Quantification of uncertainty in computational models of nuclear systems is required for assessing margins of safety for both design and operation of those systems. The largest source of uncertainty in computational models of nuclear systems derives from the nuclear cross section data used for modeling. There are two parts to cross section uncertainty data: the relative uncertainty in the cross section at a particular energy, and how that uncertainty is correlated with the uncertainty at all other energies. This cross section uncertainty and uncertainty correlation is compiled as covariance data. High fidelity covariance data exists for a few key isotopes, however the covariance data available for many structural materials is considered low fidelity, and is derived primarily from integral measurements with little meaningful correlation between energy regions. Low fidelity covariance data is acceptable for materials to which the operating characteristics of the modeled nuclear system are insensitive. However, in some cases, nuclear systems can be sensitive to isotopes with only low fidelity covariance data. Such is the case for the new U(19.5%)-10Moly foil fuel form to be produced at the Y-12 National Security Complex for use in research and test reactors. This fuel is ten weight percent molybdenum, the isotopes of which have only low fidelity covariance data. Improvements to the molybdenum isotope covariance data would benefit the modeling of systems using the new fuel form. This dissertation provides a framework for deriving high fidelity molybdenum isotope covariance data from a set of elemental molybdenum experimental cross section results. Additionally, a series of critical experiments featuring the new Y-12 fuel form was designed to address deficiencies in the critical experiment library with respect to molybdenum isotopes. Along with existing molybdenum sensitive critical experiments, these proposed experiments were used as a basis to compare the performance of the new high fidelity molybdenum covariance data set with the existing low fidelity covariance data set using the nuclear modeling code SCALE. The use of the high fidelity covariance data was found to result in reduced overall bias, reduced bias due to the molybdenum isotopes, and improved goodness-of-fit of computational results to experimental results. / text

Understanding the dynamics of agricultural basins has been difficult for decision-makers when developing cost-effective plans. An uncertainty analysis evaluates the impact of information gaps on hydrologic model’s output and performance. The Agricultural Policy/Environmental Extender (APEX v1501) was used to predict runoff, crop yield, sediment load, total phosphorus, and total nitrogen from agricultural fields in the Mississippi Delta to investigate the impact of using different input variables (climate, soils, and management practices) on mechanistic models. Results indicated that the use of surrogate information such as weather data from close weather stations, a predominant soil series, and simulated irrigation schedules, could be considered when available in situ information is restricted. Overall results provided information on model setup and output interpretation that may be useful to Mississippi Delta decision-makers.

Le secteur du bâtiment est en évolution et se doit de répondre aux problématiques environnementales actuelles aussi bien dans la construction neuve que dans la réhabilitation. La simulation thermique offre le moyen de répondre à cette problématique, mais se limite à une estimation des performances spécifiques aux paramètres qui ont été définis. Ce travail de thèse propose une méthodologie basée sur les puissances de calcul actuelles pour évaluer le comportement du bâtiment sur différentes plages de variation des entrées. Ces variations permettront, en fonction des propriétés qui leur sont attribuées, de rechercher des solutions de réhabilitation ou d'évaluer l'incertitude sur les sorties du modèle thermique. Afin de réaliser cela, des méthodes d'analyse de sensibilité performantes sont utilisées à travers un outil d'analyse spécifiquement développé pour le logiciel de simulation EnergyPlus. Cet outil permet une évaluation quasi automatique du modèle énergétique de bâtiment via : des techniques d'échantillonnage éprouvées LHS et LP-Tau, des techniques d'analyse performante pour un nombre de simulations réduit RBD-FAST et RBD-SOBOL, une gestion avancée des entrées et sorties permettant une évaluation détaillée du bâtiment, quelle que soit sa géométrie. Les techniques d'analyse de sensibilité sont employées pour proposer une vision précise des principales relations présentes au sein du modèle. Pour répondre à la contrainte du coût de calcul, la recherche de solution est réalisée à travers un méta-modèle issu d'une décomposition en polynôme de chaos. Une interface graphique utilise cette approximation du modèle complexe du bâtiment pour proposer de façon instantanée une vision claire du comportement de chaque entrée sur les sorties, ainsi que leurs tendances et les plages de variation idéales afin de choisir des solutions. Il est possible de coupler la recherche de solution à l'analyse d'incertitude afin de proposer des solutions robustes et identifier les principales raisons d'un écart entre simulation et réalité. La méthodologie des travaux de thèse encourage à l'optimisation de la compréhension du modèle plus qu'à la recherche d'une solution particulière.L'objectif de la démarche est de fournir des outils d'analyse afin d'aider l'expert dans la recherche de solution et ainsi de mettre en place des représentations graphiques qui facilitent la compréhension du système bâtiment, pour un gain en transparence et l'intégration de nombreuses contraintes de projet. / The building sector is evolving and has to meet the current environmental issues in both new construction and rehabilitation. The thermal simulation provides the mean to address this problem, but the estimated performance is limited to a specific set of parameters that have been defined. This thesis proposes a methodology based on the increased computing power to assess the building behaviour on different ranges of the inputs variation. Depending on the properties allocated to them, the changes of the variations will allow to seek solutions to restore or to assess the uncertainty in the outputs of the thermal model. To achieve this, effective sensitivity analysis methods are used across an analysis tool specifically developed for the simulation software EnergyPlus. This tool allows an almost automatic evaluation of the energy model building thanks to the proven sampling techniques such as LHS and LP-Tau; the efficient analysis techniques like RBD-FAST and RBD-SOBOL in order to reduce the number of simulations; an advanced management of the inputs and the outputs for a detailed assessment of the building regardless of its geometry. The sensitivity analysis techniques are used to provide an accurate picture of these key relationships within the model. To meet the constraints of the computational cost, the search for a solution is achieved through a metamodel from a Polynomial Chaos Expansion PCE. A GUI uses the approximation of the complex model of the building to provide instantly a clear vision of the behaviour of each entry on the outputs, their trends and their ideal ranges of variation to choose solutions. It is possible to couple the search for a solution to the uncertainty analysis to provide robust solutions and identify the main reason for the gap between simulation and reality. The methodology of the PhD work promotes the optimization of the model understanding more than looking for a specific solution. The purpose of the approach is to provide analytical tools to assist the expert in the search for a solution and so, develop graphical representations that facilitate the understanding of the building system for a gain in transparency and integration of many design constraints.

<p>When simulating reactor physics experiments, uncertainties in nuclear data result in a bias between simulated and experimental values. For new reactor designs or for power reactor designs the bias can be estimated using a set of experiments. How- ever, the experiments used to estimate the bias must be applicable to the power reactor design of interest. Similarity studies can be performed to ensure this is the case. Here, potential experiments in the ZED-2 heavy water critical facility at Chalk River Laboratories were developed that would be applicable to the multiplication factor bias calculation of three thoria plutonia fuelled power reactor designs. The power reactor designs that were analyzed were the CANDU 6 with 37-element fuel bundles and 43-element fuel bundles, and a Canadian SCWR design with 78- element fuel assemblies. The power reactors were simulated using the code package SCALE 6.1 under burnup conditions that were determined using the lattice code DRAGON 3.06H and the diffusion code DONJON 3.02A. The intermediate results from DRAGON and DONJON were used to compare the benefits of various reactor designs. Various critical core configurations were then simulated in the ZED-2 re- actor using the SCALE 6.1 package. The similarities between the potential ZED-2 reactor experiments and the power reactors were analyzed. These results were used to design a set of experiments having sufficiently high completeness that they can be used as part of a bias calculation using the generalized linear least squares method. To do so a methodology was developed to guide the experiment set design process in which the fuel type, lattice arrangements, and coolant type are modified and the effects on the sensitivity coverage analyzed. A set of six experiments was designed for which all of the power reactor designs had a completeness of 0.7 or higher.</p> / Master of Applied Science (MASc)

La généralisation planifiée du Bâtiment à Énergie POSitive est un enjeu important de développement durable, notamment dans une vision à long terme. Cependant, concevoir de tels bâtiments à moindre impact environnemental et à des coûts maîtrisés pour le marché constitue une problématique complexe pour les professionnels du secteur. Ces travaux de thèse se sont ainsi intéressés au développement d’une méthodologie permettant de fournir une aide à la décision opérationnelle et robuste pour l’écoconception de maisons à énergie positive (MEPOS). Basée sur l’optimisation multicritère (via un algorithme génétique), la démarche proposée associe au sein d’une plateforme multi-outils, simulation thermique dynamique, analyse de cycle de vie, et fonctions de coût de construction, pour identifier, sur une base multicritère (front de Pareto) et sur le cycle de vie, des concepts de MEPOS performantes et fournir au décideur une description des meilleurs compromis. Également intégrées dans la plateforme, des méthodes d’analyses de sensibilité et d’incertitude offrent la possibilité de sélectionner les variables de conception les plus influentes en amont de l’optimisation, et d’évaluer le risque de non-robustesse d’une décision en aval de l’optimisation, notamment au regard des incertitudes sur le comportement des occupants (avec un modèle stochastique d’occupation), et de l’évolution prévisible du climat jusqu’à la fin du siècle. Via une collaboration avec un constructeur, la méthodologie a été appliquée en temps réel au sein d’un processus de conception intégrée d’une MEPOS. / The planned generalisation of Plus Energy Buildings constitutes an important topic for sustainable development, especially in a long term vision. However, designing such buildings at lower environmental impact and with competitive cost for the market constitutes a complex issue for the sector’s professionals. This thesis focused on the development of a methodology providing an operational and robust decision support to eco-design plus-energy houses. Based on multi-criteria optimisation (via a genetic algorithm), the proposed approach combines in a multi-tools platform, dynamic building energy simulation, life cycle assessment, and cost construction functions, in order to identify, on a multi-criteria (Pareto front) and life cycle basis, efficient plus-energy houses concepts and to provide decision maker a description of the best compromises. Also included in the platform, sensitivity and uncertainty analysis methods offer the possibility to select the most influent design variables before optimisation, and to evaluate the non-robustness risk of a decision after optimisation, especially regarding uncertainties on occupants’ behaviour (with a stochastic model of occupancy) and climate’s predictable evolution up to the end of the century. Via collaboration with a constructor, the methodology was applied in real-time throughout a integrated design process of a plus-energy house.