In the thesis work, a code to solve the two dimensional compressible Euler equations for external flows around arbitrary geometries have been developed. A Cartesianmesh generator is incorporated to the solver. Hence the pre-processing can be performed together with the solution within a single code. The code is written in the C++ programming language and its object oriented capabilities have been exploited to save memory in the data structure developed. The Cartesian mesh is formed by dividing squares successively into its four quadrants. The main advantage of using this type of a mesh is the ability to generate meshes around geometries of arbitrary complexity quickly and to adapt the mesh easily based on the solution. The main disadvantage of this method is that the treatment of the cells that are cut by the geometry. For the solution procedure Roe&rsquo / s method as well as flux vector splitting methods are used for the flux evaluation. The flux vector splitting schemes used are van Leer, AUSM, AUSMD and AUSMV methods. Time discretization is performed using a multi-stage method. To increase the accuracy least squares reconstruction is employed. The code is validated by performing calculations around a NACA0012 airfoil profile. The effect of reconstruction is demonstrated by plotting the pressure coefficient on the airfoil. The distribution obtained using reconstruction is very close to the experimental one while there is a considerable deviation for the case without reconstruction. Also the shock capturing capabilities of different methods have been investigated. In addition the performance of each method is analyzed for flow around an NLR 7301 airfoil with a flap.

A fully automated Cartesian/Quad grid generator and laminar flow solver have been developed for external flows by using C++. After defining the input geometry by nodal points, adaptively refined Cartesian grids are generated automatically. Quadtree data structure is used in order to connect the Cartesian cells to each other. In order to simulate viscous flows, body-fitted quad cells can be generated optionally. Connectivity is provided by cut and split cells such that the intersection points of Cartesian cells are used as the corners of quads at the outmost row. Geometry based adaptation methods for cut, split cells and highly curved regions are applied to the uniform mesh generated around the geometry. After obtaining a sufficient resolution in the domain, the solution is achieved with cellcentered approach by using multistage time stepping scheme. Solution based grid adaptations are carried out during the execution of the program in order to refine the regions with high gradients and obtain sufficient resolution in these regions. Moreover, multigrid technique is implemented to accelerate the convergence time significantly. Some tests are performed in order to verify and validate the accuracy and efficiency of the code for inviscid and laminar flows.

L’objectif principal de cette thèse concerne l’étude, la conception et la mise en œuvre numérique de schémas volumes finis associés aux solveurs de type Godunov. On s’intéresse à des systèmes hyperboliques de lois de conservation non linéaires, avec une attention particulière sur les équations d’Euler et les équations MHD idéale. Tout d’abord, nous dérivons un solveur de Riemann simple et véritablement multidimensionnelle, pouvant s’appliquer à tout système de lois de conservation. Ce solveur peut être considéré comme une généralisation 2D de l’approche HLL. Les ingrédients de base de la dérivation sont : la consistance avec la formulation intégrale et une utilisation adéquate des relations de Rankine-Hugoniot. Au final nous obtenons des expressions assez simples et applicables dans les contextes des maillages structurés et non structurés. Dans un second temps, nous nous intéressons à la préservation, au niveau discret, de la contrainte de divergence nulle du champ magnétique pour les équations de la MHD idéale. Deux stratégies sont évaluées et nous montrons comment le solveur de Riemann multidimensionnelle peut être utilisé pour obtenir des simulations robustes à divergence numérique nulle. Deux autres points sont abordés dans cette thèse : la méthode de relaxation pour un système Euler-Poisson pour des écoulements gravitationnels en astrophysique, la formulation volumes finis en coordonnées curvilignes. Tout au long de la thèse, les choix numériques sont validés à travers de nombreux résultats numériques. / The main objective of this thesis concerns the study, design and numerical implementation of finite volume schemes based on the so-Called Godunov-Type solvers for hyperbolic systems of nonlinear conservation laws, with special attention given to the Euler equations and ideal MHD equations. First, we derive a simple and genuinely two-Dimensional Riemann solver for general conservation laws that can be regarded as an actual 2D generalization of the HLL approach, relying heavily on the consistency with the integral formulation and on the proper use of Rankine-Hugoniot relations to yield expressions that are simple enough to be applied in the structured and unstructured contexts. Then, a comparison between two methods aiming to numerically maintain the divergence constraint of the magnetic field for the ideal MHD equations is performed and we show how the 2D Riemann solver can be employed to obtain robust divergence-Free simulations. Next, we derive a relaxation scheme that incorporates gravity source terms derived from a potential into the hydrodynamic equations, an important problem in astrophysics, and finally, we review the design of finite volume approximations in curvilinear coordinates, providing a fresher view on an alternative discretization approach. Throughout this thesis, numerous numerical results are shown.