ according to one-sided confidence intervals for the 1st, 2nd and 3rd-order statistics, along with the two-sided confidence interval for the 1st-order statistics. The comparison will quantify that Wilks’ method is valid at the 95%/95% tolerance/confidence level as determined by the U.S. NRC for reactor safety licensing modeling.

KW - BWR spray cooling

KW - TRACE

KW - Uncertainty quantification

KW - Wilks method

UR - http://www.scopus.com/inward/record.url?scp=85052517096&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85052517096&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:85052517096

ER -