RELIABILITY OF THE EPR FUEL POOL COOLING SYSTEM USING A DYNAMIC APPROACH

摘要

One of the important issues for PSA analysis is to fully consider safety systems with their dynamic behaviour and the possibility to include operational properties and procedures. In the static traditional approach, it is not easy to introduce such dynamic phenomena in a the event tree model and one may need to use some dynamic framework to solve such problems. In this paper, we consider the Boolean Markov Driven Processes (BDMP) to model a safety system of a nuclear power plant. The main objective is to model functional dependencies, component recoveries, time dependant or conditional failures/recoveries and the possibility to use special configurations with "extra"-alignments. Thanks to a declarative knowledge based tool, these features can be embedded in such models in a compact form that may be instantiated in different ways with respect to the configuration or state of the system. Indeed, the BDMP framework allows to define such dynamic models using a fault-tree like construction with interesting mathematical properties. In particular, the possibility to reduce the combinatorial explosion problems inherent to Markov models. This allows to quantify the models and get the different reliability measures in reasonable times. The dynamic approach offered by the BDMP is particularly useful to model very redundant systems such as FA3 EPR FCPS (Fuel Pool Cooling System). The FCPS consists in three trains: two identical main trains, each equipped with two pumps in parallel, and a third train, fully independent. The complexity of the dependencies between each line can only be apprehended by a dynamic model and the BDMP allows a more realistic approach to model accident scenarios. The BDMP model of the FCPS as well as the reliability results obtained are presented in this article.