Les problèmes qui se posent aujourd'hui en mécanique numérique et domaines connexes sont complexes, et impliquent de plus en plus souvent plusieurs physiques à différentes échelles de temps et d’espace. Leur traitement numérique est en général long et difficile, d’où l’intérêt d’avoir accès à des méthodes et outils facilitant l’intégration de nouveaux modèles physiques dans des outils de simulation. Ce travail se pose dans la problématique du développement de codes de calcul numérique. L’approche proposée couvre la démarche de développement du modèle numérique depuis la formulation variationnelle jusqu’à l’outil de simulation. L’approche est appliquée à la méthode des éléments finis. Nous avons développé des concepts génériques afin d’automatiser la méthode des éléments finis. Nous nous sommes appuyés sur l'analyse tensorielle dans le contexte de la méthode des éléments finis. Le formalisme mathématique est basé sur l’algèbre tensorielle appliquée à la description de la discrétisation des formes variationnelles. Ce caractère générique est conservé grâce à l'approche logicielle choisie pour l’implantation; orientée objet en Java. Nous proposons donc un cadre orienté objet, basé sur des concepts symboliques, capables de gérer de manière symbolique les développements assistés des contributions élémentaires pour la méthode éléments finis. Ces contributions sont ensuite automatiquement programmées dans un code de calcul. L'intérêt de cette approche est la généricité de la description qui peut être étendue naturellement à tout autre modèle de discrétisation (spatiale ou temporelle). Dans ce travail, les concepts sont validés dans le cadre de problèmes linéaires simples (élasticité, chaleur,...), dans le cadre du traitement de formulations variationnelles mixtes (thermomécanique, Navier-Stokes,…) et dans un cadre Lagrangien (élasticité en grandes transformations, hyperélasticité,…). / The problems occurring today in computational mechanics and related domains are complex, and may involve several physics at different time and space scales. The numerical treatment of complex problems is in general tough and time consuming. In this context, the interest to develop methods and tools to accelerate the integration of new formulations into simulation tools is obvious. This work arises on the issue of the development of computational tool. The proposed approach covers the development process of numerical models from the variational statement to the simulation tool. The approach is applied to the finite element method. We have developed generic concepts to automate the development of the finite element method. To achieve this goal, we relied on tensor analysis applied in the context of the finite element method. The mathematical formalism is based on the tensor algebra to describe the discretization of a variational formulation. The generic character of the approach is preserved through the object-oriented approach in Java. We propose a framework based on object-oriented concepts capable of handling symbolic developments of elemental contributions for finite element codes. The advantage of this approach is the generic description that can be extended naturally to any discretization model in space or time. This concept is fully validated for simple linear problems (elasticity, heat convection, ...), for the treatment of mixed variational formulations (thermo-mechanical, Navier-Stokes for incompressible flows...) and Lagrangian frameworks (elasticity in larges transformations, hyperelasticity, ...).

A esfera é utilizada como domínio computacional na modelagem de diversos fenômenos físicos, como em previsão numérica do tempo. Sua discretização pode ser feita de diversas formas, sendo comum o uso de malha regulares em latitude/longitude. Recentemente, também para melhor uso de computação paralela, há uma tendência ao uso de malhas mais isotrópicas, dentre as quais a icosaédrica. Apesar de já existirem modelos atmosféricos que usam malhas icosaédricas, não há consenso sobre as metodologias mais adequadas a esse tipo de malha. Nos propusemos, portanto, a estudar em detalhe diversos fatores envolvidos no desenvolvimento de modelos atmosféricos globais usando malhas geodésicas icosaédricas. A discretização usual por volumes finitos para divergente de um campo vetorial utiliza como base o Teorema da Divergência e a regra do ponto médio nas arestas das células computacionais. A distribuição do erro obtida com esse método apresenta uma forte relação com características geométricas da malha. Definimos o conceito geométrico de alinhamento de células computacionais e desenvolvemos uma teoria que serve de base para explicar interferências de malha na discretização usual do divergente. Destacamos os impactos de certas relações de alinhamento das células na ordem da discretização do método. A teoria desenvolvida se aplica a qualquer malha geodésica e também pode ser usada para os operadores rotacional e laplaciano. Investigamos diversos métodos de interpolação na esfera adequados a malhas icosaédricas, e abordamos o problema de interpolação e reconstrução vetorial na esfera em malhas deslocadas. Usamos métodos alternativos de reconstrução vetorial aos usados na literatura, em particular, desenvolvemos um método híbrido de baixo custo e boa precisão. Por fim, utilizamos as técnicas de discretização, interpolação e reconstrução vetorial analisadas em um método semi-lagrangiano para o transporte na esfera em malhas geodésicas icosaédricas. Realizamos experimentos computacionais de transporte, incluindo testes de deformações na distribuição do campo transportado, que mostraram a adequação da metodologia para uso em modelos atmosféricos globais. A plataforma computacional desenvolvida nesta tese, incluindo geração de malhas, interpolações, reconstruções vetoriais e um modelo de transporte, fornece uma base para o futuro desenvolvimento de um modelo atmosférico global em malhas icosaédricas. / Spherical domains are used to model many physical phenomena, as, for instance, global numerical weather prediction. The sphere can be discretized in several ways, as for example a regular latitude-longitude grid. Recently, also motivated by a better use of parallel computers, more isotropic grids have been adopted in atmospheric global circulation models. Among those, the icosahedral grids are promising. Which kind of discretization methods and interpolation schemes are the best to use on those grids are still a research subject. Discretization of the sphere may be done in many ways and, recently, to make better use of computational resources, researchers are adopting more isotropic grids, such as the icosahedral one. In this thesis, we investigate in detail the numerical methodology to be used in atmospheric models on icosahedral grids. The usual finite volume method of discretization of the divergence of a vector field is based on the divergence theorem and makes use of the midpoint rule for integration on the edges of computational cells. The error distribution obtained with this method usually presents a strong correlation with some characteristics of the icosahedral grid. We introduced the concept of cell alignment and developed a theory which explains the grid imprinting patterns observed with the usual divergence discretization. We show how grid alignment impacts in the order of the divergence discretization. The theory developed applies to any geodesic grid and can also be used for other operators such as curl and Laplacian. Several interpolation schemes suitable for icosahedral grids were analysed, including the vector interpolation and reconstruction problem on staggered grids. We considered alternative vector reconstruction methods, in particular, we developed a hybrid low cost and good precision method. Finally, employing the discretizations and interpolations previously analysed, we developed a semi-Lagrangian transport method for geodesic icosahedral grids. Several tests were carried out, including deformational test cases, which demonstrated that the methodology is suitable to use in global atmospheric models. The computational platform developed in this thesis, including mesh generation, interpolation, vector reconstruction and the transport model, provides a basis for future development of global atmospheric models on icosahedral grids.

De récents résultats sur l’étude des systèmes de Lur’e (commutés) à temps discret mettent en avant une fonction de Lyapunov de type Lur’e avancée, dont les lignes de niveau peuvent être non convexes et non connexes. Celles-ci soulèvent de larges questions pour les systèmes de Lur’e à temps discret obtenus par la discrétisation d’un système continu. Les contributions de cette thèse sont d’apporter des éléments de réponse à ces questions. Tout d’abord, le verrou des lignes de niveau non-connexes est levé en construisant à partir de celles-ci une suite décroissante d’ensembles connexes et bornés qui converge vers l’origine et qui contient le futur de la trajectoire à temps continu. Dans un second temps, le problème de la stabilisation conjointe d’un système de Lur’e à données échantillonnées avec un échantillonnage non-uniforme est traité. Quand la période d’échantillonnage est à choisir parmi un nombre fini de valeurs, il est montré que ce problème se traduit comme la stabilisation conjointe d’un système commuté de Lur’e avec des incertitudes bornées en norme. En associant de plus à chaque mode un critère quadratique, une stratégie de type min-switching permet de résoudre cette question à l’aide d’un problème d’optimisation sous contraintes LMI. Enfin, les propriétés de la stratégie de min-switching pour les systèmes de Lur’e commutés à temps discret sont étudiées. Une extension de la notion de consistance permet de prouver que cette stratégie est consistante vis-à-vis de majorants quadratiques modaux du critère de performance et ainsi de garantir l’intérêt de la stratégie d’échantillonnage non-uniforme développée / Recent studies dealing with discrete-time (switched) Lur’e systems involve an adapted Lur’e type function exhibiting possibly non-convex and disconnected level sets. These properties raise fundamental issues in the case of discrete-time Lur’e system obtained by the sampling of a continuous time one. This PhD thesis aims at answering these questions. The first contribution is to avoid the discrete-time disconnected level sets by a decreasing sequence of bounded and connected sets that converges to the origin and that contain the future of the continuous-time trajectory. The second contribution deals with the joint stabilization of a sampled-data Lur’e system with non-uniform sampling. When the sampling period belongs to a finite set of values, this problem is reformulated as the joint stabilization of a discrete-time Lur’e switched system with norm-bounded uncertain parameters. Futhermore, if a quadratic criterion is associated with each mode, a min-switching strategy combined with LMI constraints allow to provide a solution to this problem. Finally the property of consistency for discrete-time switched Lur’e systems is investigated. It is shown that the min-switching strategy is consistent with respect to quadratic upper bounds of the performances. This result is applied on the stabilization of Lur’e systems with non-uniform sampling.

A busca por um conjunto de bases de alta qualidade é feita por vários grupos, pois a utilização destes conjuntos é importante para a obtenção de resultados exatos, tanto para propriedades atômicas quanto para propriedades moleculares. Os conjuntos usados neste trabalho são constituídos através da utilização do Método da Coordenada Geradora Hartree-Fock em sua versão polinomial e foram construídos de forma a permitirem uma maior flexibilidade à camada de valência do átomo, flexibilidade esta que foi proporcionada pelo número de funções que descrevem a última camada do átomo. Uma nova metodologia para a construção de conjuntos de base foi proposta neste trabalho. Esta nova metodologia diminui a busca artesanal pela busca pelos conjuntos de base introduzindo o Método Simplex em nosso programa gerador de base GENBASIS minimizando o tempo de procura por essa bases de qualidade. A aplicação destes conjuntos para estimativa de propriedades atômicas e moleculares foi feita em comparação com conjuntos de base largamente utilizados na literatura, envolvendo cálculos com vários níveis de teoria, desde o método Hartree-Fock até a Teoria do Funcional Densidade. / The search for high quality basis set is made by several groups since the quality of the basis sets is a very important factor to obtain accurate results for atomic and molecular properties. The basis sets used in this work are designed by using the polynomial Generate Coordinate Method and were constructed with the aim to improve the flexibility of a basis sets to describe the valence shell. A new methodology for the search of this basis set is proposed in this work. This new methodology makes the search for this basis set less artisanal with the introduction of the Simplex Method in our basis generator program, GENBASIS. This introduction minimizes the time used in the search for these quality basis set. The application of our basis sets in atomic and molecular calculations were compared with the most used basis sets in the literature and were performed by using different level of calculations, namely, from Hartree-Fock to Density Functional Theory.

Esta dissertação trata da solução numérica de problemas em dinâmica dos fluidos usando dois novos esquemas upwind de alta resolução, denominados FDPUS-C1 (Five-Degree Polynomial Upwind Scheme of \' C POT. 1\' Class) e SDPUS-C1 (Six-Degree Polynomial Upwind Scheme of \'C POT.1\' Class), para a discretização de termos convectivos lineares e não-lineares. Os esquemas são baseados nos critérios de estabilidade TVD (Total Variation Diminishing) e CBC (Convection Boundedness Criterion) e são implementados, nos contextos das metodologias de diferenças finitas e volumes finitos, no ambiente de simulação Freeflow (an integrated simulation system for Free surface Flow) para escoamentos imcompressíveis 2D, 2D-1/2 e 3D, ou no código bem conhecido CLAWPACK ( Conservation LAW PACKage) para problemaw compressíveis 1D e 2D. Vários testes computacionais são feitos com o objetivo de verificar e validar os métodos numéricos contra esquemas upwind populares. Os novos esqumas são então aplicados na resolução de uma gama ampla de problemas em CFD (Computational Fluids Dynamics), tais como propagação de ondas de choque e escoamentos incompressíveis envolvendo superfícies livres móveis. Em particular, os resultados numéricos para leis de conservação hiperbólicas 2D e equações de Navier-Stokes incompressíveis 2D, 2D-1/2 e 3D demosntram que esses novos esquemas convectivos tipo upwind polinomiais funcionam muito bem / This dissertation deals with the numerical solution of fluid dynamics problems using two new high resolution upwind schemes,. namely FDPUS-C1 and SDPUS-C1, for the discretization of the linear and non-linear convection terms. The Schemes are based on TVD and DBC stability criteria and are implemented in the context of the finite difference and finite volume methodologies, either into the Freeflow code for 2D, 2D-1/2 and 3D incompressible flows or in the well-known CLAWPACK code for 1D and 2D compressible flows. Several computational tests are performed to verify and validate the numerical methods against other popularly used upwind schemes. The new schemes are then applied to solve a wide range of problems in CFD, such as shock wave propagation and incompressible fluid flows involving moving free msurfaces. In particular, the numerical results for 2D hyperbolic conservation laws and 2D, 2D-1/2 and 3D incompressible Navier-Stokes eqautions show that new polynomial upwind convection schemes perform very well

No description available.

Les méthodes d'éléments finis de type Galerkine discontinu (DG FEM) ont démontré précision et efficacité pour résoudre des problèmes d'ondes dans des milieux complexes. Cependant, elles nécessitent un très grand nombre de degrés de liberté, ce qui augmente leur coût de calcul en comparaison du coût des méthodes d'éléments finis continus.Parmi les différentes approches variationnelles pour résoudre les problèmes aux limites, se distingue une famille particulière, basée sur l'utilisation de fonctions tests qui sont des solutions locales exactes des équations à résoudre. L'idée vient de E.Trefftz en 1926 et a depuis été largement développée et généralisée. Les méthodes variationnelles de type Trefftz-DG appliquées aux problèmes d'ondes se réduisent à des intégrales de surface, ce qui devrait contribuer à réduire les coûts de calcul.Les approches de type Trefftz ont été largement développées pour les problèmes harmoniques, mais leur utilisation pour des simulations en domaine transitoire est encore limitée. Quand elles sont appliquées dans le domaine temporel, les méthodes de Trefftz utilisent des maillages qui recouvrent le domaine espace-temps. C'est une des paraticularités de ces méthodes. En effet, les méthodes DG standards conduisent à la construction d'un système semi-discret d'équations différentielles ordinaires en temps qu'on intègre avec un schéma en temps explicite. Mais les méthodes de Trefftz-DG appliquées aux problèmes d'ondes conduisent à résoudre une matrice globale, contenant la discrétisation en espace et en temps, qui est de grande taille et creuse. Cette particularité gêne considérablement le déploiement de cette technologie pour résoudre des problèmes industriels.Dans ce travail, nous développons un environnement Tre#tz-DG pour résoudre des problèmes d'ondes mécaniques, y compris les équations couplées de l'élasto-acoustique. Nous prouvons que les formulations obtenues sont bien posées et nous considérons la difficulté d'inverser la matrice globale en construisant un inverse approché obtenu à partir de la décomposition de la matrice globale en une matrice diagonale par blocs. Cette idée permet de réduire les coûts de calcul mais sa précision est limitée à de petits domaines de calcul. Etant données les limitations de la méthode, nous nous sommes intéressés au potentiel du "Tent Pitcher", en suivant les travaux récents de Gopalakrishnan et al. Il s'agit de construire un maillage espace-temps composé de macro-éléments qui peuvent être traités indépendamment en faisant une hypothèse de causalité. Nous avons obtenu des résultats préliminaires très encourageants qui illustrent bien l'intérêt du Tent Pitcher, en particulier quand il est couplé à une méthode de Trefftz-DG formulée à partir d'intégrales de surface seulement. Dans ce cas, le maillage espace-temps est composé d'éléments qui sont au plus de dimension 3. Il est aussi important de noter que ce cadre se prête à l'utilisation de pas de temps locaux ce qui est un plus pour gagner en précision avec des coûts de calcul réduits. / Discontinuous Finite Element Methods (DG FEM) have proven flexibility and accuracy for solving wave problems in complex media. However, they require a large number of degrees of freedom, which increases the corresponding computational cost compared with that of continuous finite element methods. Among the different variational approaches to solve boundary value problems, there exists a particular family of methods, based on the use of trial functions in the form of exact local solutions of the governing equations. The idea was first proposed by Trefftz in 1926, and since then it has been further developed and generalized. A Trefftz-DG variational formulation applied to wave problems reduces to surface integrals that should contribute to decreasing the computational costs.Trefftz-type approaches have been widely used for time-harmonic problems, while their implementation for time-dependent simulations is still limited. The feature of Trefftz-DG methods applied to time-dependent problems is in the use of space-time meshes. Indeed, standard DG methods lead to the construction of a semi-discrete system of ordinary differential equations in time which are integrated by using an appropriate scheme. But Trefftz-DG methods applied to wave problems lead to a global matrix including time and space discretizations which is huge and sparse. This significantly hampers the deployment of this technology for solving industrial problems.In this work, we develop a Trefftz-DG framework for solving mechanical wave problems including elasto-acoustic equations. We prove that the corresponding formulations are well-posed and we address the issue of solving the global matrix by constructing an approximate inverse obtained from the decomposition of the global matrix into a block-diagonal one. The inversion is then justified under a CFL-type condition. This idea allows for reducing the computational costs but its accuracy is limited to small computational domains. According to the limitations of the method, we have investigated the potential of Tent Pitcher algorithms following the recent works of Gopalakrishnan et al. It consists in constructing a space-time mesh made of patches that can be solved independently under a causality constraint. We have obtained very promising numerical results illustrating the potential of Tent Pitcher in particular when coupled with a Trefftz-DG method involving only surface terms. In this way, the space-time mesh is composed of elements which are 3D objects at most. It is also worth noting that this framework naturally allows for local time-stepping which is a plus to increase the accuracy while decreasing the computational burden.

Cette thèse répond à une problématique de commande frontière d’une conduite acoustique dont l’actionnement est assuré par un réseau d’actionneurs/capteurs co-localisés constituant une peau active. Pour faire face au caractère intrinsèquement multiphysique de ce problème vibro-acoustique, nous avons choisi dans cette thèse d’employer une approche hamiltonienne à ports, approche structurée basée sur la représentation des échanges entre différents domaines énergétiques au sein d’un système et entre différents systèmes. Nous avons proposé une modélisation hamiltonienne à ports de l’équation d’onde interconnectée à la frontière au système d’actionnement distribué, correspondant à une formulation 2D du problème physique. Nous avons développé une méthode de discrétisation spatiale basée sur l’utilisation de différences finies sur plusieurs grilles en quinconce qui préserve la structure hamiltonienne à ports de l’équation d’onde. Cette méthode permet en outre d’interconnecter facilement le système discrétisé avec d’autres sous-systèmes, dans le but de mettre en place un actionnement par exemple. Son principal avantage sur d’autres méthodes préservatives de structure réside dans sa simplicité de mise en œuvre qui découle de l’utilisation de différences finies. Concernant la commande du système vibro-acoustique, nous avons proposé une méthode de synthèse de loi de commande distribuée pour les systèmes régis par deux lois de conservation en 1D. L’originalité de cette méthode réside en le fait qu’elle repose sur le calcul d’invariants structuraux (fonctions de Casimir) exploités afin de modifier la structure du système en boucle fermée. Les conditions d’application sur un système 2D sont étudiées et des résultats numériques valident les lois de commande synthétisées. / This thesis deals with the boundary control of an acoustic by a network of co-localised sensors/actuators which constitutes a smart skin. In order to cope with this multiphysical problem, we chose to place our study in the framework of port-Hamiltonian systems, a structured approach based on the representation of energy exchanges between different energy domains between different systems of subsystems. We proposed a port-Hamiltonian model of the wave equation interconnected through its boundary to the distributed actuation system, which corresponds to a 2D formulation of the physical problem. We developed a spatial discretization method based on the use of finite differences on several staggered grids that preserve the port-Hamiltonian structure of the wave equation. This method also permits to easily interconnect the discretized system with other subsystems, which is convenient for instance for control purposes. Its main advantage over other structure preserving methods is its simplicity of implementation which stems from the use of finite differences. In order to control the vibro-acoustic system, we proposed a control law synthesis method for systems governed by two conservation laws in 1D. The originality of this method lies in the fact that it relies on the computation of structural invariants (Casimir functions) exploited in order to modify the structure of the system in closed loop. The conditions of application of these laws on a 2D system are studied and numerical results validate the synthesized control laws.

Made available in DSpace on 2014-06-11T19:26:18Z (GMT). No. of bitstreams: 0 Previous issue date: 2009-01-12Bitstream added on 2014-06-13T20:54:32Z : No. of bitstreams: 1 souza_mp_me_bauru.pdf: 1121017 bytes, checksum: 8ed8fe65e5cd46111ec16064bc42b009 (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / No presente trabalho desenvolve-se um estudo de modelagem matemática e simulação numérica do escoamento bifásico gás-sólido em leito fluidizado borbulhante. Utiliza-se o modelo Euleriano de duas fases separadas formulando o tensor das tensões da fase sólida através da teoria cinética dos escoamentos granulares. As simulações numéricas são realizadas através do código fonte MFIX (Multiphase Flow with Interphase eXchanges) desenvolvido no NETL (National Energy Technology Laboratory). Os resultados de simulação numérica são avaliados por meio da análise da influência dos seguintes parâmetros: malha computacional e esquemas de discretização dos termos convectivos das equações de conservação. Com base nos estudos teóricos e resultados obtidos durante o trabalho conclui-se que esquemas de primeira, tais como FOUP são altamente difusivos, já os resultados obtidos utilizando o esquema de alta ordem, Superbee, produziu resultados de melhor qualidade para as malhas testadas neste trabalho. Além disso, os resultados mostraram-se bastante dependentes do tamanho da malha computacional. / In the present work is described a mathematical model and numerical simulation of gas-solid two-phase flow in a bubbling fluidized bed. It is used the Eulerian gas-solid two-fluid model and the solid phase stress tensor is modeled considering the kinetic theory of granular flows. The numerical simulations were developed using the MFIX (Multiphase Flow with Interphase eXchanges) code developed in NETL (National Energy Technology Laboratory). The numerical diffusion is analyzed considering a single bubbling detachment and its movement process in a two-dimensional bubbling fluidized bed using the bubble shape as a metric for results description. The influence of computacional grid it is also analyzed. It is concluded that SuperBee scheme produces the better results and analysis about estimating uncertainty in grid refinement should be studied.

Advances in data collection methods, storage and processing technology are providing a unique challenge and opportunity for automated data learning techniques which aim at producing high-level information, or models, from data. A Typical knowledge discovery process consists of data selection, data preparation, data transformation, data mining and interpretation/validation of the results. Thus, we develop automatic learning techniques which contribute to the data preparation, transformation and mining tasks of knowledge discovery. In doing so, we try to improve the prediction accuracy of the overall learning process. Our work focuses on decision tree based learning and thus, we introduce various preprocessing and transformation techniques such as discretization, fuzzy partitioning and dimensionality reduction to improve this type of learning. However, these techniques can be used in other learning methods e.g. discretization can also be used for naive-bayes classifiers. The data preparation step represents almost 80 percent of the problem and is both time consuming and critical for the quality of modeling. Discretization of continuous features is an important problem that has effects on accuracy, complexity, variance and understandability of the induction models. In this thesis, we propose and develop resampling based aggregation techniques that improve the quality of discretization. Later, we validate by comparing with other discretization techniques and with an optimal partitioning method on 10 benchmark data sets.The second part of our thesis concerns with automatic fuzzy partitioning for soft decision tree induction. Soft or fuzzy decision tree is an extension of the classical crisp tree induction such that fuzzy logic is embedded into the induction process with the effect of more accurate models and reduced variance, but still interpretable and autonomous. We modify the above resampling based partitioning method to generate fuzzy partitions. In addition we propose, develop and validate another fuzzy partitioning method that improves the accuracy of the decision tree.Finally, we adopt a topological learning scheme and perform non-linear dimensionality reduction. We modify an existing manifold learning based technique and see whether it can enhance the predictive power and interpretability of classification. / La recherche avancée dans les méthodes d'acquisition de données ainsi que les méthodes de stockage et les technologies d'apprentissage, s'attaquent défi d'automatiser de manière systématique les techniques d'apprentissage de données en vue d'extraire des connaissances valides et utilisables.La procédure de découverte de connaissances s'effectue selon les étapes suivants: la sélection des données, la préparation de ces données, leurs transformation, le fouille de données et finalement l'interprétation et validation des résultats trouvés. Dans ce travail de thèse, nous avons développé des techniques qui contribuent à la préparation et la transformation des données ainsi qu'a des méthodes de fouille des données pour extraire les connaissances. A travers ces travaux, on a essayé d'améliorer l'exactitude de la prédiction durant tout le processus d'apprentissage. Les travaux de cette thèse se basent sur les arbres de décision. On a alors introduit plusieurs approches de prétraitement et des techniques de transformation; comme le discrétisation, le partitionnement flou et la réduction des dimensions afin d'améliorer les performances des arbres de décision. Cependant, ces techniques peuvent être utilisées dans d'autres méthodes d'apprentissage comme la discrétisation qui peut être utilisées pour la classification bayesienne.Dans le processus de fouille de données, la phase de préparation de données occupe généralement 80 percent du temps. En autre, elle est critique pour la qualité de la modélisation. La discrétisation des attributs continus demeure ainsi un problème très important qui affecte la précision, la complexité, la variance et la compréhension des modèles d'induction. Dans cette thèse, nous avons proposes et développé des techniques qui ce basent sur le ré-échantillonnage. Nous avons également étudié d'autres alternatives comme le partitionnement flou pour une induction floue des arbres de décision. Ainsi la logique floue est incorporée dans le processus d'induction pour augmenter la précision des modèles et réduire la variance, en maintenant l'interprétabilité.Finalement, nous adoptons un schéma d'apprentissage topologique qui vise à effectuer une réduction de dimensions non-linéaire. Nous modifions une technique d'apprentissage à base de variété topologiques `manifolds' pour savoir si on peut augmenter la précision et l'interprétabilité de la classification.