Application of the method of uncertainty and sensitivity evaluation to results of PWR LBLOCA analysis calculated with the code ATHLET

Usually, safety analyses are based on the traditional conservative deterministic approach. However, the conservative approach does not allow evaluating the accuracy of results and their deviations from the real values. An alternative approach to reactor safety analyses uses best estimate computer codes together with quantification of uncertainties in model and plant parameters. Uncertainties in the results of safety analysis calculations are related to model uncertainties as well as deficiencies of knowledge of reactor initial conditions.
The German Gesellschaft für Anlagen- und Reaktorsicherheit (GRS) has developed an approach to uncertainty analysis based on the Wilks’ theorem. According to that theorem, the number of calculations to be performed to get a probabilistic estimation of an output parameter with certain coverage at a certain confidence level (e.g., 95%/95%) does not depend on the number of statistically independent input uncertain parameters.
An important part of the analysis is devoted to the selection of the uncertain parameters and definition of the probability distributions and intervals of parameter variation. The variations should cover all parameters which are necessary to describe all phenomena relevant for the considered scenario. The parameter selection is based on a Phenomena Identification and Ranking Table (PIRT) elaborated based on engineering judgment.
In this paper, the results of an uncertainty and sensitivity analysis for a Large Break LOCA scenario for a generic German PWR of the Konvoi design are reported about. The accident was performed using the system code ATHLET Mod 2.2 Cycle A. 47 uncertain parameters were statistically varied, among them 32 model parameters, 7 plant initial state parameters and 7 core parameters. The EXEL-integrated software SUSA developed by GRS is used for generation of the uncertainty parameter vectors, their incorporation into ATHLET input files and for the result analysis. Investigation of the sensitivity of the output parameter (peak cladding temperature) from the uncertain input parameters reveals the most relevant parameters.
Results of this reference uncertainty analysis and a related sensitivity analysis are presented in this report.