Finite volume methods employing second-order gradient reconstruction schemes are often utilized to computationally solve the governing equations of fluid mechanics and transport. These schemes, while not as dissipative as first-order schemes, frequently produce oscillatory solutions in regions of discontinuities and/or unsatisfactory levels of dissipation in smooth regions of the variable field. Limiters are often employed to reduce the inherent variable over- and under-shoot; however, they can significantly increase the numerical dissipation of a solution, eroding a scheme’s performance in smooth regions. A novel gradient reconstruction scheme, which shows significant improvement over traditional second-order schemes, is presented in this work. Two implementations of this Optimization-based Gradient REconstruction (OGRE) scheme are examined: minimizing an objective function based on the mismatch between local reconstructions at midpoints or selected quadrature points between cell stencil neighbors. Regardless of the implementation employed, the resulting gradient calculation is a compact, implicit method that can be used with unstructured meshes by employing an arbitrary computational stencil. An adjustable weighting parameter is included in the objective function that allows the scheme to be tuned towards either greater accuracy or greater stability. To address over- and undershoot of the variable field near discontinuities, non-local, non-monotonic (NLNM) and local, non-monotonic (LNM) limiters have also been developed, which operate by enforcing cell minima and maxima on dependent variable values projected to cell faces. The former determines minimum and maximum values for a cell through recursive reference to the minimum and maximum values of its upwind neighbors. The latter determines these bounding values through examination of the extrema of values of the dependent variable projected from the face-neighbor cell into the original cell. Steady state test cases on structured and unstructured grids are presented, exhibiting the low-dissipative nature of the scheme. Results are primarily compared to those produced by existing limited and unlimited second-order upwind (SOU) and first-order upwind (FOU). Solution accuracy, convergence rate and computational costs are examined.

Cette thèse porte sur une méthode cartésienne parallèle pour résoudre des problèmes elliptiques avec interfaces complexes et sur son application aux problèmes elliptiques en domaine irrégulier dans le cadre d'un modèle de croissance tumorale.La méthode est basée sur un schéma aux différences finies et sa précision est d’ordre deux sur tout le domaine. L'originalité de la méthode consiste en l'utilisation d'inconnues additionnelles situées sur l'interface et qui permettent d’exprimer les conditions de transmission à l'interface. La méthode est décrite et les détails sur la parallélisation, réalisé avec la bibliothèque PETSc, sont donnés. La méthode est validée et les résultats sont comparés avec ceux d'autres méthodes du même type disponibles dans la littérature. Une étude numérique de la méthode parallélisée est fournie.La méthode est appliquée aux problèmes elliptiques dans un domaine irrégulier apparaissant dans un modèle continue et tridimensionnel de croissance tumorale, le modèle à deux espèces du type Darcy . L'approche utilisée dans cette application est basée sur la pénalisation des conditions de transmission à l'interface, afin de imposer des conditions de Neumann homogènes sur le border d'un domaine irrégulier. Les simulations du modèle sont fournies et montrent la capacité de la méthode à imposer une bonne approximation de conditions au bord considérées. / This theses deals with a parallel Cartesian method to solve elliptic problems with complex interfaces and its application to elliptic irregular domain problems in the framework of a tumor growth model.This method is based on a finite differences scheme and is second order accurate in the whole domain. The originality of the method lies in the use of additional unknowns located on the interface, allowing to express the interface transmission conditions. The method is described and the details of its parallelization, performed with the PETSc library, are provided. Numerical validations of the method follow with comparisons to other related methods in literature. A numerical study of the parallelized method is also given.Then, the method is applied to solve elliptic irregular domain problems appearing in a three-dimensional continuous tumor growth model, the two-species Darcy model. The approach used in this application is based on the penalization of the interface transmission conditions, in order to impose homogeneous Neumann boundary conditions on the border of an irregular domain. The simulations of model are provided and they show the ability of the method to impose a good approximation of the considered boundary conditions. / Questa tesi introduce un metodo parallelo su griglia cartesiana per risolvere problemi ellittici con interfacce complesse e la sua applicazione ai problemi ellittici in dominio irregolare presenti in un modello di crescita tumorale.Il metodo è basato su uno schema alle differenze finite ed è accurato al secondo ordine su tutto il dominio di calcolo. L'originalità del metodo consiste nell'introduzione di nuove incognite sull'interfaccia, le quali permettono di esprimere le condizioni di trasmissione sull'interfaccia stessa. Il metodo viene descritto e i dettagli della sua parallelizzazione, realizzata con la libreria PETSc, sono forniti. Il metodo è validato e i risultati sono confrontati con quelli di metodi dello stesso tipo trovati in letteratura. Uno studio numerico del metodo parallelizzato è inoltre prodotto.Il metodo è applicato ai problemi ellittici in dominio irregolare che compaiono in un modello continuo e tridimensionale di crescita tumorale, il modello a due specie di tipo Darcy. L'approccio utilizzato è basato sulla penalizzazione delle condizioni di trasmissione sull'interfaccia, al fine di imporre condizioni di Neumann omogenee sul bordo di un dominio irregolare. Le simulazioni del modello sono presentate e mostrano la capacità del metodo di imporre una buona approssimazione delle condizioni al bordo considerate.