This is a book on numerical methods for singular perturbation problems - in particular stationary convection-dominated convection-diffusion problems. More precisely it is devoted to the construction and analysis of layer-adapted meshes underlying these numerical methods. An early important contribution towards the optimization of numerical methods by means of special meshes was made by N.S. Bakhvalov in 1969. His paper spawned a lively discussion in the literature with a number of further meshes being proposed and applied to various singular perturbation problems. However, in the mid 1980s this development stalled, but was enlivend again by G.I. Shishkin's proposal of piecewise- equidistant meshes in the early 1990s. Because of their very simple structure they are often much easier to analyse than other meshes, although they give numerical approximations that are inferior to solutions on competing meshes. Shishkin meshes for numerous problems and numerical methods have been studied since and they are still very much in vogue. With this contribution we try to counter this development and lay the emphasis on more general meshes that - apart from performing better than piecewise-uniform meshes - provide a much deeper insight in the course of their analysis. In this monograph a classification and a survey are given of layer-adapted meshes for convection-diffusion problems. It tries to give a comprehensive review of state-of-the art techniques used in the convergence analysis for various numerical methods: finite differences, finite elements and finite volumes. While for finite difference schemes applied to one-dimensional problems a rather complete convergence theory for arbitrary meshes is developed, the theory is more fragmentary for other methods and problems and still requires the restriction to certain classes of meshes.

An obvious reason for studying unsteady flows is the prediction of the effect of unsteady aerodynamic forces on a flight vehicle, since these effects tend to increase the likelihood of aeroelastic instabilities. This is a major concern in aerodynamic design of aircraft that operate in transonic regime, where the flows are characterised by the presence of adjacent regions of subsonic and supersonic flow, usually accompanied by weak shocks. It has been a common expectation that the numerical approach as an alternative to wind tunnel experiments would become more economical as computers became less expensive and more powerful. However even with all the expected future advances in computer technology, the cost of a numerical flutter analysis (computational aeroelasticity) for a transonic flight remains prohibitively high. Hence it is vitally important to develop an efficient, cheaper (in the sense of computational cost) and physically accurate flutter simulation tech nique which is capable of reproducing the data, which would otherwise be obtained from wind tunnel tests, at least to some acceptable engineering accuracy, and that it is essentially appropriate for industrial applications. This need motivated the present research work on exploring and developing efficient and physically accurate computational techniques for steady, unsteady and time-linearised calculations of transonic flows over an aircraft wing with moving shocks. This dissertation is subdivided into eight chapters, seven appendices and a bibliography listing all the reference materials used in the research work. The research work initially starts with a literature survey in unsteady transonic flow theory and calculations, in which emphasis is placed upon the developments in these areas in the last three decades. Chapter 3 presents the small disturbance theory for potential flows in the subsonic, transonic and supersonic regimes, including the required boundary conditions and shock jump conditions. The flow is assumed irrotational and inviscid, so that the equation of state, continuity equation and Bernoulli's equation formulated in Appendices A and B can be employed to formulate the governing fluid equation in terms of total velocity potential. Furthermore for transonic flow with free-stream Mach number close to unity, we show in Appendix C that the shocks that appear are weak enough to allow us to neglect the flow rotationality. The formulations are based on the main assumption that aerofoil slopes are everywhere small, and the flow quantities are small perturbations about their free-stream values. In Chapter 4, we developed an improved approximate factorisation algorithm that solves the two-dimensional steady subsonic small disturbance equation with nonreflecting far-field boundary conditions. The finite difference formulation for the improved algorithm is presented in Appendix D, with the description of the solver used for solving the system of difference equations described in Appendix E. The calculation of steady and unsteady nonlinear transonic flows over a realistic aerofoil are considered in Chapter 5. Numerical solution methods, based on the finite difference approach, for solving the two-dimensional steady and unsteady, general-frequency transonic small disturbance equations are presented, with the corresponding finite difference formulation described in Appendix F. The theories and solution methods for the time-linearised calculations, in the frequency and time domains, for the problem of unsteady transonic flow over a thin planar wing undergoing harmonic oscillation are presented in Chapters 6 and 7, respectively. The time-linearised calculations include the periodic shock motion via the shock jump correction procedure. This procedure corrects the solution values behind the shock, to accommodate the effect of shock motion, and consequently, the solution method will produce a more accurate time-linearised solution for supercritical flow. Appendix G presents the finite difference formulation of these time-linearised solution methods. The aim is to develop an efficient computational method for calculating oscillatory transonic aerodynamic quantities efficiently for use in flutter analyses of both two- and three-dimensional wings with lifting surfaces. Chapter 8 closes the dissertation with concluding remarks and future prospects on the current research work.

Dans ce mémoire, nous présentons une stratégie basée sur une approche hybridedans le domaine temporel, couplant une méthode de résolution des équations de Maxwelldans le domaine 3D (FDTD) avec une méthode de résolution des équations de ligne detransmission, afin de pouvoir simuler des problèmes électromagnétiques de grande échelle. Lemémoire donne les éléments d’hybridation pour deux cadres d’utilisation de cette approche :une approche multi-domaine et une approche multi-résolution ou d’échelle.L’approche multi-domaine est une extension de la méthode FDTD 3D à plusieurs sousdomainesreliés par des structures filaires sur lesquelles on résout une équation de lignes detransmission par un formalisme FDTD 1D. La difficulté est d’abord d’avoir une définitionimplicite du champ électromagnétique dans la théorie des lignes de transmission, et d’autrepart de prendre en compte les effets du sol sur les courants induits au niveau des lignes etsur les champs électromagnétiques.L’approche multi-résolution ou d’échelle est conçue pour étendre les capacités de la méthodeFDTD au traitement du routage de câbles complexes ayant une section plus petite quela taille de la cellule. Ce mémoire présente différentes techniques pour évaluer les paramètresde la ligne, basées sur la résolution d’un problème de Laplace 2D, ainsi qu’une méthode decouplage champs/câbles basée sur le courant de mode commun.L’ensemble de ce travail nous a permis de proposer une méthode numérique efficace pourcalculer les effets électromagnétiques induits par une source (type onde plane ou dipolaire)sur des sites de grande dimension, composés de plusieurs bâtiments reliés entre eux par unréseau de câbles. Dans ce cadre une application à la foudre a été réalisée. / In this thesis, we present a strategy based on a hybrid approach in the timedomain, by coupling 3D method (FDTD) with a multi-conductors transmission line (MTL)method, in order to simulate complex large scale electromagnetic problems. This reportgives the theoretical and numerical elements for coupling these approaches for two kindof problems, which are the multi domains approach and the multi scale approach. Themultiple domains approach is an extension of the classical FDTD method taking into accountseveral 3D subdomains, interconnected by a wire network, on which a 1D transmission lineformalism is used. The main issues are, on one hand to have an implicit expression ofthe electromagnetic field in the transmission line approach, and on the other hand to beable to take into account the ground effects on the induced currents, on the transmissionline parameters and on the electromagnetic field. The multi scale approach is developed toextend the capabilities of FDTD to deal with complex cables routing. We assume that thecross section of the cables are smallest than the cell size, and in these problems, the 1Dtransmission line problem is physically included in the 3D global computational domain.The work done in this thesis leaded to a new field to transmission line coupling based onthe common mode current, and an evaluation of the transmission. line parameters basedon a Laplace equation resolution in 2D. In this work, we have elaborated and proposedefficient numerical strategies for the computation of electromagnetic induced effects on largeand complex sites, composed of several interconnected distant buildings. An application tolightning problems have been done.

Submitted by JÚLIO HEBER SILVA (julioheber@yahoo.com.br) on 2016-11-10T11:15:34Z No. of bitstreams: 2 Dissertação - Hugo Naves de Jesus - 2016.pdf: 1851851 bytes, checksum: 71cb8f26f4f38eb5f89d99aafc926b66 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Approved for entry into archive by Jaqueline Silva (jtas29@gmail.com) on 2016-11-10T17:47:53Z (GMT) No. of bitstreams: 2 Dissertação - Hugo Naves de Jesus - 2016.pdf: 1851851 bytes, checksum: 71cb8f26f4f38eb5f89d99aafc926b66 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Made available in DSpace on 2016-11-10T17:47:53Z (GMT). No. of bitstreams: 2 Dissertação - Hugo Naves de Jesus - 2016.pdf: 1851851 bytes, checksum: 71cb8f26f4f38eb5f89d99aafc926b66 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2016-09-16 / Conselho Nacional de Pesquisa e Desenvolvimento Científico e Tecnológico - CNPq / Finite difference schemes belong to a class of numerical methods used to approximate derivatives. They are widely used to find approximations to differential equations. There are a lot of numerical methods, whose deductions are made through expansions in Taylor Series. Depending on the manner in which expansion is made, it can be combined with other expansions to obtain derivatives with better numerical approximations. Usually when we get numerical derivative with better approaches, it is necessary to increase the amount of points used in the grid. An alternative to this problem are compact methods, which achieve better approximations for the same derivative but without increasing the number of mesh points. This work is an attempt to develop the Compact-SSFD method for the Schrödinger Equation Nonlinear Fourth Order. SSFD methods are used to separate the parts of a differential equation so that each part can be solved separately. For example in the case of non-linear differential equations it is often used to separate the linear parts of nonlinear parts. In Compact-SSFD methods nonlinear parts are resolved exactly as the linear are resolved using compact methods. Our work is inspired in the Dehghan and Taleei’s work where was used the Compact-SSFD method for solving numerically the equation Nonlinear Schrödinger. Before we try to develop our method, the results of the authors was correctly reproduced. But when we try to deduce a method analogous to the differential equation we wanted to solve, which also involves derived from fourth order, we realized that a Compact type method does not get as trivially as in the case of used to approach second-order derivatives. / Métodos de diferenças finitas pertencem a uma classe de métodos numéricos usados para se aproximar derivadas. Eles são amplamente usados para encontrar-se soluções numéricas para equações diferenciais. Há uma grande quantidade de métodos numéricos, cuja as deduções são feitas através de expansões em séries de Taylor. Dependendo da forma em que uma expansão é feita, ela pode ser combinada com outras expansões para obter-se derivadas numéricas com melhores aproximações. Geralmente quando obtemos derivadas numéricas com aproximações melhores, é necessário aumentar-se a quantidade de pontos usados no domínio discretizado. Uma alternativa a este problema são os chamados métodos compact, que obtêm melhores aproximações para a mesma derivada mas sem precisar aumentar a quantidade de pontos da malha. Este trabalho é uma tentativa de desenvolver-se um método Compact-SSFD para a Equação de Schrödinger Não Linear de Quarta Ordem. Métodos SSFD são usados para separar-se as partes de uma equação diferencial tal que cada parte possa ser resolvida separadamente. Por exemplo no caso de equações diferenciais não lineares ele é bastante usado para separar-se as partes lineares das partes não lineares. Nos métodos Compact-SSFD as partes não lineares são resolvidas exatamente enquanto as lineares são resolvidas usando-se métodos compact. Nos baseamos no trabalho de Dehghan e Taleei onde foi usado o Método Compact-SSFD para resolver-se numericamente a Equação de Schrödinger Não Linear. Antes de tentarmos desenvolver nosso método, reproduzimos corretamente os resultados dos autores. Mas ao tentarmos deduzir um método análogo para a equação diferencial que queríamos resolver, que envolve também derivadas de quarta ordem, percebemos que um método do tipo Compact não se obtêm tão trivialmente como no caso dos usados para aproximar-se derivadas de segunda ordem.

Cette thèse est consacrée à l'étude de quelques problèmes mathématiques issus du micromagnétisme. Le but est d'analyser le comportement des modèles en fonction de différents paramètres physiques, dont les fines variations sont parfois difficilement mesurables. Nous adoptons des approches numériques, asymptotiques ou d'homogénéisation. Les modèles considérés reposent sur l'utilisation de l'équation de Landau-Lifshitz-Gilbert (LLG) décrivant l'évolution du champ d'aimantation dans un matériau ferromagnétique. Nous rappelons d'abord quelques notions importantes en ferromagnétisme. Ensuite, nous menons une étude numérique d'un modèle de la dynamique d'aimantation avec effets d'inertie. Nous proposons un schéma aux différences finies semi-implicite qui respecte de façon intrinsèque les propriétés du modèle continu. Des simulations numériques sont réalisées pour cerner l'effet du paramètre d'inertie. Ces simulations montrent aussi la performance du schéma et confirment l'ordre de convergence obtenu théoriquement. Nous étudions ensuite un modèle de la dynamique de l'aimantation avec amortissement non local. La sensibilité de la dynamique d'aimantation au paramètre d'amortissement est étudiée en donnant le problème limite pour de petites et de grandes valeurs du paramètre. Enfin, nous étudions l'homogénéisation de l'équation LLG dans deux types de matériau, à savoir les composites présentant un fort contraste des propriétés magnétiques et les matériaux périodiquement perforés avec énergie d'anisotropie de surface. Des modèles homogénéisés sont d'abord obtenus formellement puis une dérivation rigoureuse est établie en se basant principalement sur les concepts de la convergence à double échelle et de la convergence à double échelle en surface. Pour traiter les non-linéarités, nous introduisons une nouvelle méthode basée sur le couplage d'un opérateur de dilatation calibré sur les contrastes d'échelle et d'un outil de réduction de dimension, par construction de grilles emboitées adaptées à la géométrie du domaine microscopique. / This thesis is devoted to the study of some mathematical problems arising in micromagnetism. The models considered here are based on the Landau-Lifshitz-Gilbert equation (LLG) describing the evolution of the magnetization field in a ferromagnetic material. Our aim is the analysis of the behavior of the models regarding the slight variations of some physical parameters. We first recall some important notions about ferromagnetism. Then, we carry out a numerical study of a model of magnetization dynamics with inertial effects. We propose a semi-implicit finite difference scheme which intrinsically respects the properties of the continuous model. Numerical simulations are provided for emphasizing the effect of the inertia parameter. These simulations also show the performance of the scheme and confirm the order of convergence obtained theoretically. We then study a model of magnetization dynamics with a non-local damping. The sensitivity of the magnetization dynamics to the damping coefficient is studied by giving the limiting problem for small and large values of the parameter. Finally, we study the homogenization of the LLG equation in two types of structures, namely a composite material with strongly contrasted magnetic properties, and a periodically perforated material with surface anisotropy energy. The homogenized models are first obtained formally. The rigorous derivation is then performed using mainly the concepts of two-scale convergence, two-scale convergence on surfaces together with a new homogenization procedure for handling with the nonlinear terms. More precisely, an appropriate dilation operator is applied in a embedded cells network, the network being constrained by the microscopic geometry.

This is a book on numerical methods for singular perturbation problems - in particular stationary convection-dominated convection-diffusion problems. More precisely it is devoted to the construction and analysis of layer-adapted meshes underlying these numerical methods. An early important contribution towards the optimization of numerical methods by means of special meshes was made by N.S. Bakhvalov in 1969. His paper spawned a lively discussion in the literature with a number of further meshes being proposed and applied to various singular perturbation problems. However, in the mid 1980s this development stalled, but was enlivend again by G.I. Shishkin's proposal of piecewise- equidistant meshes in the early 1990s. Because of their very simple structure they are often much easier to analyse than other meshes, although they give numerical approximations that are inferior to solutions on competing meshes. Shishkin meshes for numerous problems and numerical methods have been studied since and they are still very much in vogue. With this contribution we try to counter this development and lay the emphasis on more general meshes that - apart from performing better than piecewise-uniform meshes - provide a much deeper insight in the course of their analysis. In this monograph a classification and a survey are given of layer-adapted meshes for convection-diffusion problems. It tries to give a comprehensive review of state-of-the art techniques used in the convergence analysis for various numerical methods: finite differences, finite elements and finite volumes. While for finite difference schemes applied to one-dimensional problems a rather complete convergence theory for arbitrary meshes is developed, the theory is more fragmentary for other methods and problems and still requires the restriction to certain classes of meshes.

Esta tesis está centrada en el modelado de la acústica de salas en espacios cerrados mediante el uso de una ecuación de transferencia radiativa y una ecuación de difusión En este trabajo se investiga cómo a través de estos modelos teóricos se pueden simular el campo sonoro en espacios complejos. Recientemente, el modelo de la ecuación de fusión ha sido prppuesto para ser utilizado en el modelado de la acústica de salas con superficies que reflejan el sonido de forma totalmente difusa. Este enfoque del uso de la ecuación de la disusión de sido intensamente investigado en los últimos años, ya que proporciona una alta eficiencia y flexibilidad para simular las distribuciones del campo sonoro en diferentes tipos de salas; sin embargo, sólo se han realizado unas pocas investigaciones con el objetivo de indagar sobre la precisión y las limitaciones de este método alternativo. Por lo tanto, en primer lugar se presenta un modelo basado en la ecuación de transferencia por radiación siendo meta principal el unificar una amplia gama de métodos geométricos de modelado de acústica de salas. Además, esta tesis está especialmente dedicada a establecer las bases y suposiciones que permitan obtener un modelo de difusión acústica como particularización del modelo de transferencia radiativa con el objetivo de conseguir una descripción clara y adecuada de sus ventajas y limitaciones desde el punto de vista teórico. Este trabajo permite enlazar directamente al modelo de la ecuación de difusión con el grupo de métodos de la acústica geométrica reforzando sus características y permitiendo una adecuada comparación con estos métodos ampliamente reconocidos. Una vez realizado este análisis teórico, esta tesis también se dedica a cuestiones relativas a la implementación numérica del modelo acústico de la ecuación de difusión . En este trabajo, se modela el campo sonoro a través de esquemas en diferencias finitas. Los resultados de este estudio proporcionan soluciones simples y practicas que muestran unos requerimientos computacionales bajos tanto de consumo de memoria como de tiempo. / Navarro Ruiz, JM. (2012). Discrete-time modelling of diffusion processes for room acoustics simulation and analysis [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/14861

This thesis is concerned with the investigation of numerical methods for the solution of the Hamilton-Jacobi-Bellman (HJB) equations arising in European and American option pricing with proportional transaction costs. We first consider the problem of computing reservation purchase and write prices of a European option in the model proposed by Davis, Panas and Zariphopoulou [19]. It has been shown [19] that computing the reservation purchase and write prices of a European option involves solving three different fully nonlinear HJB equations. In this thesis, we propose a penalty approach combined with a finite difference scheme to solve the HJB equations. We first approximate each of the HJB equations by a quasi-linear second order partial differential equation containing two linear penalty terms with penalty parameters. We then develop a numerical scheme based on the finite differencing in both space and time for solving the penalized equation. We prove that there exists a unique viscosity solution to the penalized equation and the viscosity solution to the penalized equation converges to that of the original HJB equation as the penalty parameters tend to infinity. We also prove that the solution of the finite difference scheme converges to the viscosity solution of the penalized equation. Numerical results are given to demonstrate the effectiveness of the proposed method. We extend the penalty approach combined with a finite difference scheme to the HJB equations in the American option pricing model proposed by Davis and Zarphopoulou [20]. Numerical experiments are presented to illustrate the theoretical findings.

Fundação Carlos Chagas Filho de Amparo a Pesquisa do Estado do Rio de Janeiro / Considera-se neste trabalho um modelo matemático para escoamentos com duas e três fases em reservatórios petrolíferos e a modelagem computacional do sistema de equações governantes para a sua solução numérica. Os fluidos são imiscíveis e incompressíveis e as heterogeneidades da rocha reservatório são modeladas estocasticamente. Além disso, é modelado o fenômeno de histerese para a fase óleo via funções de permeabilidades relativas. No caso de escoamentos trifásicos água-óleo-gás a escolha de expressões gerais para as funções de permeabilidades relativas pode levar à perda de hiperbolicidade estrita e, desta maneira, à existência de uma região elíptica ou de pontos umbílicos para o sistema não linear de leis de conservação hiperbólicas que descreve o transporte convectivo das fases fluidas. Como conseqüência, a perda de hiperbolicidade estrita pode levar à existência de choques não clássicos (também chamados de choques transicionais ou choques subcompressivos) nas soluções de escoamentos trifásicos, de difícil simulação numérica. Indica-se um método numérico com passo de tempo fracionário, baseado em uma técnica de decomposição de operadores, para a solução numérica do sistema governante de equações diferenciais parciais que modela o escoamento bifásico água-óleo imiscível em reservatórios de petróleo heterogêneos. Um simulador numérico bifásico água-óleo eficiente desenvolvido pelo grupo de pesquisa no qual o autor está inserido foi modificado com sucesso para incorporar a histerese sob as hipóteses consideradas. Os resultados numéricos obtidos para este caso indicam fortes evidências que o método proposto pode ser estendido para o caso trifásico água-óleo-gás. A técnica de decomposição de operadores em dois níveis permite o uso de passos de tempo distintos para os quatro problemas definidos pelo procedimento de decomposição: convecção, difusão, pressão-velocidade e relaxação para histerese. O problema de transporte convectivo (hiperbólico) das fases fluidas é aproximado por um esquema central de diferenças finitas explícito, conservativo, não oscilatório e de segunda ordem. Este esquema é combinado com elementos finitos mistos, localmente conservativos, para a aproximação dos problemas de transporte difusivo (parabólico) e de pressão-velocidade (elíptico). O operador temporal associado ao problema parabólico de difusão é resolvido fazendo-se uso de uma estratégia implícita de solução (Backward Euler). Uma equação diferencial ordinária é resolvida (analiticamente) para a relaxação relacionada à histerese. Resultados numéricos para o problema bifásico água-óleo em uma dimensão espacial em concordância com resultados semi-analíticos disponíveis na literatura foram reproduzidos e novos resultados em meios heterogêneos, em duas dimensões espaciais, são apresentados e a extensão desta técnica para o caso de problemas trifásicos água-óleo-gás é proposta. / We consider in this work a mathematical model for two- and three-phase flow problems in petroleum reservoirs and the computational modeling of the governing equations for its numerical solution. We consider two- (water-oil) and three-phase (water-gas-oil) incompressible, immiscible flow problems and the reservoir rock is considered to be heterogeneous. In our model, we also take into account the hysteresis effects in the oil relative permeability functions. In the case of three-phase flow, the choice of general expressions for the relative permeability functions may lead to the loss of strict hyperbolicity and, therefore, to the existence of an elliptic region or umbilic points for the system of nonlinear hyperbolic conservation laws describing the convective transport of the fluid phases. As a consequence, the loss of hyperbolicity may lead to the existence of nonclassical shocks (also called transitional shocks or undercompressive shocks) in three-phase flow solutions. We present a new, accurate fractional time-step method based on an operator splitting technique for the numerical solution of a system of partial differential equations modeling two-phase, immiscible water-oil flow problems in heterogeneous petroleum reservoirs. An efficient two-phase water-oil numerical simulator developed by our research group was sucessfuly extended to take into account hysteresis effects under the hypotesis previously annouced. The numerical results obtained by the procedure proposed indicate numerical evidence the method at hand can be extended for the case of related three-phase water-gas-oil flow problems. A two-level operator splitting technique allows for the use of distinct time steps for the four problems defined by the splitting procedure: convection, diffusion, pressure-velocity and relaxation for hysteresis. The convective transport (hyperbolic) of the fluid phases is approximated by a high resolution, nonoscillatory, second-order, conservative central difference scheme in the convection step. This scheme is combined with locally conservative mixed finite elements for the numerical solution of the diffusive transport (parabolic) and the pressure-velocity (elliptic) problems. The time discretization of the parabolic problem is performed by means of the implicit Backward Euler method. An ordinary diferential equation is solved (analytically) for the relaxation related to hysteresis. Two-phase water-oil numerical results in one space dimensional, in which are in a very good agreement with semi-analitycal results available in the literature, were computationaly reproduced and new numerical results in two dimensional heterogeneous media are also presented and the extension of this technique to the case of three-phase water-oil-gas flows problems is proposed.

Fundação Carlos Chagas Filho de Amparo a Pesquisa do Estado do Rio de Janeiro / Considera-se neste trabalho um modelo matemático para escoamentos com duas e três fases em reservatórios petrolíferos e a modelagem computacional do sistema de equações governantes para a sua solução numérica. Os fluidos são imiscíveis e incompressíveis e as heterogeneidades da rocha reservatório são modeladas estocasticamente. Além disso, é modelado o fenômeno de histerese para a fase óleo via funções de permeabilidades relativas. No caso de escoamentos trifásicos água-óleo-gás a escolha de expressões gerais para as funções de permeabilidades relativas pode levar à perda de hiperbolicidade estrita e, desta maneira, à existência de uma região elíptica ou de pontos umbílicos para o sistema não linear de leis de conservação hiperbólicas que descreve o transporte convectivo das fases fluidas. Como conseqüência, a perda de hiperbolicidade estrita pode levar à existência de choques não clássicos (também chamados de choques transicionais ou choques subcompressivos) nas soluções de escoamentos trifásicos, de difícil simulação numérica. Indica-se um método numérico com passo de tempo fracionário, baseado em uma técnica de decomposição de operadores, para a solução numérica do sistema governante de equações diferenciais parciais que modela o escoamento bifásico água-óleo imiscível em reservatórios de petróleo heterogêneos. Um simulador numérico bifásico água-óleo eficiente desenvolvido pelo grupo de pesquisa no qual o autor está inserido foi modificado com sucesso para incorporar a histerese sob as hipóteses consideradas. Os resultados numéricos obtidos para este caso indicam fortes evidências que o método proposto pode ser estendido para o caso trifásico água-óleo-gás. A técnica de decomposição de operadores em dois níveis permite o uso de passos de tempo distintos para os quatro problemas definidos pelo procedimento de decomposição: convecção, difusão, pressão-velocidade e relaxação para histerese. O problema de transporte convectivo (hiperbólico) das fases fluidas é aproximado por um esquema central de diferenças finitas explícito, conservativo, não oscilatório e de segunda ordem. Este esquema é combinado com elementos finitos mistos, localmente conservativos, para a aproximação dos problemas de transporte difusivo (parabólico) e de pressão-velocidade (elíptico). O operador temporal associado ao problema parabólico de difusão é resolvido fazendo-se uso de uma estratégia implícita de solução (Backward Euler). Uma equação diferencial ordinária é resolvida (analiticamente) para a relaxação relacionada à histerese. Resultados numéricos para o problema bifásico água-óleo em uma dimensão espacial em concordância com resultados semi-analíticos disponíveis na literatura foram reproduzidos e novos resultados em meios heterogêneos, em duas dimensões espaciais, são apresentados e a extensão desta técnica para o caso de problemas trifásicos água-óleo-gás é proposta. / We consider in this work a mathematical model for two- and three-phase flow problems in petroleum reservoirs and the computational modeling of the governing equations for its numerical solution. We consider two- (water-oil) and three-phase (water-gas-oil) incompressible, immiscible flow problems and the reservoir rock is considered to be heterogeneous. In our model, we also take into account the hysteresis effects in the oil relative permeability functions. In the case of three-phase flow, the choice of general expressions for the relative permeability functions may lead to the loss of strict hyperbolicity and, therefore, to the existence of an elliptic region or umbilic points for the system of nonlinear hyperbolic conservation laws describing the convective transport of the fluid phases. As a consequence, the loss of hyperbolicity may lead to the existence of nonclassical shocks (also called transitional shocks or undercompressive shocks) in three-phase flow solutions. We present a new, accurate fractional time-step method based on an operator splitting technique for the numerical solution of a system of partial differential equations modeling two-phase, immiscible water-oil flow problems in heterogeneous petroleum reservoirs. An efficient two-phase water-oil numerical simulator developed by our research group was sucessfuly extended to take into account hysteresis effects under the hypotesis previously annouced. The numerical results obtained by the procedure proposed indicate numerical evidence the method at hand can be extended for the case of related three-phase water-gas-oil flow problems. A two-level operator splitting technique allows for the use of distinct time steps for the four problems defined by the splitting procedure: convection, diffusion, pressure-velocity and relaxation for hysteresis. The convective transport (hyperbolic) of the fluid phases is approximated by a high resolution, nonoscillatory, second-order, conservative central difference scheme in the convection step. This scheme is combined with locally conservative mixed finite elements for the numerical solution of the diffusive transport (parabolic) and the pressure-velocity (elliptic) problems. The time discretization of the parabolic problem is performed by means of the implicit Backward Euler method. An ordinary diferential equation is solved (analytically) for the relaxation related to hysteresis. Two-phase water-oil numerical results in one space dimensional, in which are in a very good agreement with semi-analitycal results available in the literature, were computationaly reproduced and new numerical results in two dimensional heterogeneous media are also presented and the extension of this technique to the case of three-phase water-oil-gas flows problems is proposed.