Master of Science / Mechanical and Nuclear Engineering / Hitesh Bindra / This thesis extends the application of Lattice Boltzmann Methods (LBM) to radiation transport problems in thermal sciences and nuclear engineering. LBM is used to solve the linear Boltzmann transport equation through discretization into Lattice Boltzmann Equations (LBE). The application of weighted summations for the scattering integral as set forth by Bindra and Patil are used in this work. Simplicity and localized discretization are the main advantages of using LBM with fixed lattice configurations for radiation transport problems. Coupled solutions to radiation transport and material energy transport are obtained using a single framework LBM. The resulting radiation field of a one dimensional participating and conducting media are in very good agreement with benchmark results using spherical harmonics, the P₁ method. Grid convergence studies were performed for this coupled conduction-radiation problem and results are found to be first-order accurate in space. In two dimensions, angular discretization for LBM is extended to higher resolution schemes such as D₂Q₈ and a generic formulation is adopted to derive the weights for Radiation Transport Equations (RTEs). Radiation transport in a two dimensional media is solved with LBM and the results are compared to those obtained from the commercial software COMSOL, which uses the Discrete Ordinates Method (DOM) with different angular resolution schemes. Results obtained from different lattice Boltzmann configurations such as D₂Q₄ and D₂Q₈ are compared with DOM and are found to be in good agreement. The verified LBM based radiation transport models are extended for their application into coupled multi-physics problems. A porous radiative burner is modeled as a homogeneous media with an analytical velocity field. Coupling is performed between the convection-diffusion energy transport equation with the analytical velocity field. Results show that radiative transport heats the participating media prior to its entering into the combustion chamber. The limitations of homogeneous models led to the development of a fully coupled LBM multi-physics model for a heterogeneous porous media. This multi-physics code solves three physics: fluid flow, conduction-convection and radiation transport in a single framework. The LBE models in one dimension are applied to solve one-group and two-group eigenvalue problems in bare and reflected slab geometries. The results are compared with existing criticality benchmark reports for different problems. It is found that results agree with benchmark reports for thick slabs (>4 mfp) but they tend to disagree when the critical slab dimensions are less than 3 mfp. The reason for this disagreement can be attributed to having only two angular directions in the one dimensional problems.

abstract: Fission products in nuclear fuel pellets can affect fuel performance as they change the fuel chemistry and structure. The behavior of the fission products and their release mechanisms are important to the operation of a power reactor. Research has shown that fission product release can occur through grain boundary (GB) at low burnups. Early fission gas release models, which assumed spherical grains with no effect of GB diffusion, did not capture the early stage of the release behavior well. In order to understand the phenomenon at low burnup and how it leads to the later release mechanism, a microstructurally explicit model is needed. This dissertation conducted finite element simulations of the transport behavior using 3-D microstructurally explicit models. It looks into the effects of GB character, with emphases on conditions that can lead to enhanced effective diffusion. Moreover, the relationship between temperature and fission product transport is coupled to reflect the high temperature environment. The modeling work began with 3-D microstructure reconstruction for three uranium oxide samples with different oxygen stoichiometry: UO2.00 UO2.06 and UO2.14. The 3-D models were created based on the real microstructure of depleted UO2 samples characterized by Electron Backscattering Diffraction (EBSD) combined with serial sectioning. Mathematical equations on fission gas diffusion and heat conduction were studied and derived to simulate the fission gas transport under GB effect. Verification models showed that 2-D elements can be used to model GBs to reduce the number of elements. The effect of each variable, including fuel stoichiometry, temperature, GB diffusion, triple junction diffusion and GB thermal resistance, is verified, and they are coupled in multi-physics simulations to study the transport of fission gas at different radial location of a fuel pellet. It was demonstrated that the microstructural model can be used to incorporate the effect of different physics to study fission gas transport. The results suggested that the GB effect is the most significant at the edge of fuel pellet where the temperature is the lowest. In the high temperature region, the increase in bulk diffusivity due to excess oxygen diminished the effect of GB diffusion. / Dissertation/Thesis / Doctoral Dissertation Materials Science and Engineering 2014

This work is dealing with issues concerning mathematical and computer modeling of events describable with help of rst and second order differential equations. First part contains summary of basic procedures for searching differential equations using analythical methods, next chapter is focused on software used to solve chosen tasks and demonstration of speci c physics problems.

Ces travaux de thèse s’inscrivent dans le cadre du projet "IMPULSE" qui traite du développement d’un procédé innovant d’élaboration de multi-matériaux par courant pulsé et est financé par l’Agence National de la Recherche. Ils ont pour objectif d’étudier et de mettre en évidence la faisabilité, du frittage-assemblage sous charge par courants pulsés, d’un composite conducteur l’AgSn-O2 sur un support en cuivre. Cette problématique, en lien avec les applications industrielles de Schneider Electric Industries, a été abordée au travers de simulations numériques du procédé de frittage-assemblage et d’essais expérimentaux. Les travaux sur les étapes de frittage et d’assemblage ont pu être traités séparément. Les phénomènes qui interviennent lors du frittage par effet Joule et les effets spécifiques liés à l’utilisation de certaines formes ou fréquences de courant, divisent encore la communauté scientifique. Des essais de frittage et frittage-assemblage par chauffage résistif avec différents types de courant (pulsé, continu, 50 Hz) ont été réalisés et ont permis de mettre en évidence l’absence d’effets spécifiques associés aux courants pulsés dans le cas de l’Ag-SnO2. Par conséquence, un modèle électrocinétique classique stationnaire a été retenu concernant les aspects électriques du modèle macroscopique de frittage. Ces essais ont également révélé l’importance des résistances de contact électrique, présentes entre les outillages (poinçons) et l’échantillon, et de la résistance de contact thermique qui existe entre l’échantillon et la matrice. Le modèle thermique instationnaire choisi est couplé fortement au modèle électrocinétique. Les lois de comportement utilisées pour la masse volumique et les conductivités (électrique et thermique), qui interviennent dans le modèle Electro-Thermique (ET), tiennent compte des changements de microstructure grâce à l’utilisation de variables internes de « densification » et de « cohésion ». Les évolutions des résistances de contact électrique et thermique, mesurées sur un dispositif ex-situ, sont aussi implémentées dans le modèle ET.D’un point de vue mécanique, un modèle de Norton associé au critère de Green a été choisi pour modéliser le comportement viscoplastique de la matière et la compressibilité irréversible du matériau lors du frittage sous charge de l’Ag-SnO2. Les fonctions intervenant dans le critère dépendent de la densité relative, dont la cinétique de densification est calculée à partir de la trace du tenseur des vitesses de déformation irréversible. Les paramètres de la loi de comportement mécanique ont été identifiés par méthodes inverses, à l’aide des logiciels SiDoLo et Abaqus, à partir d’essais thermomécaniques spécifiques réalisés sur la machine Gleeble du LIMatB. La loi de comportement mécanique a été implémentée dans une bibliothèque spécifique du code de calcul par éléments finis Sysweld qui est utilisé pour la simulation numérique d’essais de frittage instrumentés. La concordance entre les résultats numériques et expérimentaux (tensions, températures, mesure extensométrique), est satisfaisante et les écarts restent inférieurs aux erreurs expérimentales. Concernant l’étape d’assemblage, une campagne de caractérisation de la tenue de l’assemblage Ag-SnO2/Cu, a été menée sur la machine Gleeble grâce à des essais de frittage-assemblage anisothermes. Différentes cinétiques thermiques et différentes températures maximales, ont été testées afin de mettre en évidence l’effet du temps et de la température. Des tests de cisaillement de l’assemblage, ont permis le calcul d’un observable afin de juger de la qualité de la liaison. Au vu des résultats, un modèle dépendant uniquement de la température atteinte dans l’échantillon a été développé afin d’estimer la tenue de l’assemblage Ag-SnO2/Cu. / This thesis is part of the "IMPULSE" project, which is financed by the NationalAgency of Research. This project concerns the development of innovative process to produce multimaterials by pulsed currents. The ability of sintering and joining Ag-SnO2 powder on a copper support in the same process under pressure by pulsed currents is investigated. This problematic, linked to industrial applications of Schneider Electric Industries SEI), has been approached through numerical simulations and experimental tests of sintering-joining. Sintering and joining steps have been dealt separately in this works. Sintering phenomena and specific effects of pulsed currents still divide the scientific community. Sintering and sintering-joining test by resistive heating thanks different kinds of current (pulsed, DC, AC) have been realized. They have enabled to highlight that there are not specific effects of pulsed currents in the Ag-SnO2 case. Consequently, a classical stationary electrokinetic model has been used for electrical aspects in the macroscopic sintering model. These tests have also revealed the importance of the contact resistance (CR) present between tools and sample, and more particularly the electrical CR between punches and sample and the thermal CR between die and sample. The non-stationary thermal model chose is strongly coupled with the electrokinetic model. Characterization tests have shown that electrical and thermal conductivities increase with inter-granular contact rate improvement, which is caused by strain during densification and by diffusion ("cohesion" mechanisms). The behavior laws used to calculate the density and the conductivities (electrical and thermal) of the Electrokinetic-Thermal model (ET), take into account these microstructural evolutions by mean of internal variables of "densification" and "cohesion". Electrical and thermal contact resistances, measured in LIMatB’s device versus pressure and temperature, are implemented in the ET model. From a mechanical point of view, a Norton model combined with a Green criterion has been chosen to modeling the viscoplastic behavior of matter and the irreversible compressibility of Ag-SnO2 material during sintering under pressure. The criterion functions depend on the relative density. The densification kinetic is calculated from the trace of the irreversible deformation kinetics. The properties (viscoplastic parameters, elasticity limit,...) of mechanical behavior law have been identified by inverse methods using SiDoLo and Abaqus software from thermo-mechanical tests achieved on LIMatB’s Gleeble machine. The mechanical properties don’t depend of cohesion mechanisms. The mechanical behavior law has been implemented in the finite element code Sysweld to simulate sintering tests. The agreement between numerical and experimental results (tensions, temperatures, extensometric measurements) is correct and the differences remain inferior to the experimental errors. Tests of joining of Ag-SnO2 on a copper support, non isothermal under low pressure, have been achieved on Gleeble machine. Different thermal kinetics and different maximal temperatures have been explored to highlight time and temperature effects on diffusion mechanisms at the interface. Shear tests of the joining have enabled the calculation of an observable to estimate the bonding quality. From these results, a model which only depends of temperature reached in the sample has been developed to estimate the Ag-SnO2/Cu joining resistance. This joining model could be easily integrated in the more complex sintering model.

In this dissertation we investigate methods of solving various optimization problems with PDE constraints, i.e. optimization problems that have a system of partial differential equations in the set of constraints, and develop frameworks for a number of practically inspired problems that were not considered in the literature before. Such problems arise in areas like fluid mechanics, chemical engineering, finance, and other areas where a physical system needs to be optimized. In most of the literature sources on PDE-constrained optimization only relatively simple systems of PDEs are considered, they are either linear, or the size of the system is small. On the contrary, in our case, we search for solution methods to problems constrained by large (8 to 10 equations) and non-linear systems of PDEs. More specifically, in the first part of the dissertation we consider a multiphysics phenomenon where electromagnetic and mechanical fields interact within an electrically conductive body, and develop the optimization framework to find an efficient way to control one field through another. We also apply the developed PDE-constrained optimization framework to a financial options portfolio optimization problem, and more specifically consider the case that to the best of our knowledge is not covered in the literature.

Shape memory alloys (SMAs) are materials that show reversible, thermo-elastic, diffusionless, displacive (solid to solid) phase transformation, due to the application of temperature and/ or stress (/strain). Among different SMAs, NiTi is a popular one. NiTi shows reversible phase transformation, the shape memory effect (SME), where irreversible deformations are recovered upon heating, and superelasticity (SE), where large strains imposed at high enough temperatures are fully recovered. Phase transformation process in NiTi SMA is a very complex process that involves the competition between developed internal strain and phonon dispersion instability. In NiTi SMA, phase transformation occurs over a wide range of temperature and/ or stress (strain) which involves, evolution of different crystalline phases (cubic austenite i.e. B2, different monoclinic variant of martensite i.e. B19', and orthorhombic B19 or BCO structures). Further, it is observed from experimental and computational studies that the evolution kinetics and growth rate of different phases in NiTi SMA vary significantly over a wide spectrum of spatio-temporal scales, especially with length scales. At nano-meter length scale, phase transformation temperatures, critical transformation stress (or strain) and phase fraction evolution change significantly with sample or simulation cell size and grain size. Even, below a critical length scale, the phase transformation process stops. All these aspects make NiTi SMA very interesting to the science and engineering research community and in this context, the present focuses on the following aspects. At first this study address the stability, evolution and growth kinetics of different phases (B2 and variants of B19'), at different length scales, starting from the atomic level and ending at the continuum macroscopic level. The effects of simulation cell size, grain size, and presence of free surface and grain boundary on the phase transformation process (transformation temperature, phase fraction evolution kinetics due to temperature) are also demonstrated herein. Next, to couple and transfer the statistical information of length scale dependent phase transformation process, multiscale/ multiphysics methods are used. Here, the computational difficulty from the fact that the representative governing equations (i.e. different sub-methods such as molecular dynamics simulations, phase field simulations and continuum level constitutive/ material models) are only valid or can be implemented over a range of spatiotemporal scales. Therefore, in the present study, a wavelet based multiscale coupling method is used, where simulation results (phase fraction evolution kinetics) from different sub-methods are linked via concurrent multiscale coupling fashion. Finally, these multiscale/ multiphysics simulation results are used to develop/ modify the macro/ continuum scale thermo-mechanical constitutive relations for NiTi SMA. Finally, the improved material model is used to model new devices, such as thermal diodes and smart dampers.

Les techniques sur la récupération d'énergie qui visent à permettre aux réseaux de capteurs sans fil (Wireless Sensor Network, WSN) de devenir autonomes, sont reconnues comme des élément cruciaux pour répondre aux futurs besoins des objets connectés portés par l'internet des objets (Internet of Things, IoT). C’est dans ce contexte que les matériaux fonctionnels piézoélectriques et magnétostrictifs, qui peuvent être utilisés dans une large gamme de systèmes de récupération d'énergie, ont un regain d’intérêt au cours de ces dernières années. Cette thèse porte sur la modélisation multiphysique de ces deux matériaux fonctionnels avec la méthode éléments finis et par la réduction de modèle pour les systèmes qui en résultent, sur la base de la décomposition propre généralisée (Proper Generalized Decomposition, PGD). La modélisation de ces matériaux fonctionnels reste difficile bien que la recherche dans ce domaine a été l'objet de plusieurs études depuis des décennies. Une multitude de difficultés existent, parmi lesquelles les trois suivantes qui sont largement reconnues. La première difficulté résulte de la description mathématique des propriétés de ces matériaux qui est compliquée ; ce qui est particulièrement vrai pour les matériaux magnétostrictifs pour lesquels leurs propriétés dépendent de facteurs environnementaux externes tels que la température, la contrainte et le champ magnétique d’excitation. La deuxième difficulté résulte des effets de couplage entre les champs électromagnétiques, élastiques et thermiques qui doivent être considérés mutuellement, ce qui est au-delà de la capacité de la plupart des outils de simulation existants. La troisième difficulté vient du fait que les systèmes deviennent de plus en plus compacts pour être intégrés et/ou embarqués. Dans ce cas la modélisation multi-échelle est nécessaire, ce qui signifie que des modèles numériques tridimensionnels (3D) doivent être employés. Le travail présenté ici fournit des solutions pour répondre aux difficultés mentionnées. Une modélisation multiphysique sur la base des formes différentielles est d'abord établie. Dans cette modélisation, les quantités sont discrétisés en utilisant les éléments de Whitney appropriés. Après la discrétisation, le système est résolu en un bloc unique, ce qui évite les itérations entre les solutions physiques différentes tout en conduisant à des convergences rapides. La formulation prend en compte, la loi de comportement linéaire des matériaux piézoélectriques, et une loi de comportement non linéaire pour les matériaux magnétostrictifs basée sur le principe de l’énergie libre exprimé par le modèle (Discrete Energy-Averaged Model, DEAM). La mise en œuvre de notre formulation permet de décrire les comportements des matériaux fonctionnels piézoélectriques et magnétostrictifs à des coûts numériques raisonnables. Suite à cela, deux algorithmes basés sur la PGD pour la réduction de modèle sont proposés. Ces deux algorithmes ont permis de réduire considérablement le problème dimensionnel des modèles multiphysiques tout en en conservant de très bonnes précisions. Les algorithmes proposés fournissent également des moyens pour gérer le couplage avec la non-linéarité d’une manière efficiente. L’ensemble de nos modèles sont vérifiés et validés par des exemples représentatifs. / The energy harvesting technology that aims to enable wireless sensor networks (WSN) to be maintenance-free, is recognized as a crucial part for the next generation technology mega- trend: the Internet of Things (IoT). Piezoelectric and magnetostrictive materials, which can be used in a wide range of energy harvesting systems, have attracted more and more interests during the past few years. This thesis focuses on multiphysics finite element (FE) modeling of these two materials and performing model reduction for resultant systems, based on the Prop- er Generalized Decomposition (PGD). Modeling these materials remains challenging although research in this area has been under- going over decades. A multitude of difficulties exist, among which the following three issues are largely recognized. First, mathematically describing properties of these materials is com- plicated, which is particularly true for magnetostrictive materials because their properties depend on factors including temperature, stress and magnetic field. Second, coupling effects between electromagnetic, elastic, and thermal fields need to be considered, which is beyond the capability of most existing simulation tools. Third, as systems becoming highly integrated whole-scale simulations become necessary, which means three dimensional (3D) numerical models should be employed. 3D models, on the other hand, quickly turns intractable if not properly built. The work presented here provides solutions in respond to the above challenges. A differential forms based multiphysics FE framework is first established. Within this frame- work quantities are discreted using appropriate Whitney elements. After discretization, the system is solved as a single block, thus avoiding iterations between different physics solutions and leading to rapid convergences. Next, the linear piezoelectric, and a free energy based nonlinear magnetostrictive constitutive model called Discrete Energy Averaged Model (DE- AM) are incorporated into the framework. Our implementation describes underlying material behaviors at reasonable numerical costs. Eventually, two novel PGD based algorithms for model reduction are proposed. With our algorithms, problem size of multiphysics models can be significantly reduced while final results of very good accuracy are obtained. Our algo- rithms also provide means to handle coupling and nonlinearity conveniently. All our methodologies are demonstrated and verified via representative examples.

The objective of this report was to implement battery cycling and an intermittent current interruption (ICI) method for determining battery resistance into a simple lithium-ion battery model in the finite element methods (FEM) program COMSOL Multiphysics, andevaluate how accurately the model reflects the behaviour of voltage and internal resistance with respect to experimental results. The ICI technique consists of repeating the steps of first having a longer charging period and then having a short current interruption, where the internal resistance is calculated from the voltage drop that occurs when the current is turned off. The model was evaluated against measurements, made with the same technique (ICI), on assembled NMC-graphite batteries. Codes written in the statistical programming language “R” were used to process the data from both COMSOL and the experiments. Both the batteries and the model were constructed with a reference electrode, to enable measurement of each electrode by itself. The results as documented in this report show that it is possible to simulate the measurement technique in COMSOL, but that both the resistance and voltage profiles differed quite a lot from the behaviour of the tested batteries. The resistance of the positive electrode did however give good results and it was possible to improve the model by changing some parameters. The magnitude of the resistance, which was already quite close, could be improved by changing the porosity and particle size, and the voltage profiles were improved when using voltage-data achieved from the real measurements.

The need for efficient, sustainable, and planned utilization of resources is ever more critical. In the U.S. alone, buildings consume 34.8 Quadrillion (1015) BTU of energy annually at a cost of $1.4 Trillion. Of this energy 58% is utilized for heating and air conditioning. Several building energy analysis tools have been developed to assess energy demands and lifecycle energy costs in buildings. Such analyses are also essential for an efficient HVAC design that overcomes the pitfalls of an under/over-designed system. DOE-2 is among the most widely known full building energy analysis models. It also constitutes the simulation engine of other prominent software such as eQUEST, EnergyPro, PowerDOE. Therefore, it is essential that DOE-2 energy simulations be characterized by high accuracy. Infiltration is an uncontrolled process through which outside air leaks into a building. Studies have estimated infiltration to account for up to 50% of a building’s energy demand. This, considered alongside the annual cost of buildings energy consumption, reveals the costs of air infiltration. It also stresses the need that prominent building energy simulation engines accurately account for its impact. In this research the relative accuracy of current air infiltration calculation methods is evaluated against an intricate Multiphysics Hygrothermal CFD building envelope analysis. The full-scale CFD analysis is based on a meticulous representation of cracking in building envelopes and on real-life conditions. The research found that even the most advanced current infiltration methods, including in DOE-2, are at up to 96.13% relative error versus CFD analysis. An Enhanced Model for Combined Heat and Air Infiltration Simulation was developed. The model resulted in 91.6% improvement in relative accuracy over current models. It reduces error versus CFD analysis to less than 4.5% while requiring less than 1% of the time required for such a complex hygrothermal analysis. The algorithm used in our model was demonstrated to be easy to integrate into DOE-2 and other engines as a standalone method for evaluating infiltration heat loads. This will vastly increase the accuracy of such simulation engines while maintaining their speed and ease of use characteristics that make them very widely used in building design.

Le couplage multiphysique devient important dans les domaines de l’ingénierie nucléaire et de l’informatique. La capacité d’obtenir des solutions précises pour des modèles réalistes est essentielle à la conception et l’autorisation des conceptions nouvelles de réacteurs nucléaires, surtout dans des situations d’accidents graves. Les modèles physiques qui décrivent le comportement des réacteurs nucléaires dans des conditions accidentelles sont : le transport des neutrons, la conduction/convection thermique, la thermomécanique du combustible et des structures de support, la stœchiométrie du combustible, et d’autres encore. Cependant cette thèse se concentre sur le couplage entre deux modèles, le transport des neutrons et la conduction/convection thermique.Le but de cette thèse est de développer un solveur multiphysique pour la simulation des accidents de réacteurs nucléaires. Le travail s’est focalisé à la fois sur l’environnement de simulation et sur le traitement des données pour de telles simulations.Ces travaux discutent le développement d’un solveur multiphysique basé sur la méthode Newton-Krylov sans la jacobienne (JFNK). Ce solveur inclut des solveurs linéaires et non-linéaires, accompagné des interfaces par le calcul des résidus aux codes existantes pour le transport des neutrons et la thermo hydraulique (APOLLO3 et MCTH respectivement). Une nouvelle formulation pour le résidu du transport de neutrons est explorée, qui réduit la taille de la solution et l’espace de recherche par un facteur important ; le résidu, au lieu d’être basé sur le flux angulaire, est basé sur la source de fission.La question de savoir si l’utilisation d’un flux fondamental pour l’homogénéisation des sections efficaces est suffisamment précise pendant les simulations transitoires rapides est aussi explorée. Il est montré que, dans le cas d’un milieu infini et homogène, l’utilisation des sections efficaces fabriquées avec un flux fondamental est significativement différente d’une solution de référence. Cette erreur est diminuée en utilisant un flux de pondération alternatif qui vient d’un calcul à dépendance temporelle ; soit avec un flux intégré en temps soit avec une solution asymptotique. Le flux intégré en temps vient d’une solution multiphysique sur un sous-domaine de l’accident et intégrée en temps. L’intégration en temps peut être réalisée sur plusieurs « morceaux » qui ont le même comportement temporel. La solution asymptotique vient d’un calcul de valeur propre alpha et emploie un ou plusieurs modes alpha comme flux de pondération. Entre les deux méthodes, la méthode avec un flux intégré en temps est plus précise, mais prend plus de temps.Le domaine d’application de ces nouvelles méthodes est étendu en étudiant les effets d’hétérogénéités spatiales et la discrétisation des macro-intervalles en temps. Premièrement, un cas avec des hétérogénéités spatiales et une perturbation locale est utilisé pour montrer que ces méthodes peuvent être utilisées pour l’homogénéisation au niveau des assemblages. Ces nouvelles méthodes fonctionnent mieux que la méthode traditionnelle avec un flux fondamental. Deuxièmement, une estimation a priori pour une discrétisation optimale est obtenue pour la méthode avec le flux intégré en temps. Il est montré que d’autres divisions du domaine en temps réduisent l’erreur sur plusieurs métriques jusqu’au moment où les erreurs numériques deviennent dominantes.Pour montrer que ces méthodes fonctionnent bien pour des calculs de grande taille, un calcul sur un cœur REB réduit est effectué. Cette simulation est basée sur un accident de chute de grappe dans un REB au démarrage. / Multiphysics coupling is becoming of large interest in the nuclear engineering and computational science fields. The ability to obtain accurate solutions to realistic models is important to the design and licensing of novel reactor designs, especially in design basis accident situations. The physical models involved in calculating accident behavior in nuclear reactors includes: neutron transport, thermal conduction/convection, thermo-mechanics in fuel and support structure, fuel stoichiometry, among others. However, this thesis focuses on the coupling between two models, neutron transport and thermal conduction/convection.The goal of this thesis is to develop a multiphysics solver for simulating accidents in nuclear reactors. The focus is both on the simulation environment and the data treatment used in such simulations.This work discusses the development of a multiphysics framework based around the Jacobian-Free Newton-Krylov (JFNK) method. The framework includes linear and nonlinear solvers, along with interfaces to existing numerical codes that solve neutron transport and thermal hydraulics models (APOLLO3 and MCTH respectively) through the computation of residuals. A new formulation for the neutron transport residual is explored, which reduces the solution size and search space by a large factor; instead of the residual being based on the angular flux, it is based on the fission source.The question of whether using a fundamental mode distribution of the neutron flux for cross section homogenization is sufficiently accurate during fast transients is also explored. It is shown that in an infinite homogeneous medium, using homogenized cross sections produced with a fundamental mode flux differ significantly from a reference solution. The error is remedied by using an alternative weighting flux taken from a time dependent calculation; either a time-integrated flux or an asymptotic solution. The time-integrated flux comes from the multiphysics solution of the accident on a subdomain and an integration in time. The integration can be broken into several “chunks” that capture similar time-dependent behavior. The asymptotic solution comes from an alpha-eigenvalue calculation and uses one or several alpha modes as the weighting flux. Between the two methods, the time-integrated flux is more accurate, but takes longer to obtain a solution.The usability of these new homogenization methods is further developed by studying the effects of spatial heterogeneities and of the discretization of the time-chunks. First, a case with spatial heterogeneities and a localized perturbation is used to show that these methods can be applied to assembly level homogenization. The new methods are shown to perform well with spatial heterogeneities when compared to using a traditional, fundamental mode, homogenization method. Second, an a priori estimate for an optimal time discretization is obtained for the time-integrated flux method. It is shown that further divisions of the time domain reduce the error for several metrics until numerical errors become dominant.To show that these methods work well for industrial sized calculations, a reduced size BWR core calculation is performed. This simulation is based on a rod-drop accident in a BWR core during startup.