Fuzzy clustering classification of dynamic scenarios of digital I&C in NPPs

Safety analysis of Nuclear Power Plants (NPPs) employing digital Instrumentation and Control (I&C) implies that the influence of events timing and magnitudes on the accidental scenarios be taken into account. This entails the simulation of a number of scenarios much larger than that considered for the classical Event Tree/Fault Tree (ET/FT)-based analysis. Consequently, the post-simulation retrieval of information becomes quite difficult and computationally burdensome. This paper reports the results of an investigation with respect to the classification of the numerous accidental scenarios generated in a dynamic safety analysis of a NPP with digital I&C. The method investigated is based on a Fuzzy C-Means clustering algorithm in which the classification takes into account not only the final system states reached at the end of the accidental scenarios but also the timing of the occurring events, the fault magnitudes and the characteristics of the process evolution. A case study is considered regarding the scenarios generated by a SIMULINK model of the Lead Bismuth Eutectic experimental Accelerator Driven System (LBE-XADS) embedded within a Monte Carlo (MC) sampling procedure for injecting single and multiple faults at random times and of random magnitudes. The accidental scenarios generated are classified on the basis of three different system failure modes, which relate to the value reached by the diathermic oil secondary coolant temperature with respect to maximum and minimum safety threshold values set to avoid primary coolant thermal shocks and degradation of the oil physical and chemical properties.