Ranking of uncertain parameters for dynamic event tree analysis: A case study based on a station blackout scenario

Dynamic Event Tree analysis which couples thermal-hydraulic system models with safety system and operator response models is a realistic but computationally challenging approach for risk quantification in nuclear power plants. Dynamic Event Tree analysis should also include a framework to quantify uncertainty due to the relevant aleatory and epistemic parameters of the risk assessment model. Since the computational requirements do not scale well with the number of uncertain parameters, a first necessity is to reduce the number of parameters through a rigorous selection process based on sensitivity analysis. A first step in this direction is made in this work, which presents an exemplary parameter ranking study based on a Station Black Out scenario of a Pressurized Water Reactor design (Zion power plant). Thus, a Zion power plant model for Station Black Out analysis using the TRACE thermal-hydraulic system code is employed together with Pearson and Spearman correlation coefficient methods in order to rank thirteen uncertain parameters preliminarily selected from own judgment. The candidate parameters include initial and boundary (I/B) conditions for the core (e.g. initial power, decay heat, axial core power distribution) and for the system (e.g. steam generator secondary mass inventory, water level in accumulators) as well as four physical parameters (wall to vapor heat transfer coefficient, nucleate boiling heat transfer coefficient, interfacial drag in bubbly and pre-critical-heat-flux flow regimes) that are ranked according to their correlation with core damage frequency within the validity domain of the model (no-core melt). As a result of the study based on 158 realizations of the simplified core damage frequency estimate model, the interfacial drag (pre-critical-heat-flux regime), the axial power offset and the steam generator liquid mass inventory are the high rank parameters as identified by both Pearson and Spearman correlation coefficient methods. However, both methods failed to identify another dominant contributor that is the initial water volume in the accumulators. The reason is that neither Pearson nor Spearman can capture the contribution to the output of cross interactions between the input parameters. But in this scenario the intricate interaction of the accumulator with the primary system during the intermittent passive injection phase is shown to have a non-negligible impact on the distribution of the core failure probability. This result hints at non-linear interactions between the parameters of the risk model, and shows the limits of sensitivity analysis methods based on linear or monotonic correlation coefficients.