No description available.

La multiplication des sources d'information et le développement de nouvelles technologies ont engendré des bases données complexes, souvent caractérisées par un nombre de variables relativement élevé par rapport aux individus. En particulier, dans les études environnementales sur la pollution de l'air intérieur, la collecte des informations sur les individus se fait au regard de plusieurs thématiques, engendrant ainsi des données de grande dimension avec une structure multi-blocs définie par les thématiques. L'objectif de ce travail a été de développer des méthodes de classification adaptées à ces jeux de données de grande dimension et structurées en blocs de variables. La première partie de ce travail présente un état de l'art des méthodes de classification en général et dans le cas de la grande dimension. Dans la deuxième partie, trois nouvelles approches de classification d'individus décrits par des variables structurées en blocs ont été proposées. La méthode 2S-SOM (Soft Subspace-Self Organizing Map), une approche de type subspace clustering basée sur une modification de la fonction de coût de l'algorithme des cartes topologiques à travers un double système de poids adaptatifs défini sur les blocs et sur les variables. Nous proposons ensuite des approches CSOM (Consensus SOM) et Rv-CSOM de recherche de consensus de cartes auto-organisées basées sur un système de poids déterminés à partir des partitions initiales. Enfin, la troisième partie présente une application de ces méthodes sur le jeu de données réelles de la campagne nationale logement (CNL) menée par l'OQAI afin de définir une typologie des logements au regard des thématiques : qualité de l'air intérieur, structure du bâtiment, composition des ménages et habitudes des occupants. / The multiplication of information source and the development of news technologies generates complex databases, often characterized by relatively high number of variables compared to individuals. In particular, in the environmental studies on the indoor air quality, the information's collection is done according to several thematic, yielding column partitioned or multi-block data set. However, in case of high dimensional data, classical clustering algorithms are not efficient to find clusters which may exist in subspaces of the original space. The goal of this work is to develop clustering algorithms adapted to high dimensional data sets with multi-block structure. The first part of the work shows the state of art on clustering methods. In the second part, three new methods of clustering: the subspace clustering method 2S-SOM (Soft Subspace-Self Organizing Map)is based on a modified cost function of the Self Organizing Maps method across a double system of weights on the blocks and the variables. Then we propose two approaches to find the consensus of self-organized maps CSOM (Consensus SOM) and Rv-CSOM based on weights determined from initial partitions. The last part presents an application of these methods on the OQAI data to determine a typology of dwellings relatively to the following topics: indoor air quality, dwellings structure, household characteristics and habits of the inhabitants.

Les tumeurs cérébrales malignes sont la première cause de mortalité par cancer chez l’enfant avec une survie médiane de 12 à 14 mois et une survie globale à 5 ans de 20%, pour les gliomes de haut grade. Ce travail de thèse propose des méthodes innovantes pour l’analyse de blocs de données de génomiques, dans le but d’accroître les connaissances biologiques sur ces tumeurs. Les méthodes proposées étendent les travaux de Tenenhaus et al (2011), introduisant le cadre statistique général : Regularized Generalized Canonical Correlation Analysis (RGCCA). Dans un premier temps, nous étendons RGCCA à la gestion de données en grande dimension via une écriture duale de l’algorithme initial (KGCCA). Dans un deuxième temps, la problématique de la sélection de variables dans un contexte multi-Blocs est étudiée. Nous en proposons une solution avec la méthode SGCCA, qui pénalise la norme L1 des poids des composantes. Dans un troisième temps, nous nous intéressons à la nature des liens entre blocs avec deux autres adaptations. D’une part, la régression logistique multi-Blocs (multiblog) permet de prédire une variable binaire, comme la réponse à un traitement. D’autre part, le modèle de Cox multi-Blocs (multiblox) permet d’évaluer, par exemple, le risque instantané de rechute. Enfin, nous appliquons ces méthodes à l’analyse conjointe des données de transcriptome et d’aberrations du nombre de copies, acquises sur une cohorte de 53 jeunes patients avec un gliome de haut grade primaire. Les résultats sont décrits dans le dernier chapitre du manuscrit. / Cerebral malignant tumors are the leading cause of death among pediatric cancers with a median survival from 12 to 14 months and an overall survival of 20% at 5 years for high grade gliomas. This work proposes some innovative methods for the analysis of heterogeneous genomic multi-Block data, with the main objective of increasing biological knowledge about such tumors. These methods extend works of Tenenhaus and Tenenhaus (2011), who introduce Regularized Generalized Canonical Correlation Analysis (RGCCA) as a general statistical framework for multi-Block data analysis. As a first step, we extended RGCCA to handle large-Scale data with kernel methods (KGCCA). As a second step, SGCCA for variable selection within the RGCCA context is studied and leads to an additional constraint on the L1-Norm of the weight vectors. Then, as a third step, we focused on the nature of the links between blocks, with 2 other developments. On one hand, multi-Block logistic regression (multiblog) enables to predict a binary variable, such as response to treatment. On the other hand, the Cox model for multi-Block data (multiblox) enables the assessment of the instant risk, for instance, of relapse. We applied these methods to the joint analysis of Gene Expression and Copy Number Aberrations, acquired on a cohort of 53 young patients with a primary High Grade Glioma. Results are detailed in the last chapter of this work.

In order to make use of Newton&rsquo / s method for complex flow domains, an Euler multi-block Newton solver is developed. The generated Newton solver uses Analytical Jacobian derivation technique to construct the Jacobian matrices with different flux discretization schemes up to the second order face interpolations. Constructed sparse matrices are solved by parallel and series matrix solvers. In order to use structured grids for complex domains, multi-block grid construction is needed. Each block has its own Jacobian matrices and during the iterations the communication between the blocks should be performed. Required communication is performed with &ldquo / halo&rdquo / nodes. Increase in the number of grids requires parallelization to minimize the solution time. Parallelization of the analyses is performed by using matrix solvers having parallelization capability. In this thesis, some applications of the multi-block Newton method to different problems are given. Results are compared by using different flux discretization schemes. Convergence, analysis time and matrix solver performances are examined for different number of blocks.

Předložená diplomová práce se zabývá přípravou hydrogelů odvozených od polyakrylonitrilu a charakterizací jejich fyzikálních a chemických vlastností. Teoretická část shrnuje základní poznatky z oblasti hydrogelů, a také o polyakrylonitrilu. Dále se zabývá možností zpracování polyakrylonitrilu do reaktivní formy, takzvaného aquagelu a jeho zásadité hydrolýzy za účelem přípravy multi-blokových kopolymerů schopných tvořit 3D síť (gel). Experimentální část prezentuje výsledky charakterizace hydrogelů z polyakrylonitrilu neboli HYPANů, které byly připraveny bazicky katalyzovanou hydrolýzou aquagelu. Aquagel byl připraven rozpuštěním a následnou extruzí polyakrylonitrilu. Vzniklé vlákno bylo zpracováno do formy pelet, které byly dále užity pro zmiňovanou hydrolýzu. Hydrolýza byla prováděna za různých podmínek (teplota, NaOH koncentrace, reakční čas) za účelem přípravy produktů s různým stupněm konverze -CN skupiny. Hydrolýzou vytvořené hydrofóbní a hydrofilní bloky, mohou zformovat 3D síť o různých vlastnostech, závisejících na poměru mezi počtem a délkou bloků. Z hydrolyzátů byly připraveny hydrogely jejichž visko-elastické a optické vlastnosti byly dále charakterizovány. Optimalizací přípravy bylo dozaženo multi-blokového kopolymeru schopného vytvořit gel s vhodnými fyzikálními vlastnostmi. Tento gel může najít uplatnění v medicíně, například jako implantáty v oftamologii.

No description available.

The objective of this thesis is to accomplish the three dimensional parallel thin-layer Navier-Stokes solutions for low aspect ratio rectangular flat wings in compressible flow. Two block parallel Navier Stokes solutions of an aspect ratio 1.0 flat plate with sharp edges are obtained at different Mach numbers and angles of attack. Reynolds numbers are of the order of 1.0E5-3.0E5. Two different grid configurations, the coarse and the fine grids, are applied in order to speed up convergence. In coarse grid configuration, 92820 total grid points are used in two blocks, whereas it is 700,000 in fine grid. The flow field is dominated by the vortices and the separated flows. Baldwin Lomax turbulence model is used over the flat plate surface. For the regions dominated by the strong side edge vortices, turbulence model is modified using a polar coordinate system whose origin is at the minimum pressure point of the vortex. In addition, an algebraic wake-type turbulence model is used for the wake region behind the wing. The initial flow variables at the fine grid points are obtained by the interpolation based on the coarse grid results previously obtained for 40000 iterations. Iterations are continued with the fine grid about 20000-40000 more steps. Pressures of the top surface are predicted well with the exception of leading edge region, which may be due to unsuitable turbulence model and/or grid quality. The predictions of the side edge vortices and the size of the leading edge bubble are in good agreement with the experiment.

A three-dimensional, parallel, anisotropic, block-based, adaptive mesh refinement (AMR) algorithm is proposed and described for the solution of fluid flows on body-fitted, multi-block, hexahedral meshes. Refinement and de-refinement in any grid block computational direction, or combination of directions, allows the mesh to rapidly adapt to anisotropic flow features such as shocks, boundary layers, or flame fronts, common to complex flow physics. Anisotropic refinements and an efficient and highly scalable parallel implementation lead to a potential for significant reduction in computational cost as compared to a more typical isotropic approach. Unstructured root-block topology allows for greater flexibility in the treatment of complex geometries. The AMR algorithm is coupled with an upwind finite-volume scheme for the solution of the Euler equations governing inviscid, compressible, gaseous flow. Steady-state and time varying, three-dimensional, flow problems are investigated for various geometries, including the cubed-sphere mesh.

A three-dimensional, parallel, anisotropic, block-based, adaptive mesh refinement (AMR) algorithm is proposed and described for the solution of fluid flows on body-fitted, multi-block, hexahedral meshes. Refinement and de-refinement in any grid block computational direction, or combination of directions, allows the mesh to rapidly adapt to anisotropic flow features such as shocks, boundary layers, or flame fronts, common to complex flow physics. Anisotropic refinements and an efficient and highly scalable parallel implementation lead to a potential for significant reduction in computational cost as compared to a more typical isotropic approach. Unstructured root-block topology allows for greater flexibility in the treatment of complex geometries. The AMR algorithm is coupled with an upwind finite-volume scheme for the solution of the Euler equations governing inviscid, compressible, gaseous flow. Steady-state and time varying, three-dimensional, flow problems are investigated for various geometries, including the cubed-sphere mesh.

A computer code is developed using finite volume technique for solving steady twodimensional and axisymmetric compressible Euler and Navier-Stokes equations for internal flows by &ldquo / multi-block&rdquo / technique. For viscous flows, both laminar and turbulent flow properties can be used. Explicit one step second order accurate Lax-Wendroff scheme is used for time integration. Inviscid solutions are verified by comparing the results of test cases of a support project which was supported by ONERA/France for Turkey T-108, named &ldquo / 2-D Internal Flow Applications for Solid Propellant Rocket Motors&rdquo / . For laminar solutions, analytical flat plate solution is used for planar case and theoretical pipe flow solution is used for axisymmetric case for verification. Prandtl turbulent flow analogy is used in a flat plate solution to verify the turbulent viscosity calculation. The test cases solved with single-block code are compared with the ones solved with multi-block technique to verify the multi-block algorithm and good similarity is observed between single-block solutions and multi-block solutions. For the burning simulation of propellant of Solid Propellant Rocket Motors, injecting boundary is used. Finally, a segmented solid propellant rocket motor case is solved to show the multi-block algorithm&rsquo / s flexibility in solving complex geometries.