Lessons learned for nuclear safety studies from the quantification of input parameters uncertainties applied to cathare thermal-hydraulics code within the premium benchmark

Fabrice Fouet (1) , Pierre Probst (1)

1 IRSN/PSN-RES/SEMIA/BASICC - Bureau d’Analyse et de Simulation des Incertitudes, des Cœurs et des Combustibles

In nuclear safety studies, Best-Estimate (BE) codes may be used provided that uncertainties are added to the relevant output parameters before comparing them with the acceptance criteria. The uncertainty of output parameters comes mainly from the lack of knowledge of the input parameters (initial and boundary conditions, parameters of physical models…). Moreover, the BEMUSE (Best Estimate Methods for Uncertainty and Sensitivity Evaluation) benchmark clearly showed that the quantification of input parameters uncertainty is a key point in uncertainty quantification. In the great majority of the studies, the application of the Best-Estimate Plus Uncertainty (BEPU) analysis relies on three steps with a probabilistic representation of uncertainty: 1. The determination of the input parameters statistical characteristics 2. The propagation of uncertainties from inputs to output parameters 3. The evaluation of the 95th percentile of the output with a high degree of confidence. The first step is usually based on expert judgments and comparing experimental data, especially for physical models parameters which cannot be directly measured. The user effect is also very important in the determination of the statistical characteristics: range of variation and probability law. To reduce this user effect and to help the experts in their evaluation, some mathematical methods have been specifically developed. However, determining the intrinsic uncertainty of such input parameters appears very complex. The PREMIUM (Post BEMUSE REflood Models Input Uncertainty Methods) benchmark is a follow-up activity of the BEMUSE programme dedicated to the quantification of uncertainties of the physical models in thermal-hydraulics system codes. For its contribution to this exercise, IRSN has used a quantification methodology called DIPE: Determination of Input Parameters uncertaintiEs. This paper presents a numerical application of the DIPE methodology related to the physical models involved in the prediction of core reflooding based on FEBA and PERICLES experiments. This work has been done with CATHARE V2.5_2, a thermal-hydraulics system code used in safety studies. Lessons learned from this exercise are discussed in the paper.

Mots clés: Safety studies; Statistics; Uncertainty; Input parameters; DIPE; CATHARE; PREMIUM

Domaines: Statistiques [math.ST]

Fabrice Fouet, Pierre Probst. Lessons learned for nuclear safety studies from the quantification of input parameters uncertainties applied to cathare thermal-hydraulics code within the premium benchmark. ICAPP 2016, Apr 2016, San Francisco, United States. ⟨irsn-04106694⟩