Application of uncertainty analysis methods to MELCOR simulation of postulated severe accidents in a Nordic BWR

Abstract

Different uncertainty analysis methods are applied to MELCOR simulation of two postulated severe accidents in a Nordic boiling water reactor (BWR): (i) station blackout (SBO) accident, and (ii) large break loss-of-coolant accident (LBLOCA) combined with SBO, with the objective to compare their performances in the estimates of 95/95 tolerance limits of two figures of merit (FOMs) – the hydrogen mass produced from core degradation and the timing of vessel failure. Given 17 uncertain input parameters of MELCOR with probability density functions (PDFs), the 95/95 estimates of the two FOMs are obtained through the uncertainty analysis. From the uncertainty analysis results, it is found that for the quantification of single FOM a larger sample size leads to a much more accurate and stable 95/95 estimate at a higher computational cost, and the three nonparametric methods (Wilks’ method, Beran and Hall’s linear interpolation method as well as Hutson fractional statistics method) behave similarly in both accidents, while the goodness-of-fit test method performs differently and tends to provide a more realistic 95/95 estimate in both accidents. For the quantification of multiple FOMs the bracketing method tends to provide a smaller 95/95 estimate than the Wald and Guba method does, in consistent with their mathematical definitions. The Wald and Guba method is more stringent than the bracketing method when all percentiles (coverage) are set as the same. The sensitivity analysis results show that the several most significant input parameters are ranked almost identically by Spearman rank correlation coefficient (SRCC) and Pearson correlation coefficient (PCC), but these coefficients are dependent on accident scenarios and output parameters. Among the 17 parameters chosen, molten cladding drainage rate is the most influential to the output parameters (timing of initial melt relocation, timing of vessel failure, residual heat, etc.) considered in the present study, probably due to its impacts on molten Zr exposure to steam and resulting oxidation.