Hybrid dynamic event tree sampling strategy in raven code

The RAVEN code has been under development at the Idaho National Laboratoiy since 2012. Its main goal is to create a mulli.purpose platform/or the deploying of oil the capabilities needed for Probabilistic Risk Assessment, uncertainty quan:flcatian and data mining analysis. RAVEN is currently equ4ped with three dWerent sampling sfrategies: Once-through samplers (Monte Carlo, Latin Hyper Cube, Straqfled and Grid Sampler), Adaptive Samplers (Adaptive Point Sampler) and Dynamic Even: Tree samplers (Traditional and Adaptive Iiynamic Event Trees). The main subject of this paper is about the development of a Dynamic Event Tree (DEl) sampler named "Hybrid Dynamic Event Tree" HD.E7. As other authors have already reported, among the dlirent type of uncertainties, it is possible to discern two principle types: aleatory and epistemic uncertainties. The classical Dynamic Even: Tree is in charge of treating the first class (aleatory) uncertainties: the dependence of the probabilistic risk assessment and analysis on the epistemic uncertainties are treated lv an initial MOnte Carlo sampling MCDET1. From each Monte Carlo sample, a DET analysis is ran (ln total, N :rees1. The Monte carlo employs a pre-sampling of the input space characterized b,v epistemic uacertanties. The consequent Dynamic Event Tree performs the exploration of the aleatc'ry space. In the R4 PEN code, a more general approach has been developed, flat limiting the exploration qf the epistemic space through a Monte Carlo method but using all the once-through sampling strategies RAVEN currently employs The user con combine a Latin Hyper Cube, Grid, Stratfled and Monte carlo sampling in order to explore the epistemic space, without any limitation. From this pre-sampllng, the Dynamic Event Tree sampler starts its aleatory space exploration.