Welcome to the PSAM 18 Abstract Status page.
Abstract KI176Full Paper + Presentation
Contribution Analysis of Uncertainty Parameters in MAAP5 Dynamic Event Tree Simulations of Station Blackout
Dynamic Event Tree (DET) analysis integrates thermal-hydraulic system models with safety system and operator response models and is widely used for dynamic risk quantification in nuclear power plants. While DET frameworks represent aleatory uncertainties using branching methods such as ADAPT, severe accident system codes employed in DET analyses contain significant epistemic uncertainties associated with physical modeling of severe accident phenomena. Therefore, a systematic framework to evaluate the impact of epistemic uncertainties on risk metrics is essential.
As a first step toward this objective, this study presents a parameter contribution analysis for a station blackout scenario in a pressurized water reactor. The Zion nuclear power plant model implemented in the MAAP5 severe accident code was adopted. Important physical phenomena during station blackout were categorized into thermal-hydraulics, core pre-damage, core post-damage, and lower head behavior. Thirty-one physical model parameters associated with these phenomena were selected as epistemic uncertainty parameters.
A total of 93 MAAP5 simulations were performed, and parameter importance was evaluated with respect to reactor vessel breach timing using SHAP-based contribution analysis. The results identified the maximum shear stress of the penetration tube, the heat sink heat transfer scale factor, lower head debris pool convective heat transfer, and collapse node porosity as dominant parameters influencing vessel failure progression.
The analysis further indicated that complex interactions between the accumulator system and the primary system during intermittent passive injection significantly affect the distribution of core damage progression, while having limited influence on reactor vessel breach probability. These findings suggest that explainable machine learning–based contribution analysis provides a useful tool for interpreting epistemic uncertainty effects in dynamic event tree–based severe accident risk assessments. The proposed framework contributes to enhancing transparency and interpretability in uncertainty propagation within dynamic PRA.
← Check another abstractAs a first step toward this objective, this study presents a parameter contribution analysis for a station blackout scenario in a pressurized water reactor. The Zion nuclear power plant model implemented in the MAAP5 severe accident code was adopted. Important physical phenomena during station blackout were categorized into thermal-hydraulics, core pre-damage, core post-damage, and lower head behavior. Thirty-one physical model parameters associated with these phenomena were selected as epistemic uncertainty parameters.
A total of 93 MAAP5 simulations were performed, and parameter importance was evaluated with respect to reactor vessel breach timing using SHAP-based contribution analysis. The results identified the maximum shear stress of the penetration tube, the heat sink heat transfer scale factor, lower head debris pool convective heat transfer, and collapse node porosity as dominant parameters influencing vessel failure progression.
The analysis further indicated that complex interactions between the accumulator system and the primary system during intermittent passive injection significantly affect the distribution of core damage progression, while having limited influence on reactor vessel breach probability. These findings suggest that explainable machine learning–based contribution analysis provides a useful tool for interpreting epistemic uncertainty effects in dynamic event tree–based severe accident risk assessments. The proposed framework contributes to enhancing transparency and interpretability in uncertainty propagation within dynamic PRA.