A Deterministic/probalistic analysis of Ex-Vessel melt risk in a BWR

Abstract

The present study is concerned with deterministic and probabilistic analysis of ex-vessel melt risks in a Swedish designed BWR plant. The focus is placed on a station blackout (SBO) scenario, with immediate SCRAM and subsequent activation of the main steam valve isolation (at 52 s). Four sequences were examined in detail to study the effect of two valves systems related to the operation of ADS (Automatic Depressurization System), and cavity flooding by water from suppression pool. The later action constitutes a cornerstone in the SAM (Severe Accident Management) strategy adopted in the newer Swedish and Finnish BWR plants. On the deterministic side, the US NRC MELCOR code was used to simulate the BWR system behaviours and phenomena that govern threats to the containment integrity. Remarkably, time factor (such as delays in instrumentation activation) was found to be of paramount importance to the sequence’s dynamics and consequences. On the probabilistic side, a review of methods in Level 2 Probabilistic Risk Analysis (PRA) based on classical concepts of event trees-fault trees reveals that the existing methods ignore the time dependence, suggesting the need and potential value to explore a dynamic approach. In the present work, a recently-developed concept of dynamic fault trees was adopted. Specifically, dynamic structures were built into the tree’s branching points to enable capturing the effect of time-competing events in Level 2 PRA (severe accidents). The technique was successfully implemented in Matlab\Simulink. The method was then applied to four cases to compute the probability of occurrence of various key events. For the four cases studied, MELCOR results show the time for cavity rupture due to debris ablation varied between the 17.5 h and the 30.5 h. They also show direct containment heating (DCH) when ADS fails to operate. Interestingly, the probabilistic results exhibit associative and competitive behaviours in the two valve systems, confirming the need to account for their interdependence.