In the uncertainty analysis of severe accidents, which is becoming an integral part of the nuclear power plant safety assessment, the results of the Monte Carlo transformation of uncertain input parameter probability density functions may, in some cases, reveal several well-differentiated groups with significantly different consequences. Within the framework of The Nuclear Safety Institute (IBRAE) IBRAE’s participation in the International Atomic Energy Agency’s coordinated research project I31033, an uncertainty analysis of the SOCRAT code simulation of the QUENCH-06 experiment was performed. The experiment was conducted with a pressurized water reactor bundle under conditions that are expected during a severe accident terminated by reflooding the uncovered core.

Consideration of measurement uncertainties in the test simulation revealed a cliff edge effect that manifested itself in the far stronger degradation of the assembly with extensive melt formation. To supplement an earlier finding, sensitivity investigations were performed, which are reported in this paper. A new analysis was conducted using the previously obtained computational data and the reliability sensitivity index based on model output classification, which is widely used in system reliability analyses. The results showed that the occurrence of the cliff edge effect was mostly determined by the uncertainty of the electric power measurement. The obtained results were consistent with the physical understanding of the processes occurring in the test facility.