Nuclear safety demonstration has traditionally relied on deterministic approaches, in which predefined accident scenarios are analyzed to verify compliance with safety criteria and to evaluate safety margins. While probabilistic methods have been developed in parallel to assess risk and identify significant accident sequences, their integration into safety margin quantification remains limited. This work proposes an Integrated Safety Margin Quantification (ISMQ) framework that leverages probabilistic considerations to enhance deterministic safety analyses.
The proposed methodology incorporates probabilistic insights to systematically explore a wide range of accident scenarios, accounting for different configurations of available safety systems as well as uncertainties in physical and modeling parameters. By embedding uncertainty quantification within a two-loop framework, the approach captures both variability in system conditions and uncertainties affecting accident progression, while maintaining computational efficiency through the use of optimized deterministic sampling techniques.
In addition, a machine-learning-based exploration is employed to identify the most penalizing scenario, thereby complementing the realistic safety margin assessment.
By integrating probabilistic considerations into the safety demonstration, the proposed approach extends beyond existing deterministic evaluations to provide a more comprehensive characterization of safety margin. This includes the identification of relevant and risk-informed scenarios, the assessment of the distance between best-estimate system responses and safety limits, and the quantification of the likelihood of exceeding these limits. The framework thus supports a more informed and realistic evaluation of safety margins.