DETERMINATION OF THE NEUTRON PARAMETER'S UNCERTAINTIES USING THE STOCHASTIC METHODS OF UNCERTAINTY PROPAGATION AND ANALYSIS

Main objectives of reactor dosimetry are the determination of the neutron flux and fluence. In industrial power reactors, they are also used to follow the vessel and internal structures neutron induced damages. The knowledge of the associated uncertainties represents a significant stake for nuclear industry due to the high uncertainty value, 10% 15% (one standard deviation) commonly found for the calculated neutron flux (E > I MeV). Thus, the CENSPEX is engaged in the present study in order to analyze and then to reduce uncertainties associated to the reactor dosimetry interpretation process. The first part of this paper presents the main steps of data processing. Dosimeters activities are treated using nuclear data, neutron computation results and irradiation conditions. This treatment constitutes a complex process where many entries are more or less correlated. Thus, assessment of the output uncertainties requires the implementation of rather complex methods. In order to achieve uncertainty determination, a specific software tool has been developed to automate the process and to perform Monte Carlo uncertainty propagation and sensitivity analysis. Data uncertainties identification and quantification are performed in particular with regard to covariance. Then the stochastic uncertainty propagation methodology is described and carried out on representative cases. In complement, a Monte Carlo sensitivity study based on Sobol indexes is achieved on these cases to find the most penalizing input uncertainties. A fine analysis is performed to point out whether the uncertainty is induced by the incriminate input data uncertainties or by the methodology process. The paper concludes by pointing out the needs in input data knowledge.