Current codes related to severe accidents in nuclear power plants and the mathematical models implemented in them have a certain degree of reliability, given the complexity of the phenomena involved. After the Fukushima Daiichi accident, this type of peculiarity became relevant due to the need to reproduce the data reported during the progression of the accident, and thus confirm the fidelity of the simulation of the codes. However, due to the lack of knowledge of the phenomena, the prediction of results by specialized codes sometimes differs, and the implementation of methods for statistical analysis associated with sensitivity and uncertainty helps infer the importance of key parameters during a severe accident.

The objective of this study is to show the capabilities of the Monte Carlo filtering technique methodology to handle simultaneously quantitative (physical variables) and qualitative (code options) input variables for a more robust sensibility analysis applied to uncertain variables with a major impact on two common figures of merit in boiling water reactor severe accident analysis: hydrogen production in the core and time of reactor pressure vessel breach during a severe accident short-term station blackout using the MAAP5 code. The resulting uncertain variables of major significance were compared against those resulting from applying the Spearman rank correlation coefficient and the CobWeb plot methods for the case of quantitative variables. In the case of qualitative variables, only the CobWeb plot method results could be used for comparison.

For the figure of merit hydrogen production, it was shown that all three methods were in agreement in determining that the temperature for cladding rupture and the porosity of the collapsed core region were the quantitative variables of major impact. Regarding the qualitative uncertain variables, the options for the UZrO mixture oxidation model (with two models) and the options of the Zr oxidation model (with four models) were found to be the variables of significance by the Cobweb plot and the Monte Carlo filtering technique methods.

The highlights are as follows:

1. Sensitivity and uncertainty analysis using MAAP5 for quantitative and qualitative variables.

2. Hydrogen production during a short-term station blackout.

3. Indicator of Correlation base on Monte Carlo filtering and Spearman rank correlation coefficient.