The aim of this project was to develop a numerical model simulating the global transport of air pollution. The model currently used by SMHI, the Swedish Meteorological and Hydrological Institute, was intended to perform such simulations at regional scales. However, the usage of a grid based on longitudes and latitudes makes this model work poorly in global simulations. The problem with this grid is that the nodes are concentrated near the North and South Poles. Due to this characteristic, the Courant, Friedrich and Lewy-condition for numerical stability puts a sharp bound on the timestep length. In order to keep the transport algorithms stable, the timestep length has to be very small. To avoid too short timesteps, we adopted a new kind of grid. In principle, we defined the global grid on the surface of a cube. By replacing straight lines on the cube with great circles we took the curvature of the Earth into account. The result was almost equal distances between the nodes, and thus we were able to use an acceptable timestep length. / Målet med detta examensarbete var att utveckla en datormodell för global transportsimulering av luftföroreningar. Den modell som hittills använts av SMHI, Sveriges meterologiska och hydrologiska institut, utvecklades för att göra transportsimuleringar över begränsade områden. Detta innebar att ett gitter baserat på latituder och longituder kunde utnyttjas. Problemet som uppkommer då detta gitter används i globala simuleringar är att noderna ligger alltför tätt i närheten av polerna. Enligt CFL-villkoret för numerisk stabilitet så kommer då de numeriska algoritmer som simulerar vindtransport att bli instabila i polarområdena om inte tidssteglängden görs mycket liten. För att undvika en alltför kort tidssteglängd så har vi i detta projekt begagnat oss av ett nytt slags gitter. Grundideen bakom det nya gittret var att utgå från en numerisk metod definierad på en kub. För att tillämpa metoden på en sfär så ersatte vi kubens räta linjer med storcirklar till sfären. Resultatet blev en någorlunda jämn geografisk spridning av beräkningsnoderna, vilket innebar att en godtagbar tidssteglängd kunde användas.

We consider the reaction-diffusion-advection (RDA) equations resulting in air pollution mod- eling problems. We employ the finite element method (FEM) for solving the RDA equations in two dimensions. Linear triangular finite elements are used in the discretization of problem domains. The instabilities occuring in the solution when the standard Galerkin finite element method is used, in advection or reaction dominated cases, are eliminated by using an adap- tive stabilized finite element method. In transient problems the unconditionally stable Crank- Nicolson scheme is used for the temporal discretization. The stabilization is also applied for reaction or advection dominant case in the time dependent problems. It is found that the stabilization in FEM makes it possible to solve RDA problems for very small diffusivity constants. However, for transient RDA problems, although the stabilization improves the solution for the case of reaction or advection dominance, it is not that pronounced as in the steady problems. Numerical results are presented in terms of graphics for some test steady and unsteady RDA problems. Solution of an air pollution model problem is also provided.

À l'échelle macroscopique, le transport d'un composant d'un mélange fluide dans un milieu poreux est décrit par des flux advectif, diffusif et dispersif. Ces deux derniers peuvent être formulés en utilisant les tenseurs de diffusion et de dispersion homogénéisés. Nous nous intéressons d'abord au cas monophasique où le milieu est saturé par un liquide incompressible. En utilisant la technique de développement asymptotique à double échelle, le champ de vitesse est d'abord obtenu en résolvant le problème de Stokes à l'échelle microscopique, puis le champ de vitesse est utilisé pour résoudre un problème d'advection-diffusion vectoriel dont la solution permet d'évaluer les tenseurs de diffusion et de dispersion. Nous considérons une approximation par éléments finis des problèmes posés à l'échelle microscopique dont nous effectuons une analyse numérique complète. Dans le cas du problème de Stokes, nous comparons trois types d'éléments finis en fonction de la qualité de la divergence du champ de vitesse discret. A titre d'application, nous calculons les valeurs des tenseurs de diffusion et de dispersion pour des réseaux cubiques et cubiques centrés de sphères et nous étudions l'influence de l'intensité de l'advection et de la morphologie des pores sur les tenseurs. La méthodologie ci-dessus est étendue au cas diphasique où nous considérons le transport de vapeur en équilibre avec des ménisques liquides localisés dans les pores. Nous mettons en évidence l'influence de la concentration de vapeur et de la taille des pores dans l'équilibre liquide-vapeur. Des simulations sur des réseaux cubiques et cubiques centrés de sphères fournissent les coefficients des tenseurs de diffusion et de dispersion. Enfin, un problème de transport macroscopique est résolu afin d'étudier les effets non-linéaires dus au caractère multi-échelles du problème.

Ce travail présente une étude analytique, numérique et expérimentale de l'hydrodynamique d'un cylindre soumis à un choc sinus. L'étude analytique consiste à développer un modèle général permettant de prédire les forces et la répartition de pressions sur le cylindre soumis à un mouvement transitoire quelconque dans un fluide visqueux bidimensionel. La simulation numérique s'attache à la modélisation en maillage mobile d'un cylindre évoluant dans un domaine fluide infini bidimensionel sous le code généraliste STAR-CD. Un dispositif expérimental avec la chaîne de mesure associée est conçu permettant d'imposer différents types de mouvements (choc sinus, mouvement oscillatoire entretenu) à un cylindre baignant dans un bassin. Ces trois approches complémentaires montrent que l'écoulement est régi par deux nombres adimensionnels : le nombre de Stokes, Beta, et le nombre de Keulegan-Carpenter, KC. Beta mesure le rapport entre le temps de diffusion visqueuse et le temps caractéristique du choc et KC le rapport entre le déplacement maximum du corps et son diamètre. Dans le cas où Beta est très grand (fluide parfait), la force sur le cylindre est régie par l'effet inertiel quelque soit KC. A petit KC, la pression est contrôlée uniquement par l'effet inertiel alors que pour des grands KC, un effet d'advection vient s'ajouter. A Beta modéré et pour des petites valeurs de KC, la force et la pression subissent trois effets : un effet d'inertie, d'histoires et de traînée. Pour des grandes valeurs de KC, l'écoulement devient plus riche : des zones dynamiques de vorticité apparaissent. Elles induisent de fortes fluctuations locales de pression sans modifier la force totale. Ce travail se termine par l'étude d'un cas d'interaction fluide-structure où les outils de prédiction et de simulation développés récédemment sont mis en oeuvre. Les phénomènes d'inertie, de trainée et les effets d'histoire apparaissent simultanément et sont couplés avec le mouvement du cylindre.

No description available.

The modelling of the processes of advection and dispersion-diffusion is the most crucial factor in solute transport simulations. It is generally appreciated that the first order upwind difference scheme gives rise to excessive numerical diffusion, whereas the conventional second order central difference scheme exhibits severe oscillations for advection dominated transport, especially in regions of high solute gradients or discontinuities. Higher order schemes have therefore become increasingly used for improved accuracy and for reducing grid scale oscillations. Two such schemes are the QUICK (Quadratic Upwind Interpolation for Convective Kinematics) and TOASOD (Third Order Advection Second Order Diffusion) schemes, which are similar in formulation but different in accuracy, with the two schemes being second and third order accurate in space respectively for finite difference models. These two schemes can be written in various finite difference forms for transient solute transport models, with the different representations having different numerical properties and computational efficiencies. Although these two schemes are advectively (or convectively) stable, it has been shown that the originally proposed explicit QUICK and TOASOD schemes become numerically unstable for the case of pure advection. The stability constraints have been established for each scheme representation based upon the von Neumann stability analysis. All the derived schemes have been tested for various initial solute distributions and for a number of continuous discharge cases, with both constant and time varying velocity fields. The 1-D QUICKEST (QUICK with Estimated Streaming Term) scheme is third order accurate both in time and space. It has been shown analytically and numerically that a previously derived quasi 2-D explicit QUICKEST scheme, with a reduced accuracy in time, is unstable for the case of pure advection. The modified 2-D explicit QUICKEST, ADI-TOASOD and ADI-QUICK schemes have been developed herein and proved to be numerically stable, with the bility sta- region of each derived 2-D scheme having also been established. All these derived 2-D schemesh ave been tested in a 2-D domain for various initial solute distributions with both uniform and rotational flow fields. They were further tested for a number of 2-D continuous discharge cases, with the corresponding exact solutions having also been derived herein. All the numerical tests in both the 1-D and 2-D cases were compared with the corresponding exact solutions and the results obtained using various other difference schemes, with the higher order schemes generally producing more accurate predictions, except for the characteristic based schemes which failed to conserve mass for the 2-D rotational flow tests. The ADI-TOASOD scheme has also been applied to two water quality studies in the U. K., simulating nitrate and faecal coliform distributions respectively, with the results showing a marked improvement in comparison with the results obtained by the second order central difference scheme. Details are also given of a refined numerical representation of flooding and drying of tidal flood plains for hydrodynamic modelling, with the results showing considerable improvements in comparison with a number of existing models and in good agreement with the field measured data in a natural harbour study.

A novel parallel block-based anisotropic adaptive mesh refinement (AMR) technique for multi-block body-fitted grids is proposed and described. Rather than adopting the more usual isotropic approach to mesh refinement, an anisotropic refinement procedure is proposed which allows refinement of grid blocks in each coordinate direction in an independent fashion. This allows for more efficient and accurate treatment of narrow layers and/or discontinuities which occur, for example, in the boundary and mixing layers of viscous flows, and in regions of strong non-linear wave interactions with shocks. The anisotropic AMR technique is implemented within an existing finite-volume framework, which encompasses both explicit and implicit solution methods, and is capable of performing calculations with second- and higher-order spatial accuracy. To clearly demonstrate the feasibility of the proposed technique, it is applied to the unsteady and steady-state solutions of both the advection diffusion equation, as well as the Euler equations, in two space dimensions.

A novel parallel block-based anisotropic adaptive mesh refinement (AMR) technique for multi-block body-fitted grids is proposed and described. Rather than adopting the more usual isotropic approach to mesh refinement, an anisotropic refinement procedure is proposed which allows refinement of grid blocks in each coordinate direction in an independent fashion. This allows for more efficient and accurate treatment of narrow layers and/or discontinuities which occur, for example, in the boundary and mixing layers of viscous flows, and in regions of strong non-linear wave interactions with shocks. The anisotropic AMR technique is implemented within an existing finite-volume framework, which encompasses both explicit and implicit solution methods, and is capable of performing calculations with second- and higher-order spatial accuracy. To clearly demonstrate the feasibility of the proposed technique, it is applied to the unsteady and steady-state solutions of both the advection diffusion equation, as well as the Euler equations, in two space dimensions.

The research presented in this thesis was motivated by the desire to understand the flow field within a new digital microfluidic device currently under development. This required an investigation of the dynamics of a droplet migrating along the surface of another fluid due to interfacial surface tension gradients. The quantitative analysis of the flow field presented in this thesis provides the first known solution for the velocity field in a migrating droplet confined to an interface. The first step towards gaining insight into the flow field was accomplished by using the method of reflections to obtain an analytical model for a submerged droplet migrating near a free surface. The submerged droplet model enabled the analysis of the velocity field and droplet migration speed and their dependence on the fluid properties. In general, the migration velocity of a submerged droplet was found to differ dramatically from the classic problem of thermocapillary migration in an unbounded substrate. A boundary-collocation scheme was developed to determine the flow field and migration velocity of a droplet floating trapped at the air-substrate interface. The numerical method was found to produce accurate solutions for the velocity and temperature fields for nearly all parameters. This numerical scheme was used to judge the accuracy of the flow field obtained by the submerged droplet model. In particular, the model was tested using parameter values taken from a digital microfluidic device. It was determined that the submerged droplet model captured most of the flow structure within the microfluidic droplet. However, for a slightly different choice of parameters, agreement between the two methods was lost. In this case, the numerical scheme was used to uncover novel flow structures.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Uma análise utilizando a série de Taylor é apresentada para se estimar a priori os erros envolvidos na solução numérica da equação de advecção unidimensional com termo fonte, através do Método dos Volumes Finitos em uma malha do tipo uniforme e uma malha não uniforme. Também faz-se um estudo a posteriori para verificar a magnitude do erro de discretização e corroborar os resultados obtidos através da análise a priori. Por meio da técnica de solução manufaturada tem-se uma solução analítica para o problema, a qual facilita a análise dos resultados numéricos encontrados, e estuda-se ainda a influência das funções de interpolação UDS e CDS e do parâmetro u na solução numérica. / An analysis based on Taylor series is presented for estimating a priori the errors involved in the numerical solution of advection equation one-dimensional with source term, using the Finite Volume Method in a mesh uniform and a nonuniform mesh. Also is accomplished a study to determine the magnitude of discretization error and corroborate the results obtained on analyzing a priori. By using the technique of solution manufactured is produced an analytical solution for the problem, which facilitates analysis of the numeric results, and was also studied the influence functions of interpolation UDS and CDS and of parameter u in the numerical solution.