• Citer

Integrated Safety Margin Quantification by Combining Probabilistic and Deterministic Approaches

par Yu, Shizhen
Président du jury Henneaux, Pierre
Promoteur Labeau, Pierre-Etienne
Publication Non publié, 2026-03-13

Thèse de doctorat

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

The thesis develops an Integrated Safety Margin Quantification (ISMQ) methodology designed to advance the framework for nuclear safety assessment by both combining deterministic and probabilistic approaches and addressing a wide range of uncertainties more realistically than conventional approaches. The ISMQ approach moves beyond unnecessary conservatism in scenario selection and input data uncertainties by offering not only realistic measurements of the safety margin but also explicitly evaluating the most penalizing cases and estimating their probabilities. In doing so, it improves insight into the residual risk associated with rare but impactful accident sequences.A central theme of the methodology is the characterization of safety margin along two complementary dimensions: first, the probabilistic safety margin, which measures the likelihood that regulatory safety limits will be met, and second, the probability-weighted margin of those scenarios where the safety limit is not exceeded, which reflects how far the system remains from dangerous conditions, averaged over their weighted likelihood. Together, these metrics provide a richer and more nuanced view of plant safety than traditional approaches that may focus primarily on single threshold criteria.In developing the ISMQ framework, the thesis compares and analyses random, deterministic, and grid-based sampling strategies, examines alternative approaches to demonstrate the non-exceedance of safety limits—including brief excursion and multi-sequence considerations—and positions the method as broadly applicable to different types of nuclear safety studies. An extended version of the ISMQ methodology is formulated for Design Extension Condition (DEC) analyses, integrating scenario definition, uncertainty propagation, safety margin assessment, and sequence categorization into DEC A or B (being DEC with limited or significant core degradation), all while including uncertainties related to system availability, model parameters, and accident dynamics—even accounting for operator action timing. Two established approaches for uncertainty quantification are examined to strengthen this dimension of realism in the analysis.Furthermore, the thesis demonstrates the special relevance of the ISMQ methodology to passive system reactor designs through two case studies that utilize machine-learning-based sensitivity analysis to explore how uncertainties—both in physical parameters and logical controls—can interact in complex ways and impact accident progression. In one case (ESBWR), interactions among logical uncertainties governing passive system actuation are shown to have substantial effects on core cooling performance, while in another (CAP1400), the combined effects of physical and dynamic uncertainties amplify output variability and challenge traditional assessment techniques.In conclusion, the ISMQ methodology offers a robust and flexible methodological foundation for comprehensive, realistic, and risk-informed nuclear safety assessments. With its two-loop approach to uncertainty propagation, it not only meets the challenge of modern nuclear plant scenarios—including those with passive systems or complex operator actions—but also sets a solid basis for future developments in performance-based and reliability-informed safety analysis frameworks for both existing and advanced reactor designs.