Dynamic Event Trees for Probabilistic Safety Analysis

Abstract: In technical systems like nuclear power plants, an accident sequence starts with an initiating event and evolves over time through the interaction of dynamics and stochastics. This interaction is capable of producing infinitely many different sequences. Along the time line they define a continuous dynamic event tree with infinitely many branch points. At each point of time, the stochastic variability of the accident consequences is summarized by a multivariate probability distribution. A probabilistic safety analysis (PSA) requires an approximation to this distribution for selected consequence variables. It is felt that the conventional event tree analysis of Level 1 and of Level 2 PSA does often not permit a satisfactory probabilistic representation for PSA purposes. For this reason various methods of probabilistic dynamics have been suggested over the past decade. This paper presents a recent development that combines dynamic event tree analysis with Monte Carlo simulation. The advantages of this combination are explained and illustrated by a practical application involving MELCOR as the dynamics code