• BEPU
• water cooled reactor
• reactor licensing
• uncertainty
• best estimate
• GRS method
• CIAU method

ABSTRACT
Safety analysis is of vital importance in the licensing process of a nuclear power plant. Historically it has been carried out with a conservative approach, a sufficient safety margins with respect to the “licensing limits” is provided using conservative assumption on boundary and operational conditions in the power plant. Today, the safety analyses are mostly performed using best estimate system codes, the results of such analysis are accepted by the authority control only if an appropriate evaluation of the uncertainty related to the results is provided.
The conservative assumptions were introduced, in the framework of the safety analysis, during the ’70 [1] to account for the uncertainty associated with the early system codes. The quantification of the safety margins is however impossible with this methodology, leading to a reduction of the economic performance of the plants.
Since that time, an extensive experimental effort has been carried out aiming to improve the knowledge of the nuclear power plant behaviour under transient scenarios. The development of Best Estimate (BE) computer codes are direct consequence of these huge experimental efforts. These codes are capable of providing more realistic information on the status of the plant, allowing the prediction of the “real” safety margins. Establishing The knowing of the real operational and transient conditions allows the operator to increase the performances of the plant without decreasing the safety margins.
The results of the best estimates codes (for instance Relap, Cathare, TRACE, Cathena, Athlet) are always affected by error, due to the intrinsic nature of the codes models and boundary conditions. Errors arise from the approximations necessary for the solution of the complex differential equations, from the uncertainty that affect the power plant parameters, and from errors made by the user of the codes. The last one, usually represents the biggest source in the final uncertainty.
Adequate procedures and methods have been developed to account for this lack of knowledge and user effect so that, estimates of the final uncertainty in the results of the BE codes can be made.
The uncertainty methods developed in different countries, from various industry and research groups, are classifiable in to three distinct typologies: propagation of input errors, propagations of output errors and the deterministic treatment of the uncertainty. Uncertainty methodologies based on the first two approaches have already been used for licensing purposes and constitute the state of the art for the industrial application of BEPU analysis.
Taking into consideration the framework described above, the main objective of this thesis is to contribute to the further development of these methodologies, two uncertainties methods has been applied in the course of this work, namely the GRS method [9] and the CIAU method [59]. The thesis provids also a state of the art of the uncertainty methodologies developed recently, and a discussion on the various approaches used for the licensing of nuclear power reactor recognized by the IAEA.
The GRS method [9] has been applied to evaluate the uncertainty on the average void production in the ATUCHA-2 reactor core following a LBLOCA. The ATUCHA-2 nuclear power plant is undergoing licensing in Argentina. The reactor has been designed with a positive void reactivity coefficient. Therefore establishing the void generation is of vital importance to evaluate the power peak following a LOCA, in order to verify that the performance of the safety systems are sufficient to guarantee an adequate safety margins with respect to the licensing limits.
The second methodology applied is the CIAU [51]. The goal of this activity is to demonstrate the maturity level of this tool, which has already been applied in the licensing of nuclear power plant [61, 62]. The development of a Relap5 model of the LOBI/MOD1 experimental facility, used to simulated two LOCA transient, in this thesis is an important achievement that allows for the simulation of other experiments performed in the facility itself, at some point in the future. Increasing the number of the simulation included in the uncertainty database increase the statistical performance of the CIAU tool.
The thesis work highlights the increasing role of the new approach for the licensing of the NPP. Due to the indisputable advantages of the BEPU method, it is possible to state that in the future the new plant will be licensed with this new approach. The application of the GRS methodology for addressing the uncertainty during the licensing of ATUCHA-2 is an example of this trend.