Correlations between nuclear data and results of integral slab experiments. Case of hafnium

The aim of this thesis was to evaluate how much integral slab experiments can both reduce discrepancies between experimental results and calculations, and improve the knowledge of hafnium isotopes neutronic parameters by an adapted sensitivity and uncertainty method. A statistical approach, based on the generalized least squares method and perturbation theory, has been incorporated into our calculation system in order to deduce microscopic cross-section adjustments from observed integral measurements on this particular 'mock-up' reactor. In this study it has been established that the correlations between integral parameters and hafnium capture cross-sections enable specific variations in the region of resolved resonances at the level of multigroup and punctual cross-sections recommended data (JEF-2.2 evaluation) to be highlighted. The use of determinist methods (APOLLO2 code) together with Monte Carlo- type simulations (TRIPOLI4 code) enabled a depth analysis of the modelling approximations to be carried out. Furthermore, the sensitivity coefficient validation technique employed leads to a reliable assessment of the quality of the new basic nuclear data. In this instance, the adjustments proposed for certain isotope 177Hf resonance parameters reduce, after error propagation, by 3 to 5 per cent the difference between experimental results and calculations related to this absorbent's efficiency. Beyond this particular application, the qualification methodology integrated in our calculation system should enable other basic sizing parameters to be treated (chemical / geometric data or other unexplored nuclear data) to make technological requirements less stringent. (author)