A methodology for determining interactions in probabilistic safety assessment models by varying one parameter at a...

Area of Science:

• Risk Analysis and Management
• Quantitative Modeling
• Nuclear Safety Engineering

Background:

• Risk analysis relies on quantitative models for decision-making.
• Understanding factor interactions is crucial for interpreting model results.
• Computational complexity often limits sensitivity analysis to one-parameter-at-a-time methods, hindering interaction assessment.

Purpose of the Study:

• To develop and illustrate a methodology for quantifying interactions in probabilistic safety assessment (PSA) models.
• To enable the assessment of individual factor contributions versus joint actions in risk analysis.
• To address the limitations of traditional sensitivity methods in complex models.

Main Methods:

• Utilizing a property of functional ANOVA decomposition for finite changes.
• Applying a one-parameter-at-a-time variation approach to quantify interactions.
• Testing the methodology with a case study on a nuclear reactor's core damage frequency.

Main Results:

• The methodology successfully quantifies the relevance of individual factors and their interactions.
• Numerical results revealed a nonadditive model structure in the analyzed nuclear reactor accident.
• The direction of risk change (increase or decrease) due to factor variations and cooperation was identified.

Conclusions:

• The proposed method effectively quantifies interactions in complex PSA models.
• This approach enhances the understanding of risk apportionment between individual and joint factor effects.
• It provides valuable insights for risk management and decision-making in safety-critical systems.