Dans cette thèse différents outils issus de l'automatique non linéaire ont été mis en œuvre et ont permis d'apporter une première solution au problème de stabilité large signal des dispositifs électriques. A l'aide de modèles flous de type Takagi-Sugeno, on a montré qu'il était possible de résoudre le problème de stabilité dans le cas de deux applications électrotechniques à savoir un hacheur contrôlé en tension et l'alimentation par l'intermédiaire un filtre d'entrée d'un dispositif électrique fonctionnant à puissance constante. Dans le cas du hacheur, la taille estimée des bassins d'attraction reste modeste. Les raisons essentielles à l'échec obtenu dans la recherche de bassin de grande taille peut résulter dans le fait que d'une part , la mise sous forme TS du système n'est pas unique et que d'autre part les matrices du sous modèle TS du système ne sont de Hurwitz que dans une gamme très restreinte de variations du rapport cyclique. Dans le cas de l'alimentation par l'intermédiaire d'un filtre d'entrée d'un dispositif fonctionnant à puissance constante, on a montré que l'utilisation d'un modèle flou de type Takagi-Sugeno permettait d'exhiber un domaine d'attraction de taille significative. On a fourni des outils permettant de borner la plage de variations des pôles du système dans un domaine donné de l'espace d'état, domaine dans lequel la stabilité du modèle TS est prouvée. L'utilisation de la D-stabilité permet de connaitre les dynamiques maximales du système. La notion de stabilité exponentielle permet de connaître les dynamiques minimales du système. L'approche utilisée pour prouver la stabilité du système en présence de variations paramétriques, pour les deux systèmes étudiés, n'autorise que des variations extrêmement faibles de la valeur du paramètre autour de sa valeur nominale / In this thesis, various tools resulting from the nonlinear automatic were implemented and made it possible to bring a first solution to the problem of large signal stability of the electric systems. Using Takagi-Sugeno fuzzy models, one showed that it was possible to in the case of solve the problem of stability two electrotechnical applications to knowing a Boost converter controlled in tension and an electric system constituted by an input filter connected to an actuator functioning at constant power. In the case of the Boost converter, the estimated size of attraction domain remains modest. The reasons essential with the failure obtained in the search for domain of big size can result in the fact that on the one hand, the setting TS fuzzy models of the system is not single and that on the other hand the matrices of local model of TS model of the system are of Hurwitz only in one very restricted range of variations of the cyclic ratio. In the case of the electric system via a filter of entry of a functioning device at constant power, one showed that the use of a Takagi-Sugeno fuzzy model allowed exhibit a attraction domain of significant size. One provided tools allowing to limit the variations of the poles of the system in a given field of the state space, domain in which the stability of model TS is proven. The use of D-stability makes it possible to know dynamic maximum system. The concept of exponential stability makes it possible to know dynamic minimal system. The approach used to prove the stability of the system in the presence of parametric variations, for the two studied systems, authorizes only extremely weak variations of the value of the parameter around its maximal value

Made available in DSpace on 2016-08-17T14:53:17Z (GMT). No. of bitstreams: 1 Freud Sebastian Bach Carvalho Lima.pdf: 1668786 bytes, checksum: 953b0905b2e9bcd1cefb5702b8c6b362 (MD5) Previous issue date: 2011-07-29 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / In activated sludge systems, the dissolved oxygen is used by microorganisms to decompose organic matter and treatment of wastewater. In these systems the dynamics of dissolved oxygen has been seen as the main source of information to obtain online information about the treatment process. Thus, the development of an appropriate model for the dynamics of dissolved oxygen and estimation of the parameters of this model can improve the quality of the estimates of process parameters and, consequently, the measurement system. In this study, a model of the bench-scale reactor is developed considering the dynamics associated with the operation of the aeration system and the dissolved oxygen sensor. The aeration system consists of air pumps with two operating states: on or off, and dissolved oxygen measurement is made by electrochemical cell based on Clark. Approach was used identification systems in continuous time and state variable filters with the least squares estimator for estimating the parameters of the model developed. Simulations and experimental results, using a bench scale reactor developed in the laboratory, are presented to illustrate the proposed model. / Em sistemas de lodo ativado, o oxigênio dissolvido é utilizado por microorganismos para decomposição de matéria orgânica e tratamento de água residuárias. Nestes sistemas, a dinâmica de oxigênio dissolvido tem sido vista como a principal fonte de informação para obtenção de informações online sobre o processo de tratamento. Com isso, o desenvolvimento de um modelo apropriado para dinâmica do oxigênio dissolvido e a estimação dos parâmetros deste modelo, pode melhorar a qualidade das estimativas dos parâmetros do processo e, consequentemente, do sistema de medição. No presente trabalho, um modelo do reator em escala de bancada é desenvolvido considerando as dinâmicas relacionadas com a operação do sistema de aeração e com o sensor de oxigênio dissolvido. O sistema de aeração é composto por bombas de ar com dois estados de operação: ligada ou desligada, e a medição de oxigênio dissolvido é realizada por sensor eletroquímico baseado em célula de Clark. Usa-se abordagem de identificação de sistemas em tempo contínuo e filtros de variável de estado junto com o estimador dos mínimos quadrados para estimação dos parâmetros do modelo desenvolvido. Simulações e resultados experimentais, utilizando um reator em escala de bancada desenvolvido no laboratório, são apresentados para ilustrar o modelo proposto.

Dans le cadre de la protection du bâti face au risque sismique, les techniques d’analyse simplifiées, basées sur des calculs quasi-statiques en poussée progressive, se sont fortement développées au cours des deux dernières décennies. Le travail de thèse a pour objectif d’optimiser une stratégie d’analyse simplifiée proposée par Chopra et al. (2001) et adoptée par les normes américaines FEMA 273. Il s’agit d’une analyse modale non linéaire découplée, dénommée par les auteurs UMRHA qui se caractérisent principalement par : des calculs de type pushover selon les modes de vibration dominants de la structure, la création de modèles à un degré de liberté non linéaire à partir des courbes de pushover, puis le calcul de la réponse temporelle de la structure en recombinant les réponses temporelles associées à chaque mode de vibration. Dans ce travail, la méthode UMRHA a été améliorée en investiguant les points suivants. Tout d’abord, plusieurs modèles à un degré de liberté non linéaire déduits des courbes de pushover modal sont proposés afin d’enrichir la méthode UMRHA originelle qui emploie un simple modèle élasto-plastique : autres modèles élasto-plastiques avec des courbes enveloppes différentes, le modèle de Takeda prenant en compte un comportement hystérétique propre aux structures sous séismes, et enfin, un modèle simplifié basé sur la dégradation de fréquence en fonction d’un indicateur de dommage. Ce dernier modèle à un degré de liberté privilégie la vision de la chute de fréquence au cours du processus d’endommagement de la structure par rapport à une description réaliste des boucles d’hystérésis. La réponse totale de la structure est obtenue en sommant les contributions non linéaires des modes dominants aux contributions linéaires des modes non dominants. Enfin, la dégradation des déformées modales, due à l’endommagement subi par la structure au cours de la sollicitation sismique, est prise en compte dans la méthode M-UMRHA proposée dans ce travail, en généralisant le concept précédent de dégradation des fréquences modales en fonction d’un indicateur de dommage : la déformée modale devient elle-aussi dépendante d’un indicateur de dommage, le déplacement maximum en tête de l’ouvrage ; l’évolution de la déformée modale en fonction de cet indicateur est directement identifiée à partir des calculs de pushover modal. La pertinence de la nouvelle méthode M-UMRHA est investiguée pour plusieurs types de structures, en adoptant des modélisations éprouvées dans le cadre de la simulation des structures sous séismes : portique en béton armé modélisé par des éléments multifibres pour le béton et les armatures, remplissage en maçonnerie avec des éléments barres diagonales résistant uniquement en compression, bâti existant contreventé (Hôtel de Ville de Grenoble) avec des approches coques multicouches. Les résultats obtenus par la méthode simplifiée proposée sont comparés aux résultats de référence issus de l'analyse temporelle non linéaire dynamique. / In the context of building’s protection against seismic risk, simplified analysis techniques, based on quasi-static analysis of pushover, have strongly developed over the past two decades. The thesis aims to optimize a simplified method proposed by Chopra and Goel in 2001 and adopted by American standards FEMA 273. This method is a nonlinear decoupled modal analysis, called by the authors UMRHA (Uncoupled Modal for Response History Analysis) which is mainly characterized by: pushover modal analysis according to the dominant modes of vibration of the structure, setting up nonlinear single degree of freedom systems drawn from modal pushover curves, then determining the history response of the structure by combining of the temporal responses associated with each mode of vibration. The decoupling of nonlinear history responses associated with each mode is the strong assumption of the method UMRHA. In this study, the UMRHA method has been improved by investigating the following points. First of all, several nonlinear single degree of freedom systems drawn from modal pushover curves are proposed to enrich the original UMRHA method, in which a simple elastic-plastic model is used, other elastic-plastic models with different envelope curves, Takeda model taking into account an hysteretic behavior characteristic of structures under earthquakes, and finally, a simplified model based on the frequency degradation as a function of a damage index. The latter nonlinear single degree of freedom model privileges the view of the frequency degradation during the structure damage process relative to a realistic description of hysteresis loops. The total response of the structure is obtained by summing the contributions of the non linear dominant modes to those of linear non dominant modes. Finally, the degradation of the modal shapes due to the structure damage during the seismic loading is taken into account in the new simplified method M-UMRHA (Modified UMRHA) proposed in this study. By generalizing the previous model of frequency degradation as a function of a damage index: the modal shape becomes itself also dependent on a damage index, the maximum displacement at the top of the structure; the evolution of the modal shape as a function of this index is directly obtained from the modal pushover analysis. The pertinence of the new method M-UMRHA is investigated for several types of structures, by adopting tested models of structures simulation under earthquakes: reinforced concrete frame modeled by multifibre elements with uniaxial laws under cyclic loading for concrete and steel, infill masonry wall with diagonal bars elements resistant only in compression, existing building (Grenoble City Hall) with multilayer shell elements and nonlinear biaxial laws based on the concept of smeared and fixed cracks. The obtained results by the proposed simplified method are compared to the reference results derived from the nonlinear response history analysis.

Ma thèse en automatique s’inscrit dans la thématique de recherche «Transport» du LAMIH. L’objectif est d’améliorer le fonctionnement des moteurs thermiques (essence), notamment en réduisant la consommation et la pollution. Face à cet enjeu écologique et économique, et compte tenu des nouvelles normes et des stratégies court-termistes de l’industrie (scandale Volkswagen...), de nouveaux contrôleurs doivent être conçus pour piloter l’arrivée d’air et d’essence au sein du moteur. En considérant l’aspect hautement non-linéaire du système, la représentation Takagi-Sugeno et le background théorique du LAMIH sont utilisés. Un premier contrôleur est synthétisé pour régler le problème de la vitesse de ralenti du moteur. Cependant, la complexité du système impose l’utilisation d’un contrôleur très coûteux d’un point de vue computationnel. Un contrôleur alternatif est donc synthétisé afin d’être implémenté dans l’ordinateur embarqué du moteur. Un second contrôleur est obtenu pour maintenir la richesse en proportions stoechiométriques afin de réduire la pollution. Ce système étant sujet à un retard de transport variable, un changement de domaine est réalisé afin de rendre ce retard constant, et de concevoir un contrôleur simple et efficace. Des essais réels sur le banc d’essai moteur du LAMIH sont réalisés afin de valider la méthodologie présentée. / My PhD in Automatic Control is part of the research theme “Transport” of the LAMIH. The objective is to improve the functioning of the gasoline engines, mainly by reducing the fuel consumption and the pollution. With this ecologic and economic challenge, and taking into account the new norms and the short-term strategies of the industry (scandal of Volkswagen...), new controllers have to be designed to control the air valve and the fuel injection inside the engine. Considering the highly nonlinear aspect of the system, the Takagi-Sugeno representation and the theoretical background of the LAMIH have been used. A first controller is designed to solve the problem of idle engine speed. However, the complexity of the system forces the use of a controller that is very costly from a computational point of view. An alternative controller is then designed in order to be implemented inside the embedded computer of the engine. A second controller is obtained to maintain the air-fuel ratio in stoichiometric proportions in order to reduce the pollution. This system being subject to a variable transport delay, a change of domain is realized to make this delay constant, and to design a simple and efficient controller. Real-time experiments have been realized on the engine test bench of the LAMIH in order to validate the presented methodology.

The estimation problem of stochastic nonlinear parametric models is recognized to be very challenging due to the intractability of the likelihood function. Recently, several methods have been developed to approximate the maximum likelihood estimator and the optimal mean-square error predictor using Monte Carlo methods. Albeit asymptotically optimal, these methods come with several computational challenges and fundamental limitations. The contributions of this thesis can be divided into two main parts. In the first part, approximate solutions to the maximum likelihood problem are explored. Both analytical and numerical approaches, based on the expectation-maximization algorithm and the quasi-Newton algorithm, are considered. While analytic approximations are difficult to analyze, asymptotic guarantees can be established for methods based on Monte Carlo approximations. Yet, Monte Carlo methods come with their own computational difficulties; sampling in high-dimensional spaces requires an efficient proposal distribution to reduce the number of required samples to a reasonable value. In the second part, relatively simple prediction error method estimators are proposed. They are based on non-stationary one-step ahead predictors which are linear in the observed outputs, but are nonlinear in the (assumed known) input. These predictors rely only on the first two moments of the model and the computation of the likelihood function is not required. Consequently, the resulting estimators are defined via analytically tractable objective functions in several relevant cases. It is shown that, under mild assumptions, the estimators are consistent and asymptotically normal. In cases where the first two moments are analytically intractable due to the complexity of the model, it is possible to resort to vanilla Monte Carlo approximations. Several numerical examples demonstrate a good performance of the suggested estimators in several cases that are usually considered challenging. / <p>QC 20171128</p>

科学研究費補助金 研究種目:基盤研究(C) 課題番号:10650238 研究代表者:安田 仁彦 研究期間:1998-1999年度