Inverse uncertainty quantification using the modular Bayesian approach based on Gaussian process, Part 1: Theory

Xu Wu, Tomasz Kozlowski, Hadi Meidani, Koroush Shirvan

Research output: Contribution to journalArticle

Abstract

In nuclear reactor system design and safety analysis, the Best Estimate plus Uncertainty (BEPU) methodology requires that computer model output uncertainties must be quantified in order to prove that the investigated design stays within acceptance criteria. “Expert opinion” and “user self-evaluation” have been widely used to specify computer model input uncertainties in previous uncertainty, sensitivity and validation studies. Inverse Uncertainty Quantification (UQ) is the process to inversely quantify input uncertainties based on experimental data in order to more precisely quantify such ad-hoc specifications of the input uncertainty information. In this paper, we used Bayesian analysis to establish the inverse UQ formulation, with systematic and rigorously derived metamodels constructed by Gaussian Process (GP). Due to incomplete or inaccurate underlying physics, as well as numerical approximation errors, computer models always have discrepancy/bias in representing the realities, which can cause over-fitting if neglected in the inverse UQ process. The model discrepancy term is accounted for in our formulation through the “model updating equation”. We provided a detailed introduction and comparison of the full and modular Bayesian approaches for inverse UQ, as well as pointed out their limitations when extrapolated to the validation/prediction domain. Finally, we proposed an improved modular Bayesian approach that can avoid extrapolating the model discrepancy that is learnt from the inverse UQ domain to the validation/prediction domain.

Original languageEnglish (US)
Pages (from-to)339-355
Number of pages17
JournalNuclear Engineering and Design
Volume335
DOIs
StatePublished - Aug 15 2018

Fingerprint

formulations
design analysis
Bayesian analysis
nuclear reactors
prediction
guy wires
predictions
systems engineering
acceptability
Uncertainty
specifications
safety
physics
methodology
evaluation
causes
output
Nuclear reactors
sensitivity
estimates

Keywords

  • Bayesian calibration
  • Gaussian process
  • Inverse uncertainty quantification
  • Model discrepancy
  • Modular Bayesian

ASJC Scopus subject areas

  • Nuclear and High Energy Physics
  • Nuclear Energy and Engineering
  • Materials Science(all)
  • Safety, Risk, Reliability and Quality
  • Waste Management and Disposal
  • Mechanical Engineering

Cite this

Inverse uncertainty quantification using the modular Bayesian approach based on Gaussian process, Part 1 : Theory. / Wu, Xu; Kozlowski, Tomasz; Meidani, Hadi; Shirvan, Koroush.

In: Nuclear Engineering and Design, Vol. 335, 15.08.2018, p. 339-355.

Research output: Contribution to journalArticle

@article{818448109b2141cf9e8331110ab452c5,
title = "Inverse uncertainty quantification using the modular Bayesian approach based on Gaussian process, Part 1: Theory",
abstract = "In nuclear reactor system design and safety analysis, the Best Estimate plus Uncertainty (BEPU) methodology requires that computer model output uncertainties must be quantified in order to prove that the investigated design stays within acceptance criteria. “Expert opinion” and “user self-evaluation” have been widely used to specify computer model input uncertainties in previous uncertainty, sensitivity and validation studies. Inverse Uncertainty Quantification (UQ) is the process to inversely quantify input uncertainties based on experimental data in order to more precisely quantify such ad-hoc specifications of the input uncertainty information. In this paper, we used Bayesian analysis to establish the inverse UQ formulation, with systematic and rigorously derived metamodels constructed by Gaussian Process (GP). Due to incomplete or inaccurate underlying physics, as well as numerical approximation errors, computer models always have discrepancy/bias in representing the realities, which can cause over-fitting if neglected in the inverse UQ process. The model discrepancy term is accounted for in our formulation through the “model updating equation”. We provided a detailed introduction and comparison of the full and modular Bayesian approaches for inverse UQ, as well as pointed out their limitations when extrapolated to the validation/prediction domain. Finally, we proposed an improved modular Bayesian approach that can avoid extrapolating the model discrepancy that is learnt from the inverse UQ domain to the validation/prediction domain.",
keywords = "Bayesian calibration, Gaussian process, Inverse uncertainty quantification, Model discrepancy, Modular Bayesian",
author = "Xu Wu and Tomasz Kozlowski and Hadi Meidani and Koroush Shirvan",
year = "2018",
month = "8",
day = "15",
doi = "10.1016/j.nucengdes.2018.06.004",
language = "English (US)",
volume = "335",
pages = "339--355",
journal = "Nuclear Engineering and Design",
issn = "0029-5493",
publisher = "Elsevier BV",

}

TY - JOUR

T1 - Inverse uncertainty quantification using the modular Bayesian approach based on Gaussian process, Part 1

T2 - Theory

AU - Wu, Xu

AU - Kozlowski, Tomasz

AU - Meidani, Hadi

AU - Shirvan, Koroush

PY - 2018/8/15

Y1 - 2018/8/15

N2 - In nuclear reactor system design and safety analysis, the Best Estimate plus Uncertainty (BEPU) methodology requires that computer model output uncertainties must be quantified in order to prove that the investigated design stays within acceptance criteria. “Expert opinion” and “user self-evaluation” have been widely used to specify computer model input uncertainties in previous uncertainty, sensitivity and validation studies. Inverse Uncertainty Quantification (UQ) is the process to inversely quantify input uncertainties based on experimental data in order to more precisely quantify such ad-hoc specifications of the input uncertainty information. In this paper, we used Bayesian analysis to establish the inverse UQ formulation, with systematic and rigorously derived metamodels constructed by Gaussian Process (GP). Due to incomplete or inaccurate underlying physics, as well as numerical approximation errors, computer models always have discrepancy/bias in representing the realities, which can cause over-fitting if neglected in the inverse UQ process. The model discrepancy term is accounted for in our formulation through the “model updating equation”. We provided a detailed introduction and comparison of the full and modular Bayesian approaches for inverse UQ, as well as pointed out their limitations when extrapolated to the validation/prediction domain. Finally, we proposed an improved modular Bayesian approach that can avoid extrapolating the model discrepancy that is learnt from the inverse UQ domain to the validation/prediction domain.

AB - In nuclear reactor system design and safety analysis, the Best Estimate plus Uncertainty (BEPU) methodology requires that computer model output uncertainties must be quantified in order to prove that the investigated design stays within acceptance criteria. “Expert opinion” and “user self-evaluation” have been widely used to specify computer model input uncertainties in previous uncertainty, sensitivity and validation studies. Inverse Uncertainty Quantification (UQ) is the process to inversely quantify input uncertainties based on experimental data in order to more precisely quantify such ad-hoc specifications of the input uncertainty information. In this paper, we used Bayesian analysis to establish the inverse UQ formulation, with systematic and rigorously derived metamodels constructed by Gaussian Process (GP). Due to incomplete or inaccurate underlying physics, as well as numerical approximation errors, computer models always have discrepancy/bias in representing the realities, which can cause over-fitting if neglected in the inverse UQ process. The model discrepancy term is accounted for in our formulation through the “model updating equation”. We provided a detailed introduction and comparison of the full and modular Bayesian approaches for inverse UQ, as well as pointed out their limitations when extrapolated to the validation/prediction domain. Finally, we proposed an improved modular Bayesian approach that can avoid extrapolating the model discrepancy that is learnt from the inverse UQ domain to the validation/prediction domain.

KW - Bayesian calibration

KW - Gaussian process

KW - Inverse uncertainty quantification

KW - Model discrepancy

KW - Modular Bayesian

UR - http://www.scopus.com/inward/record.url?scp=85048576082&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85048576082&partnerID=8YFLogxK

U2 - 10.1016/j.nucengdes.2018.06.004

DO - 10.1016/j.nucengdes.2018.06.004

M3 - Article

AN - SCOPUS:85048576082

VL - 335

SP - 339

EP - 355

JO - Nuclear Engineering and Design

JF - Nuclear Engineering and Design

SN - 0029-5493

ER -