ABSTRACT
The fault tree linking method for building Probabilistic Safety Assessment (PSA) models of nuclear power plants models accident sequences—combinations of safety system failures following an initiating event—by relatively small event trees. Failures of individual safety systems are modelled by fault trees. Such a model allows us to analyze selected accident scenarios or scenarios leading to a specific consequence. The analysis algorithm implemented in RiskSpectrum decomposes function events which fail along a sequence into minimal cutsets and (optionally) summarizes successful function events in an aggregate event, a so called success module. First order algorithms for quantification of a list of such minimal cutsets yield an approximate result. The new MCS BDD algorithm implemented in RiskSpectrum aims at improving this approximation. When computing resources suffice, it has the capability to quantify the minimal cutset list exactly. We evaluate the performance of this algorithm on real life models.
