Les méthodes actuelles d’optimisation de formes permettent des gains importants vis-à-vis des fonctions à optimiser. Elles sont largement utilisées par les industriels, notamment dans l’automobile comme chez Renault. Parmi ces approches, l’optimisation de formes paramétriques permet d’obtenir rapidement une géométrie optimale sous réserve que l’espace de conception soit assez restreint pour pouvoir être parcouru en un temps de simulation raisonnable. Ce manuscrit présente deux méthodes d’optimisation de formes paramétriques en grande dimension pour des applications nécessitant des coûts de calculs importants, par exemple en mécanique des fluides. Une manière originale de reconstruire un modèle CAO paramétré à partir d’une surface morte est présentée au préalable. Nous proposons une approche pour identifier les zones interdites de l’espace de conception ainsi que leurs gestions dans une boucle d’optimisation par plans d’expériences. La première méthode s’appuie sur des techniques statistiques pour lever le verrou du nombre de degrés de liberté et utilise une optimisation à deux niveaux de fidélité pour minimiser les temps de calcul. Cette méthode en rupture avec le processus industriel habituel a été appliquée pour optimiser le coefficient de trainée aérodynamique d’un véhicule. La seconde méthode se base sur l’exploitation des gradients fournis par les solveurs adjoints, c’est-à-dire sur les sensibilités du critère (comme l’uniformité d’un écoulement par exemple) par rapport aux degrés de liberté de l’optimisation. Cette méthode innovante et en rupture avec les approches classiques permet de lever très naturellement le verrou du nombre de paramètres. Cependant, les gradients fournis par les logiciels ne sont pas donnés par rapport aux paramètres CAO mais par rapport aux nœuds du maillage. Nous proposons donc une façon d’étendre ces gradients jusqu’aux paramètres CAO. Des exemples académiques ont permis de montrer la pertinence et la validité de notre approche. / Current design loops for shape optimizations allow significant improvements in relation to the functions that need to be optimized. They are widely used in industry, particularly in the car industry like Renault. Among these approaches, parametric shape optimization allows rapid enhancement of the shape, on the condition that the design space is confined enough in order to be explored within a reasonable computational time. This Thesis introduces two CAD-based large-scale shape optimization methods for products requiring significant computational cost, for example in fluid mechanics. An original way to create a parameterized CAD model developed from a dead geometry is presented first. We propose an approach to identify the restricted areas of the design space and their managements in an optimization loop that uses a design of experiments. The first method is based on statistical techniques to circumvent the difficulties of large-scale optimizations and uses a two-level multi-fidelity modelling approach to minimize the computational time. This method, breaking away from the usual industrial process, was applied to optimize the aerodynamic drag coefficient of a car body. The second method is based on the gradients provided by adjoint solvers, that is to say on the sensitivity of the cost function (such as the uniformity deviation for example) with respect to the design points or displacement boundaries. This innovative method breaking away from classical approaches naturally gets over the number of degrees of freedom. However, the sensitivities provided by softwares are not computed with respect to CAD parameters but with respect to the coordinates of the vertices of the surface mesh. Thus, we propose a way to extend these gradients to CAD parameters. Academic test cases have proved the efficiency and accuracy of our method.

We develop a stochastic parametrization, based on a 'simple' deterministic model for the dynamics of steady longshore currents, that produces ensembles that are statistically consistent with field observations of these currents. Unlike deterministic models, stochastic parameterization incorporates randomness and hence can only match the observations in a statistical sense. Unlike statistical emulators, in which the model is tuned to the statistical structure of the observation, stochastic parametrization are not directly tuned to match the statistics of the observations. Rather, stochastic parameterization combines deterministic, i.e physics based models with stochastic models for the "missing physics" to create hybrid models, that are stochastic, but yet can be used for making predictions, especially in the context of data assimilation. We introduce a novel measure of the utility of stochastic models of complex processes, that we call consistency of sensitivity. A model with poor consistency of sensitivity requires a great deal of tuning of parameters and has a very narrow range of realistic parameters leading to outcomes consistent with a reasonable spectrum of physical outcomes. We apply this metric to our stochastic parametrization and show that, the loss of certainty inherent in model due to its stochastic nature is offset by the model's resulting consistency of sensitivity. In particular, the stochastic model still retains the forward sensitivity of the deterministic model and hence respects important structural/physical constraints, yet has a broader range of parameters capable of producing outcomes consistent with the field data used in evaluating the model. This leads to an expanded range of model applicability. We show, in the context of data assimilation, the stochastic parametrization of longshore currents achieves good results in capturing the statistics of observation that were not used in tuning the model.

Tato práce se zabývá extrakcí vhodných příznaků pro rozpoznávání řečníka z delších časových úseků. Po představení současných technik pro extrakci takových příznaků navrhujeme a popisujeme novou metodu pracující v časovém rozsahu fonémů a využívající známou techniku i-vektorů. Velké úsilí bylo vynaloženo na nalezení vhodné reprezentace temporálních příznaků, díky kterým by mohly být systémy pro rozpoznávání řečníka robustnější, zejména modelování prosodie. Náš přístup nemodeluje explicitně žádné specifické temporální parametry řeči, namísto toho používá kookurenci řečových rámců jako zdroj temporálních příznaků. Tuto techniku testujeme a analyzujeme na řečové databázi NIST SRE 2008. Z výsledků bohužel vyplývá, že pro rozpoznávání řečníka tato technika nepřináší očekávané zlepšení. Tento fakt diskutujeme a analyzujeme ke konci práce.

Numerical weather prediction (NWP) and climate models have shown to have a challenge to correctly simulate stable boundary layers and diurnal cycles. This aim of this study is to evaluate, describe and give suggestions for improvements of the descriptions of stable boundary layers in operational NWP models. Two papers are included. Paper I focuses on the description of the surface and the interactions between the surface and the boundary layer in COAMPSR, a regional NWP model. The soil parametrization showed to be of great importance to the structure of the boundary layer. Moreover, it showed also that a low frequency of radiation calculations caused a bias in received solar energy at the surface. In paper II, the focus is on the formulation of the turbulent transport in stable boundary layers. There, an implementation of a diffusion parametrization based on the amount of turbulent kinetic energy (TKE) is tested in a single column model (SCM) version of the global NWP model Integrated Forecast System (IFS). The TKE parametrization turned out to behave similarly as the currently operational diffusion parametrization in convective regimes and neutral regimes, but showed to be less diffusive in weakly stable and stable conditions. The formulations of diffusion also turned out to be very dependent on the length scale formulation. If the turbulence and the gradients of wind temperature and wind are weak, the magnitude of turbulence can enter an oscillating mode. This oscillation can be avoided with the use of a lower limit of the length scale. / Det har visat sig att det är en stor utmaning för numeriska väderprognosmodeller (NWP-modeller) att simulera stabilt skiktade atmosfäriska gränsskikt och gränsskiktets dygnscykel på ett korrekt sätt. Syftet med denna studien är att utvärdera, beskriva och ge förslag på förbättringar av beskrivningen av gränsskiktet i NWP-modeller. Studien innehåller två artiklar. Den första fokuserar på beskrivningen av markytan och interaktionen mellan marken och gränsskiktet i den regionala NWP-modellen COAMPS R . Det visade sig att beskrivningen av markytan har en signifikant inverkan på gränsskiktets struktur. Det framkom också att strålningsberäkningarna endast görs en gång i timmen vilket bland annat orsakar en bias i inkommande solinstrålning vid markytan. Den andra artikeln fokuserar på beskrivningen av den turbulenta transporten i stabila skiktade gränsskikt. En implemenering av en diffusionsparametrisering som bygger på turbulent kinetisk energy (TKE) testas i en endimensionell version av NWP-modellen Integrated Forecast System (IFS), utvecklat vid European Center for Medium Range Weather Forecasts (ECMWF). Den TKE-baserade diffussionsparametriseringen är likvärdigt med den nuvaran de operationella parametriseringen i neutrala och konvektiva gränsskikt, menär mindre diffusivt i stabila gränsskikt. Diffusionens intensitet är beroende påden turbulenta längdskalan. Vidare kan turbulensen i TKE-formuleringen hamna i ett oscillerande läge om turbulensen är svag samtidigt som temperatur- och vindgradienten är kraftig. Denna oscillation kan förhindras om längdskalans minsta tillåtna värde begränsas.

京都大学 / 0048 / 新制・課程博士 / 博士(理学) / 甲第18766号 / 理博第4024号 / 新制||理||1580(附属図書館) / 31717 / 京都大学大学院理学研究科数学・数理解析専攻 / (主査)准教授 伊藤 哲史, 教授 上田 哲生, 教授 雪江 明彦 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DFAM

No description available.

The work deals with the use of geometric parameterization for shape description of some parts of the airplane. Geometric parametrization is used for creating a parametric model airfoil. This parametric model allows local deformations pobrchu profile and can easily be applied to generate the geometry of the wing or other parts letoumu. Some properties of the parametric model were tested applications in aerodynamic optimization. Furthermore, the work deals with the parametric description of the blades, the aerodynamic optimization and noise analysis. For propeller blade were created distribution function of the control parameters that can be used in aerodynamic optimization of the blades. Geometric parameterization is used for identifying the location and other characteristics of noise sources.

Les équipements de communications numériques intègrent de plus en plus de standards. La commutation d’un standard à l’autre doit pouvoir se faire au prix d’un surcoût matériel modéré, ce qui impose l’utilisation de ressources communes dans des instanciations différentes. La plateforme matérielle nécessaire à l’exécution d’une couche physique multistandard est le segment du système présentant le plus de contraintes par rapport à la reconfiguration : réactivité, consommation et occupation de ressources matérielles. Nos travaux se focalisent sur la paramétrisation qui vise une implémentation multistandards efficace. L’objectif de cette technique est d’identifier des traitements communs entre les standards, voire entre blocs de traitement au sein d’un même standard, afin de définir des blocs génériques pouvant être réutilisés facilement. Nous définissons le management d’opérateurs mutualisés (opérateurs communs) et nous étudions leur implémentation en se basant essentiellement sur des évaluations de complexité pour quelques standards utilisant la modulation OFDM. Nous proposons en particulier l’architecture d’un opérateur commun permettant la gestion efficace des ressources matérielles entre les algorithmes FFT et décodage de Viterbi. L’architecture, que nous avons proposé et implémenté sur FPGA, permet d’adapter le nombre d’opérateurs communs alloués à chaque algorithme et donc permet l’accélération des traitements. Les résultats montrent que l’utilisation de cette architecture commune offre des gains en complexité pouvant atteindre 30% dans les configurations testées par rapport à une implémentation classique avec une réduction importante de l’occupation mémoire. / Today's telecommunication systems require more and more flexibility, and reconfiguration mechanisms are becoming major topics especially when it comes to multistandard designs. In typical hardware designs, the communication standards are implemented separately using dedicated instantiations which are difficult to upgrade for the support of new features. To overcome these issues, we exploit a parameterization approach called the Common Operator (CO) technique that can be considered to build a generic terminal capable of supporting a large range of communication standards. The main principle of the CO technique is to identify common elements based on smaller structures that could be widely reused across signal processing functions. This technique aims at designing a scalable digital signal processing platform based on medium granularity operators, larger than basic logic cells and smaller than signal processing functions. In this thesis, the CO technique is applied to two widely used algorithms in wireless communication systems: Viterbi decoding and Fast Fourier Transform (FFT). Implementing the FFT and Viterbi algorithms in a multistandard context through a common architecture poses significant architectural constraints. Thus, we focus on the design of a flexible processor to manage the COs and take advantage from structural similarities between FFT and Viterbi trellis. A flexible FFT/Viterbi processor was proposed and implemented on FPGA and compared to dedicated hardware implementations. The results show a considerable gain in flexibility. This gain is achieved with no complexity overhead since the complexity if even decreased up to 30% in the considered configurations.

No description available.

O objetivo deste trabalho é desenvolver uma ferramenta de automatização de cálculo para projeto estrutural de pórticos rolantes. Com o apoio da geração automática de um modelo de elementos finitos e um memorial de cálculo, o projeto estrutural desses equipamentos pode ser rapidamente verificado quanto ao limite de escoamento do material e resistência à flambagem. Optou-se pela utilização do método dos elementos finitos para o cálculo estrutural, pois se trata de uma ferramenta de cálculo moderna, que permite avaliar soluções para as quais não há ferramentas analíticas disponíveis. Porém, o tempo para geração de modelos de cálculo pode ser longo em relação ao cronograma do projeto, principalmente se houver a necessidade de se alterar a geometria inicial ou se existirem várias condições de carregamento a serem analisadas. A utilização de um pré-processador permite que várias alternativas sejam analisadas para escolha da que melhor atenda aos requisitos de projeto e de custo. Assim como ocorre com outras estruturas de engenharia, não existe uma equação de dimensionamento, mas sim de verificação; as estruturas ótimas são procuradas por tentativa e erro com base na experiência do projetista. Para facilitar a busca de uma estrutura ótima, implementou-se também uma rotina para otimizar as estruturas metálicas do pórtico. Adotou-se como função objetivo nesta implementação a minimização da massa, o que no caso dos pórticos implica em redução da área da seção transversal das vigas. Como restrições a esta redução adotaram-se o limite de escoamento do material e limite de estabilidade da estrutura (flambagem). Foram utilizados os aplicativos Excel (Microsoft), Ansys (Ansys Inc.) e Mathcad (Mathsoft) de maneira integrada a fim de se obter uma interface amigável, uma análise estrutural confiável e a elaboração automática de um memorial de cálculo. / The objective of this work is to develop a tool to generate an automatic structural design of gantry cranes. With an automatic generation of finite element models and also a design report, this routine allows a fast verification against yield of material and structural instability. The use of the finite element method was chosen for the structural design because it is a modern analysis tool that permits the evaluation of geometric configurations for which there are no analytical formulations available. However, the time necessary to build these models may be high, especially if there are changes in the initial geometry and many load cases. The use of a pre-processor allows the evaluation of a series of geometric alternatives, within which would be chosen the one with the lowest cost that attends the client’s specifications. Like many other engineering problems, there are no direct equations to find an adequate structure; there are only verification procedures available; the optimum structures are searched by trial and error, based on the designers’ experience. To accelerate this search process, an optimization routine was developed. Mass reduction was adopted as the objective function, which leads to reduction of the cross section area of the beams. Yield strength and buckling were adopted as restrictions to this optimization. Excel (Microsoft), Ansys (Ansys Inc) and Mathcad (Mathsoft) software were integrated to provide an user-friendly interface, reliable structural analysis and an automatic report generation.