Cette thèse traite avec des équations aux dérivées partielles provenant de la physique mathématique. En particulier, à partir de modèles 3D ferromagnétisme et ferroélectricité, nous obtenons des modèles 1D et 2D par l'intermédiaire de processus asymptotiques basés sur des méthodes de réduction de dimension. Le modèle 3D ferromagnétisme a été proposé par W.F. Brown depuis lesannées 40. Il est également possible d'utiliser un modèle dynamique, décrivant l'aimantation au cours du temps, en utilisant un système introduit par L.D. Landau et E.M. Lifschitz en 1935. Pour le modèle ferroélectrique, nous nous référons aux papiers de P. Chandra et P.B. Littlewood, W. Zhang et K. Bhattacharya et au livre de T. Mitsui, I. Taksuzaki et E. Nakamura.Ma thèse est constituée de trois parties :Au début, je considère l'énergie micromagnétique avec des coefficients dégénératifs dans un fil mince. Après avoir montrer l'existence de minimiseurs du problème, j'identifie l'énergie limite lorsque la section du fil tend vers zéro.Dans la deuxième partie, j'étudie le comportement asymptotique des solutions dépendant du temps des problèmes micromagnétique dans une multi-structure constituée de la jonction de deux fils minces. En supposant que les volumes des deux fils tendent vers zéro avec la même vitesse. On obtient un problème limite couplé par une condition de jonction. Le problème limite reste non-convexe, mais devient complètement local.Dans le dernier chapitre, à partir d’un modèle variationnel 3D non convexe et non-local pour la polarisation électrique dans un matériau ferroélectrique, et à l'aide d'un processus asymptotique basé sur la réduction de dimension, j'analyse des phénomènes de jonction pour deux films minces ferroélectriques joints orthogonaux. Selon la façon dont la réduction se passe, on obtienttrois modèles différents de dimension 2. On remarque qu’un effet de mémoire du processus de réduction apparaît, ce dernier dépend de la compétition entre les épaisseurs des deux films: Le paramètre de guidage est la limite du rapport des épaisseurs des deux films / This thesis deals with partial differential equations coming from mathematical physics. Particularly, starting from 3D models for ferromagnetism and ferroelectricity, we derive 1D and 2D models via asymptotic processes based on dimensional reduction methods. The 3D model for ferromagnetism was proposed by W.F. Brown in the 40s and it is based on a system introduced by L.D. Landau and E.M. Lifschitz in 1935. About the ferroelectric model, we refer tothe papers of P. Chandra and P.B. Littlewood, W. Zhang and K. Bhattacharya and to the book of T. Mitsui, I. Taksuzaki, and E. Nakamura.This thesis based on three works:At the beginning, we consider micromagnetic energy, with some degenerating coefficients, in a thin wire. After showing the existence of minimizers, we identify the limit energy as the section of the wire vanishes.In the second part, we study the asymptotic behavior of the solutions of a time dependent micromagnetic problem in a multi-structure consisting of two joined thin wires. We assume that the volumes of the two wires vanish with same rate. We obtain two 1D limit problems coupled by a junction condition on the magnetization. The limit problem remains non-convex, but now it becomes completely local.In the last chapter, starting from a non-convex and nonlocal 3D variational model for the electric polarization in a ferroelectric material, and using an asymptotic process based on dimensional reduction, we analyze junction phenomena for two orthogonal joined ferroelectric thin films. We obtain three different 2D-variational models for joined thin films, depending on how the reduction happens. We note that, a memory effect of the reduction process appears, and it depends on the competition of the relative thickness of the two films: The guide parameter is the limit of the ratio between these two small thickness

Este projeto de doutorado apresenta uma nova abordagem MAP-MRF para a classificação contextual de imagens multiespectrais utilizando combinação de modelos de Campos Aleatórios Markovianos definidos em sistemas de ordens superiores. A modelagem estatística para o problema de classificação segue o paradigma Bayesiano, com a definição de um modelo Markoviano para os dados observados (Gaussian Markov Random Field multiespectral) e outro modelo para representar o conhecimento a priori (Potts). Nesse cenário, o parâmetro β do modelo de Potts atua como um parâmetro de regularização, tendo papel fundamental no compromisso entre as observações e o conhecimento a priori, de modo que seu correto ajuste é necessário para a obtenção de bons resultados. A introdução de sistemas de vizinhança de ordens superiores requer a definição de novos métodos para a estimação dos parâmetros dos modelos Markovianos. Uma das contribuições desse trabalho é justamente propor novas equações de pseudo-verossimilhança para a estimação desses parâmetros no modelo de Potts em sistemas de segunda e terceira ordens. Apesar da abordagem por máxima pseudo-verossimilhança ser amplamente utilizada e conhecida na literatura de campos aleatórios, pouco se conhece acerca da acurácia dessa estimação. Foram derivadas aproximações para a variância assintótica dos estimadores propostos, caracterizando-os completamente no caso limite, com o intuito de realizar inferências e análises quantitativas sobre os parâmetros dos modelos Markovianos. A partir da definição dos modelos e do conhecimento dos parâmetros, o próximo estágio é a classificação das imagens multiespectrais. A solução para esse problema de inferência Bayesiana é dada pelo critério de estimação MAP, onde a solução ótima é determinada maximizando a probabilidade a posteriori, o que define um problema de otimização. Como não há solução analítica para esse problema no caso de prioris Markovianas, algoritmos iterativos de otimização combinatória foram empregados para aproximar a solução ótima. Nesse trabalho, adotam-se três métodos sub-ótimos: Iterated Conditional Modes, Maximizer of the Posterior Marginals e Game Strategy Approach. Porém, é demonstrado na literatura que tais métodos convergem para máximos locais e não globais, pois são altamente dependentes de sua condição inicial. Isto motivou o desenvolvimento de uma nova abordagem para combinação de classificadores contextuais, que utiliza múltiplas inicializações simultâneas providas por diferentes classificadores estatísticos pontuais. A metodologia proposta define um framework MAP-MRF bastante robusto para solução de problemas inversos, pois permite a utilização e a integração de diferentes condições iniciais em aplicações como classificação, filtragem e restauração de imagens. Como medidas quantitativas de desempenho, são adotados o coeficiente Kappa de Cohen e o coeficiente Tau de Kendall para verificar a concordância entre as saídas dos classificadores e a verdade terrestre (amostras pré-rotuladas). Resultados obtidos mostram que a inclusão de sistemas de vizinhança de ordens superiores é de fato capaz de melhorar significativamente não apenas o desempenho da classificação como também a estimação dos parâmetros dos modelos Markovianos, reduzindo tanto o erro de estimação quanto a variância assintótica. Além disso, a combinação de classificadores contextuais através da utilização de múltiplas inicializações simultâneas melhora significativamente o desempenho da classificação se comparada com a abordagem tradicional com apenas uma inicialização. / This work presents a novel MAP-MRF approach for multispectral image contextual classification by combining higher-order Markov Random Field models. The statistical modeling follows the Bayesian paradigm, with the definition of a multispectral Gaussian Markov Random Field model for the observations and a Potts MRF model to represent the a priori knowledge. In this scenario, the Potts MRF model parameter (β) plays the role of a regularization parameter by controlling the tradeoff between the likelihood and the prior knowledge, in a way that a suitable tunning for this parameter is required for a good performance in contextual classification. The introduction of higher-order MRF models requires the specification of novel parameter estimation methods. One of the contributions of this work is the definition of novel pseudo-likelihood equations for the estimation of these MRF parameters in second and third order neighborhood systems. Despite its widely usage in practical MRF applications, little is known about the accuracy of maximum pseudo-likelihood approach. Approximations for the asymptotic variance of the proposed MPL estimators were derived, completely characterizing their behavior in the limiting case, allowing statistical inference and quantitative analysis. From the statistical modeling and having the model parameters estimated, the next step is the multispectral image classification. The solution for this Bayesian inference problem is given by the MAP criterion, where the optimal solution is obtained by maximizing the a posteriori distribution, defining an optimization problem. As there is no analytical solution for this problem in case of Markovian priors, combinatorial optimization algorithms are required to approximate the optimal solution. In this work, we use three suboptimal methods: Iterated Conditional Modes, Maximizer of the Posterior Marginals and Game Strategy Approach, a variant approach based on non-cooperative game theory. However, it has been shown that these methods converge to local maxima solutions, since they are extremelly dependent on the initial condition. This fact motivated the development of a novel approach for combination of contextual classifiers, by making use of multiple initializations at the same time, where each one of these initial conditions is provided by different pointwise pattern classifiers. The proposed methodology defines a robust MAP-MRF framework for the solution of general inverse problems since it allows the use and integration of several initial conditions in a variety of applications as image classification, denoising and restoration. To evaluate the performance of the classification results, two statistical measures are used to verify the agreement between the classifiers output and the ground truth: Cohens Kappa and Kendalls Tau coefficient. The obtained results show that the use of higher-order neighborhood systems is capable of significantly improve not only the classification performance, but also the MRF parameter estimation by reducing both the estimation error and the asymptotic variance. Additionally, the combination of contextual classifiers through the use of multiple initializations also improves the classificatoin performance, when compared to the traditional single initialization approach.

Mathematical theory of partial differential equations (PDEs) is a pretty old classical area with wide range of applications to almost every branch of science and engineering. With the advanced development of functional analysis and operator theory in the last century, it became a topic of analysis. The theory of homogenization of partial differential equations is a relatively new area of research which helps to understand the multi-scale phenomena which has tremendous applications in a variety of physical and engineering models, like in composite materials, porous media, thin structures, rapidly oscillating boundaries and so on. Hence, it has emerged as one of the most interesting and useful subject to study for the last few decades both as a theoretical and applied topic. In this thesis, we study asymptotic analysis (homogenization) of second-order partial differential equations posed on an oscillating domain. We consider a two dimensional oscillating domain (comb shape type) consisting of a fixed bottom region and an oscillatory (rugose) upper region. We introduce optimal control problems for the Laplace equation. There are mainly two types of optimal control problems; namely distributed control and boundary control. For distributed control problems in the oscillating domain, one can apply control on the oscillating part or on the fixed part and similarly for boundary control problem (control on the oscillating boundary or on the fixed part the boundary). We consider all the four cases, namely distributed and boundary controls both on the oscillating part and away from the oscillating part. The present thesis consists of 8 chapters. In Chapter 1, a brief introduction to homogenization and optimal control is given with relevant references. In Chapter 2, we introduce the oscillatory domain and define the basic unfolding operators which will be used throughout the thesis. Summary of the thesis is given in Chapter 3 and future plan in Chapter 8. Our main contribution is contained in Chapters 4-7. In chapters 4 and 5, we study the asymptotic analysis of optimal control problems namely distributed and boundary controls, respectively, where the controls act away from the oscillating part of the domain. We consider both L2 cost functional as well as Dirichlet (gradient type) cost functional. We derive homogenized problem and introduce the limit optimal control problems with appropriate cost functional. Finally, we show convergence of the optimal solution, optimal state and associate adjoint solution. Also convergence of cost-functional. In Chapter 6, we consider the periodic controls on the oscillatory part together with Neumann condition on the oscillating boundary. One of the main contributions is the characterization of the optimal control using unfolding operator. This characterization is new and also will be used to study the limiting analysis of the optimality system. Chapter 7 deals with the boundary optimal control problem, where the control is applied through Neumann boundary condition on the oscillating boundary with a suitable scaling parameter. To characterize the optimal control, we introduce boundary unfolding operators which we consider as a novel approach. This characterization is used in the limiting analysis. In the limit, we obtain two limit problems according to the scaling parameters. In one of the limit optimal control problem, we observe that it contains three controls namely; a distributed control, a boundary control and an interface control.

Les systèmes hyperboliques interviennent dans de nombreuses branches des sciences : théorie cinétique, mécanique des fluides non visqueux, magnéto hydrodynamique, dynamique des gaz non visqueux, trafic routier, flux d’une rivière ou d’un glacier, processus de sédimentation, processus d’échanges chimiques, etc. Et souvent, les systèmes qui régissent ces évènements font intervenir des petits paramètres, dont l’étude asymptotique permet d’envisager des simplifications mathématiques et/ou informatiques notoires. L’existence locale et l’existence globale de solutions, uniformément par rapport à ces paramètres, sont des questions fondamentales. Cette thèse regroupe à la fois des résultats généraux sur l’existence locale uniforme de solutions pour des systèmes hyperboliques quasi-linéaires du premier ordre ; et sur l’existence globale uniforme de solutions autour d’un équilibre constant pour ces mêmes systèmes. Le cas du système d’Euler-Maxwell ne satisfaisant pas les conditions requises pour l’existence uniforme globale, nous le traitons à part. / Hyperbolic systems arise in a large field of sciences : kinetic theory, inviscid reactive flow, magnetohydrodynamics, inviscid gas dynamics, traffic flow, river or glacier flow, sedimentation processes, chemical exchange processes, etc. In these kind of systems, small paramaters often appear, and an asymptotic study may lead to mathematical or computational simplifications. One fundamental problem that we may work on is local and global existence of solutions for these systems, uniformly with respect to these parameters. This Ph.D. thesis includes, on one hand, general results on uniform local existence of solutions for first order quasi-linear hyperbolic systems ; and on the other hand, results on uniform global existence of solutions near constant equilibriums for these same systems. In the case of Euler-Maxwell systems, required conditions are not fulfilled for uniform global existence, then we treat it separately.

Cette thèse est consacrée à l’étude mathématique des écoulements complexes dans des tubes minces. Les difficultés ne sont pas seulement liées à la rhéologie complexe, mais aussi aux conditions au bord sur la pression en entrée et en sortie (qui sont moins habituelles, mais réalistes du point de vue physique). Dans une première partie, des écoulements quasi-newtoniens stationnaires sont étudiés. D’abord, on utilise la petitesse du domaine pour montrer l’existence de la solution. Ensuite, on écrit un développement asymptotique de cette solution et on calcule formellement ses coefficients. Finalement, on justifie rigoureusement la validité de ce développement en démontrant des estimations d’erreur. Dans une deuxième partie, on considère des écoulements de fluides visco-élastiques décrits par la loi d’Oldroyd en régime stationnaire. Le modèle que nous avons choisi contient un terme diffusif en contrainte, dont l’ordre de grandeur est lié à la petitesse du domaine. Similairement à la première partie, un développement asymptotique est complètement justifié du point de vue mathématique. Dans le cas particulier de domaines axisymétriques une solution numérique est cherchée afin de la comparer à la solution obtenue via la technique asymptotique. Dans une dernière partie, on étudie les équations de Navier-Stokes non stationnaires. Un résultat d’existence des solutions fortes pour des données petites est démontré. Malheureusement, la méthode directe ne nous a pas permis pas d’avoir suffisamment de contrôle par rapport à la petitesse du domaine. Pour obtenir le résultat désiré, on utilise l’approche à la Kato, basé sur la théorie de C0 semigroupes. / This thesis is devoted to the mathematical analysis of complex flows in thin pipes. The difficulties stem not only from the complex rheology, but also from the boundary conditions used involving the pressure (which are rather atypical, but realistic from a physical point of view).In the first part, we study stationary, quasi-newtonian flows. The existence of a solution is shown using the smallness of the domain as a key ingredient. Furthermore, an asymptotic expansion of this solution is sought and its coefficients are formally computed. Lastly, the validity of this expansion is rigorously justified by proving error estimates. In the second part, we consider visco-elastic flows represented by Oldroyd’s law in stationary regime. The model which we have chosen contains a diffusive stress term, whose order of magnitude is related to the smallness of the domain. Similarly to the first part, a complete asymptotic expansion in mathematically justified. For the special case of axisymmetric domains a numerical solution is sought in order to compare it against the one obtained via the asymptotic technique. In the last part we study the non stationary Navier-Stokes equations. An existence result of strong solutions for small initial data is proven. Unfortunately, the direct method – based on energy estimates – doesn’t give us an optimal control of the smallness constant with respect to the size of the domain. To obtain the desired result, we employ the method of C 0 semigroups of linear operators.

Magister Scientiae - MSc / This thesis deals with the design and implementation of some novel numerical methods for non-linear singular pertubations problems (NSPPs). It provide a survey of asymptotic and numerical methods for some NSPPs in the past decade. By considering two test problems, rigorous asymptotic analysis is carried out. Based on this analysis, suitable numerical methods are designed, analyzed and implemented in order to have some relevant results of physical importance. Since the asymptotic analysis provides only qualitative information, the focus is more on the numerical analysis of the problem which provides the quantitative information. / South Africa

The work developed in this thesis focusses on the electrical activity of the heart, from the modeling of the action potential originating from cardiac cells and propagating through the heart, as well as its electrical manifestation at the body surface. The study is divided in two main parts: modeling the action potential, and numerical simulations. For modeling the action potential a dimensional and asymptotic analysis is done. The key advance in this part of the work is that this analysis gives the steps to reliably control the action potential. It allows predicting the time/space scales and speed of any action potential that is to say the shape of the action potential and its propagation. This can be done as the explicit relations on all the physiological constants are defined precisely. This method facilitates the integrative modeling of a complete human heart with tissue-specific ionic models. It even proves that using a single model for the cardiac action potential is enough in many situations. For efficient numerical simulations, a numerical method for solving the heart-torso coupling problem is explored according to a level set description of the domains. This is done in the perspective of using directly medical images for building computational domains. A finite element method is then developed to manage meshes not adapted to internal interfaces. Finally, an anisotropic adaptive remeshing methods for unstructured finite element meshes is used to efficiently capture propagating action potentials within complex, realistic two dimensional geometries.

Dans ce travail de thèse, il s'agit essentiellement de déterminer les propriétés mécaniques et physiques linéaires effectives des composites dans lesquels l'interface entre deux phases n'est pas lisse mais très rugueuse. Une approche efficace pour surmonter les difficultés provenant de la présence de rugosités d'interface consiste d'abord à homogénéiser une zone d'interface rugueuse comme une interphase équivalente par une analyse asymptotique et ensuite à appliquer des schémas micromécaniques pour estimer les propriétés effectives en tenant en compte de la présence de l'interphase équivalente. L'objectif principal de ce travail est de développer cette approche dans un cadre général où la surface autour de laquelle l'interface oscille périodiquement et rapidement peut être courbée et les phénomènes physiques concernés peuvent être couplés. Pour atteindre cet objectif, la conduction thermique est premièrement étudiée comme un prototype des phénomènes de transport non couplés pour élaborer dans un cadre simple les éléments essentiels de notre approche. Cette étude, préliminaire mais très utile au vu de l'importance des phénomènes de transport, montre que des résultats généraux et compacts peuvent s'obtenir quand l'interface est ondulée dans une seule direction et que des méthodes numériques sont en général nécessaires dans le cas où l'interface oscille suivant deux directions. L'approche développée et les résultats obtenus pour la conduction thermique sont étendus d'abord à l'élasticité linéaire et ensuite aux phénomènes physiques linéaires couplés tels que la thermoélectricité et la piézoélectricité. Dans ces cas plus complexes, des résultats généraux sont obtenus pour les composites stratifiés avec les interfaces ondulées dans une seule direction et des méthodes numériques sont élaborées pour les composites dans lesquels les interfaces oscillent suivant deux directions / This work is essentially concerned with determining the effective linear mechanical and physical properties of composites in which the interface between two phases is not smooth but very rough. An efficient approach to overcome the difficulties arising from the presence of interfacial roughness is first to homogenize a rough interface zone as an equivalent interphase by an asymptotic analysis and then to apply micromechanical schemes to estimation of the effective properties while accounting for the equivalent interphase. The present work aims mainly to develop this approach in a general situation where the surface around which an interface oscillates periodically and quickly can be curved and the physical phenomena involved can be coupled. To achieve this goal, thermal conduction is first studied as a prototype of transport phenomena so as to elaborate key elements of our approach in a simple situation. This study,even preliminary but very useful in view of the importance of transport phenomena, shows that general and compact results can be obtained when the interface is corrugated in only one direction and that numerical methods are generally required when an interface is curved along two directions. The approach developed and the results obtained for thermal conduction are extended first to linear elasticity and then to linear coupled physical phenomena such as thermoelectricity and piezoelectricity. In these more complex cases, general results are obtained for composite laminates with interfaces oscillating in only one direction, and numerical methods are elaborated for composites in which the interfaces oscillate in two directions

Ce manuscrit de thèse traite de l'analyse asymptotique de problèmes paraboliques possédant des termes raides. Dans un premier temps, on fait l'analyse asymptotique d'un système parabolique possédant des termes de transport raide. Une analyse à deux échelles, basée sur des résultats de théorie ergodique, nous permet de dériver un système limite effectif. Ce système effectif se trouve être, de nouveau, un système parabolique dont le champ de diffusion peut être explicité par une moyenne du champ de diffusion initial le long d'un groupe d'opérateurs unitaires. L'introduction d'un correcteur nous permet d'obtenir un résultat de convergence forte, avec un ordre de convergence, pour des données initiales non nécessairement bien préparées. On propose dans un second temps une méthode numérique permettant de calculer le champ de diffusion effectif. Celle-ci est basée sur la combinaison d'un schéma Runge-Kutta et d'un schéma de type semi-Lagrangien. L'ordre de convergence obtenu théoriquement est mis en évidence de manière numérique. On propose une méthode numérique basée sur un splitting d'opérateur pour la résolution du système parabolique avec termes de transport raide. Enfin, on effectue l'analyse asymptotique d'un système parabolique fortement anisotrope. Sous de bonnes hypothèses de régularité, un système variationnel effectif est proposé et l'introduction d'un correcteur adapté permet d'obtenir un résultat de convergence forte avec un ordre de convergence. Les arguments utilisés relèvent une nouvelle fois de l'analyse à deux échelles et de la théorie ergodique. / This manuscript is devoted to the asymptotic analysis of parabolic equations with stiff terms. First, we perform the asymptotic analysis of a parabolic equation with stiff transport terms. An effective limit model is obtained by a two-scale analysis based on ergodic theory results. This effective system is again a parabolic system whose diffusion field is an average of the initial diffusion field along a group of unitary operators. The introduction of a corrector allows us to obtain a strong convergence result, with an order of convergence, for initial data not necessarily well prepared. We propose a numerical method to compute the effective diffusion field. This method is based on a Runge-Kutta scheme and a semi-Lagrangian scheme. The theoretically order of convergence is obtained numerically. We propose a numerical method based on operator splitting for the resolution of the parabolic system with stiff transport terms. Finally, we perform the asymptotic analysis of a strongly anisotropic parabolic problem. Under suitable smoothness hypotheses, an effective variational system is proposed. By using a suitable corrector, we obtain a strong convergence result and we are able to perform the error analysis. The arguments relate again to the two-scale analysis and the ergodic theory.

>Magister Scientiae - MSc / This thesis deals with the design and implementation of some novel numerical methods for nonlinear singular perturbations problems (NSPPs). We provide a survey of asymptotic and numerical methods for some NSPPs in past decade. By considering two test problems, rigorous asymptotic analysis is carried out. Based on this analysis, suitable numerical methods are designed, analyzed and implemented in order to have some relevant results of physical importance. Since the asymptotic analysis provides only qualitative information, the focus is more on the numerical analysis of the problem which provides the quantitative information.