The collision probability approach to neutron transport can be used to obtain the energy-dependent neutron spectrum in nuclear reactor systems as well as other quantities of interest. This method makes the approximation that the neutron distribution is constant within homogeneous regions, or cells, in the system. This assumption restricts geometries that can be modeled by the collision probability approach. The geometry modeled is typically an infinite lattice of two homogeneous cells: a fuel pin cylinder and the coolant that surrounds it. The transport of neutrons between the homogeneous cells is done using probabilities describing the chance that a neutron having a collision in one cell has its next collision in another cell. These collision probabilities can be cast in terms of escape and transmission probabilities for each cell. Some methods exist that extend the collision probability approach to systems composed of more than two homogeneous cells. In this work, we present a novel collision probability method, based on previous work by Schneider et al. (2006a), for an arbitrary number of cells. The method operates by averaging the transmission probabilities across cells of the same shape, and thus assumes a certain level of homogeneity across all cells. When using multigroup cross sections, which the collision probability approach requires, it is necessary to consider the effect that a system's geometry and composition has on those multigroup cross sections. The cross sections must be computed in a way that accounts for the resonance self-shielding that may reduce the reaction rates in the resonance region. The process of developing self-shielded cross sections in a heterogeneous system utilizes an escape cross section. We compute this escape cross section using the same collision probabilities used to obtain the energy spectrum. Results are presented for simple two-cell systems, and preliminary results for four-cell simulations are also given. An extension to the method is provided that accounts for the fact that in thermal systems the assumption of homogeneity is not always valid. / text

Dans cette thèse, nous présentons une étude rigoureuse des barres d'erreurs et des sensibilités de paramètres neutroniques (tels le keff) aux données nucléaires de base utilisées pour les calculer. Notre étude commence au niveau fondamental, i.e. les fichiers de données ENDF et leurs incertitudes, fournies sous la forme de matrices de variance/covariance, et leur traitement. Lorsqu'un calcul méthodique et consistant des sensibilités est consenti, nous montrons qu'une approche déterministe utilisant des formalismes bien connus est suffisante pour propager les incertitudes des bases de données avec un niveau de précision équivalent à celui des meilleurs outils disponibles sur le marché, comme les codes Monte-Carlo de référence. En appliquant notre méthodologie à trois exercices proposés par l'OCDE, dans le cadre des Benchmarks UACSA, nous donnons des informations, que nous espérons utiles, sur les processus physiques et les hypothèses sous-jacents aux formalismes déterministes utilisés dans cette étude. / This dissertation presents a comprehensive study of sensitivity/uncertainty analysis for reactor performance parameters (e.g. the k-effective) to the base nuclear data from which they are computed. The analysis starts at the fundamental step, the Evaluated Nuclear Data File and the uncertainties inherently associated with the data they contain, available in the form of variance/covariance matrices. We show that when a methodical and consistent computation of sensitivity is performed, conventional deterministic formalisms can be sufficient to propagate nuclear data uncertainties with the level of accuracy obtained by the most advanced tools, such as state-of-the-art Monte Carlo codes. By applying our developed methodology to three exercises proposed by the OECD (UACSA Benchmarks), we provide insights of the underlying physical phenomena associated with the used formalisms.