Novel methods of system identification are developed in this dissertation. First set of methods are designed to realize time varying linear dynamical system models from input-output experimental data. The preliminary results obtained in a recent paper by the author are extended to establish a new algorithm called the Time Varying Eigensystem Realization Algorithm (TVERA). The central aim of this algorithm is to obtain a linear, time varying, discrete time model sequence of the dynamic system directly from the input-output data. Important results relating to concepts concerning coordinate systems for linear time varying systems are developed (discrete time theory) and an intuitive understanding of equivalent realizations is provided. A procedure to develop first few time step models is detailed, providing a unified solution to the time varying identification problem. The practical problem of identifying the time varying generalized Markov parameters required for TVERA is presented as the next result. In the process, we generalize the classical time invariant input output AutoRegressive model with an eXogenous input (ARX) models to the time varying case and realize an asymptotically stable observer as a byproduct of the calculations. It is further found that the choice of the generalized time varying ARX model (GTV-ARX) can be set to realize a time varying dead beat observer. Methods to use the developed algorithm(s) in this research are then considered for application to the identification of system models that are bilinear in nature. The fact that bilinear plant models become linear for constant inputs is used in the development of an algorithm that generalizes the classical developments of Juang. An intercept problem is considered as a candidate for application of the time varying identification scheme, where departure motion dynamics model sequence is calculated about a nominal trajectory with suboptimal performance owing to the presence of unstructured perturbations. Control application is subsequently demonstrated. The dynamics of a particle in a rotating tube is considered next for identification using the time varying eigensystem realization algorithm. Continuous time bilinear system identification method is demonstrated using the particle example and the identification of an automobile brake model.

Esse trabalho compara análises modais, obtidas por dois métodos distintos, de propriedades de imagens de chamas de um forno industrial. Visto que a combustão de gases e líquidos é caracterizada pelo tipo de injeção de combustível, pela formação aleatória de vórtices e pela distribuição de reagentes, o exato sinal de excitação do sistema é, na prática, muito pouco conhecido, cabendo aos métodos de análise a identificação dos parâmetros modais sem essa informação. Para vencer este obstáculo, na presente dissertação foram utilizados o método de Ibrahim no Domínio do Tempo (ITDM) e o método Eigensystem Realization Algorithm (ERA) associados à técnica do decremento aleatório (Random Decrement - RandomDec). A técnica RandomDec permite a obtenção de um sinal proporcional ao decaimento livre do sistema, a partir de uma excitação aleatória. Essa proporcionalidade permite que os métodos no domínio do tempo citados anteriormente sejam empregados na obtenção das características de vibração do sistema. O estudo da dinâmica de chamas com base em propriedades de suas imagens é tema muito pouco abordado na literatura; por esse motivo, inicialmente foi realizado, por meio de simulações numéricas em um sistema hipotético, um completo teste de sensibilidade de todos os métodos e da técnica a serem implementados no problema real do forno industrial. Posteriormente, os dados de chamas foram utilizados para obtenção dos modos de vibrar das propriedades de imagens. Os resultados mostram que a modelagem realizada pelo ERA para sistema SIMO (single input and multiple output) fornece modos de vibrar com amortecimentos mais condizentes com a dinâmica da chama e, assim, sugerem que, em princípio, está técnica poderia ser utilizada prontamente na identificação desse tipo de sistema, sem a necessidade da realização de qualquer outra análise modal para corroborar os resultados. Além disso, a identificação de um sistema complexo como o analisado, utilizando métodos de análise modal operacional, enseja a possibilidade de pesquisas visando obter sistemas de controle baseado em imagens para plantas com características semelhantes. / In this present study, two different modal analysis methods are applied to obtain and compare models for properties of flame images captured in an industrial furnace. Since combustion of gases and liquids is characterized by the type of fuel injection, random formation of vortices and distribution of reagents, the system excitation signal isn\'t precisely known in the reality, thus the analysis methods must identify the modal parameters without this information. To overcome this obstacle, Ibrahim Time Domain Method (ITDM) and Eigensystem Realization Algorithm (ERA) associated with the Random Decrement (RandomDec) technique were implemented in this thesis. The RandomDec technique give the means to obtain a signal which is proportional to the free decay of the system when it is randomly forced. This proportional signal allows the previously mentioned analysis methods to be used to identify vibration characteristics of the system. The study of flame dynamics based on its image characteristics is barely approached in the literature; for this reason, a complete sensitivity test of all the methods and techniques to be implemented in the real problem of industrial furnaces was initially performed through numerical simulations of a hypothetical system. Afterwards, flame data was used to obtain the vibration modes of image properties. The results lead to conclude that the SIMO (single input and multiple output) model obtained by ERA has vibration modes with more consistent damping regarding flame dynamics and suggest also that this technique could be readily used in the identification of this kind of system, without the need of any other modal analysis for results corroboration. Furthermore, the utilization of operational modal analysis methods to identify a complex system such as the analyzed one may instigate researches on image-based control systems of industrial plants with suchlike characteristics.

Esse trabalho compara análises modais, obtidas por dois métodos distintos, de propriedades de imagens de chamas de um forno industrial. Visto que a combustão de gases e líquidos é caracterizada pelo tipo de injeção de combustível, pela formação aleatória de vórtices e pela distribuição de reagentes, o exato sinal de excitação do sistema é, na prática, muito pouco conhecido, cabendo aos métodos de análise a identificação dos parâmetros modais sem essa informação. Para vencer este obstáculo, na presente dissertação foram utilizados o método de Ibrahim no Domínio do Tempo (ITDM) e o método Eigensystem Realization Algorithm (ERA) associados à técnica do decremento aleatório (Random Decrement - RandomDec). A técnica RandomDec permite a obtenção de um sinal proporcional ao decaimento livre do sistema, a partir de uma excitação aleatória. Essa proporcionalidade permite que os métodos no domínio do tempo citados anteriormente sejam empregados na obtenção das características de vibração do sistema. O estudo da dinâmica de chamas com base em propriedades de suas imagens é tema muito pouco abordado na literatura; por esse motivo, inicialmente foi realizado, por meio de simulações numéricas em um sistema hipotético, um completo teste de sensibilidade de todos os métodos e da técnica a serem implementados no problema real do forno industrial. Posteriormente, os dados de chamas foram utilizados para obtenção dos modos de vibrar das propriedades de imagens. Os resultados mostram que a modelagem realizada pelo ERA para sistema SIMO (single input and multiple output) fornece modos de vibrar com amortecimentos mais condizentes com a dinâmica da chama e, assim, sugerem que, em princípio, está técnica poderia ser utilizada prontamente na identificação desse tipo de sistema, sem a necessidade da realização de qualquer outra análise modal para corroborar os resultados. Além disso, a identificação de um sistema complexo como o analisado, utilizando métodos de análise modal operacional, enseja a possibilidade de pesquisas visando obter sistemas de controle baseado em imagens para plantas com características semelhantes. / In this present study, two different modal analysis methods are applied to obtain and compare models for properties of flame images captured in an industrial furnace. Since combustion of gases and liquids is characterized by the type of fuel injection, random formation of vortices and distribution of reagents, the system excitation signal isn\'t precisely known in the reality, thus the analysis methods must identify the modal parameters without this information. To overcome this obstacle, Ibrahim Time Domain Method (ITDM) and Eigensystem Realization Algorithm (ERA) associated with the Random Decrement (RandomDec) technique were implemented in this thesis. The RandomDec technique give the means to obtain a signal which is proportional to the free decay of the system when it is randomly forced. This proportional signal allows the previously mentioned analysis methods to be used to identify vibration characteristics of the system. The study of flame dynamics based on its image characteristics is barely approached in the literature; for this reason, a complete sensitivity test of all the methods and techniques to be implemented in the real problem of industrial furnaces was initially performed through numerical simulations of a hypothetical system. Afterwards, flame data was used to obtain the vibration modes of image properties. The results lead to conclude that the SIMO (single input and multiple output) model obtained by ERA has vibration modes with more consistent damping regarding flame dynamics and suggest also that this technique could be readily used in the identification of this kind of system, without the need of any other modal analysis for results corroboration. Furthermore, the utilization of operational modal analysis methods to identify a complex system such as the analyzed one may instigate researches on image-based control systems of industrial plants with suchlike characteristics.

Les écoulements impactants auto-oscillants sont sources de bruit intenses dans diverses applications. La cavité représente une configuration privilégiée pour les étudier. L’objectif de la thèse est de fournir une solution concrète aux problèmes d’oscillation de la couche cisaillée de l’écoulement de cavité, en exploitant des outils issus de la communauté des automaticiens. Un rappel des travaux menés ces dernières années est présenté ainsi que les difficultés rencontrées pour le contrôle d’un tel dispositif. La principale source de complexité est les équations de Navier- Stockes qui régissent sont fonctionnement. Il est actuellement impossible d’exploiter directement ce type d’équations dans la synthèse de correcteur. Outre cette difficulté, les oscillations de la couche limite sont qualitativement induites par une instabilité non-linéaire soumise à saturation. Par conséquent, l’identification empirique basée sur la réponse fréquentielle ne peut être appliquée, sauf si un correcteur stabilisant le point fixe est trouvé, de manière à préserver le principe de superposition lors de l’identification. Ainsi, comme première solution, une commande empirique en boucle fermée est obtenue à partir de la pression retardée. Un travail sur un choix optimal des paramètres est réalisé permettant ainsi l’élimination des oscillations de la couche cisaillée. De plus, ce contrôle se montre robuste face aux changements de configuration de cavité. Une variante de cette loi de commande aboutit à un signal de contrôle évanescent, induisant ainsi un gain considérable en terme de consommation énergétique. Malheureusement, une telle loi commande ne permet pas d’identifier, ni d’expliquer les phénomènes dynamiques associés à un tel bouclage. Par conséquent, une approche linéaire est utilisée pour identifier les dynamiques de l’écoulement de cavité. La méthode d’identification ERA (Eigensystem Realization Algorithm) en boucle fermée est utilisée pour extraire cette dynamique. Les modèles identifiés nous informent sur les modes responsables de l’apparition des oscillations de la couche cisaillée. En outre, nous avons une information précise sur le phénomène de commutation de mode observé dans plusieurs régimes d’écoulement. Le modèle permet d’extraire les modes instables de la dynamique linéaire de la cavité et les associer aux oscillations du régime saturé. Pour valider la procédure linéaire, un correcteur optimal a été synthétisé à partir du modèle identifié. Cette commande se révèle robuste aux variations paramétriques du modèle. / The cavity flow represents a preferred configuration to study the self-oscillating impingement flows, which constitute an intense noise source in various applications (rail, aviation, etc.). Several studies have been conducted in order to mitigate these oscillations by open or closed-loop control strategies. A reminder of undertaken studies in recent years is presented as well as the encountered difficulties to control of such a device. The main difficulty to control such flows is that their state results from the nonlinear saturation of instabilities by developing on the steady base flow (stable limit cycle against unstable fixed point). The identification of a model is not easy in such a situation. Indeed, the empirical identification based on the frequency response can't be applied, unless a controller stabilizing the unstable fixed point is previously given, in order to preserve the superposition principle during the identification.Our study is mainly based on a closed-loop control because it brings a better energetic gain. We show that a simple time-delayed feedback control is able to stabilize the base flow and robust to changes in cavity configuration. To extract the responsible dynamics of the oscillations appearance in the shear layer, ERA (Eigensystem Realization Algorithm) identification technique is used in closed-loop. The model identifies the unstable modes of the linear dynamics of the cavity as those actually responsible for the oscillation frequencies of the saturated regime. An optimal controller was synthesized from this dynamics. This controller reveals robust to parametric variations of the model.

No description available.

Apresenta uma nova técnica de identificação estocástica para estruturas sujeitas a excitação ambiente utilizando um método não recursivo, a análise estatística e a transformada wavelet. A análise estatística contribui para a seleção da melhor ordem para o sistema e automação dos procedimentos computacionais. Em geral há dificuldade para a identificação de baixas freqüências. A transformada wavelet é uma ferramenta fundamental para a compressão de dados e torna possível a identificação completa incluindo baixas freqüências e modos acoplados, além de melhorar significativamente a eficiência computacional do método. É apresentado o estudo de três sistemas simulados e os resultados obtidos são comparados aos parâmetros modais teóricos. / A new form to carry out stochastic identification of structures in operational conditions using a non recursive method, the statistic analysis and the wavelet transform, is presented. The statistic analysis contributed to select the best system order and to automation of computational procedures. In general the identification of low frequencies is a difficult task. The wavelet transform is an essential tool for compression of data making possible the complete identification including low frequencies. In addition it improves the computational efficiency. The study of three simulated system is presented and the results are compared with the analytical modal parameters.

Apresenta uma nova técnica de identificação estocástica para estruturas sujeitas a excitação ambiente utilizando um método não recursivo, a análise estatística e a transformada wavelet. A análise estatística contribui para a seleção da melhor ordem para o sistema e automação dos procedimentos computacionais. Em geral há dificuldade para a identificação de baixas freqüências. A transformada wavelet é uma ferramenta fundamental para a compressão de dados e torna possível a identificação completa incluindo baixas freqüências e modos acoplados, além de melhorar significativamente a eficiência computacional do método. É apresentado o estudo de três sistemas simulados e os resultados obtidos são comparados aos parâmetros modais teóricos. / A new form to carry out stochastic identification of structures in operational conditions using a non recursive method, the statistic analysis and the wavelet transform, is presented. The statistic analysis contributed to select the best system order and to automation of computational procedures. In general the identification of low frequencies is a difficult task. The wavelet transform is an essential tool for compression of data making possible the complete identification including low frequencies. In addition it improves the computational efficiency. The study of three simulated system is presented and the results are compared with the analytical modal parameters.

The aim of this diploma thesis is a creation of modal parameter estimation application. Modal properties (natural frequencies, damping factors and mode shapes) are used in many dynamics analysis and their accurate determination is very important therefore the modal parameter estimation is one of the most significant part of the experimental modal analysis. Many methods have been developed for modal parameter estimation, each of them with different assumptions and with different accuracy. In the beginning of this thesis, a theory connected with modal analysis and a theory which is necessary for understanding to presented modal parameter methods are given. Then four different modal parameter estimation methods are presented - Peak Picking, Circle Fit, Least Square method and Eigensystem Realization Algorithm. The application for the modal parameter estimation is the output of this diploma thesis. In addition, the application allows performing all experimental modal analysis such as estimation of frequency response functions, animation of the found mode shapes, different kinds of comparison etc. In the conclusion, three structures are shown on which the application and modal parameter estimation methods were tested.

No description available.