Assessment of smoothed particle hydrodynamics in application of high-wind risk analysis

Risk Informed Safety Margin Characterization (RISMC) is a modern methodology for the risk analysis of Nuclear Power Plant that uses probability margin approach to quantify impacts of risk to economics, reliability, and safety. In the traditional Probabilistic Risk Analysis (PRA) for NPP safety analysis, the status of certain Structures, Systems and Components (SSCs) is determined by comparing the load against the capacity, where the load is calculated deterministically by system codes. As for the uncertainty during the safety analysis, it’s usually accounted with a conservative analysis. However, for RISMC methodology, the uncertainty is treated directly by determining the load and capacity distribution of each SSCs, which also avoids conservatisms. In RISMC application, the load is treated as a distribution that varies under different scenarios and calculated with RISMC simulation toolkit, which contains multiple advanced simulation tools, and use modeling and simulations to predict the event progression and plant response. However, in the RISMC application for external hazards, the landscape simulations usually take days or weeks to finish. Therefore, an assessment process is needed for the RISMC simulation tool to ensure both accuracy and efficiency. Smoothed Particle Hydrodynamics (SPH), as a Lagrangian method, is one of the particle-based Computational Fluid Dynamics tools. Due to the capability of SPH in simulating fluid problem with complex interfacial structure, an SPH-based software named NEUTRINO [1] is selected as the RISMC tool for simulating the generation, propagation, and interaction of high wind with Nuclear Power Plant. Based on the assumption that NEUTRINO is applicable for the external-hazards risk analysis, the objective of this work is to assess the capability of NEUTRINO as the RISMC simulation tool in high wind risk analysis. Towards this objective, Figure of Merits and Quantity of Interests are determined firstly with PIRT process, based on which numerical benchmarks are set. Next, the results are compared against experimental data. If the simulation is stable and the accuracy is acceptable, simulation parameters are varied to identify the major source of uncertainty. Otherwise, numerical models are developed or suggested until the accuracy requirements are reached. In the future work, a relationship between the simulation accuracy and source of uncertainty will be determined such that the uncertainty can be predicted and suggestion on parameter selection can be made in real-scale scenario simulations.