The masters thesis was focused on design and realization power converter which should worked as welding machine. It was necessary choose suitable type converter. Next, it was needs design of pulsed transformer, output filter choke and design of semiconductors which were in power part welding machine. It was also designed control and driver circuits. This parts which were used on construct the converter were made and tested too.

This thesis tries to research the term “robust controllers”. Its aim is to compare the robustness of discrete PID controllers (Discrete Equivalent Continuous Controller, Discrete Impulse Area Invariant, Takahashi, Feed-Forward), adaptive discrete PID controllers (Discrete Impulse Area Invariant, Takahashi, Feed-Forward), optimal controllers (quadratic optimal), and adaptive optimal controllers (quadratic optimal) on chosen transfer functions. Its aim is also to check the influence of A/D and D/A converters. The aims to obtain are demarked at the beginning of the text and also there is an explanation of the term “robustness.” Later on there is a description and an approximation to each of the chosen kinds of controllers and the identification methods used in the thesis (for adaptive controllers the method of recursive least-squares was used). The Kharitonov's Theorem are made on the chosen transfer function. Next there is a description of the methods with which the robustness of the controllers will be tested. The first method is the integral criteria, particular ITAE criterion and quadratic criterion. The second one is the analysis of the generalised circle criterion. Furthermore there are various displays of the results obtained and their corresponding comments. The results obtained are graphically displayed and by means of these schemes the particular types of controllers are compared. All of the simulations and results obtained were acquired through the use of the program MATLAB- Simulink. In the end of the thesis there is an overall evaluation.

Na literatura há vários trabalhos elucidando a importância de ferramentas para auditar malhas de controle. Na indústria, o interesse por esse tipo de software é crescente, devido ao benefício trazido: garantir o exato desempenho para cada controlador significa atingir pontos de operação mais lucrativos. Todavia, as metodologias disponíveis na literatura e nas ferramentas comerciais não provêm medidas conclusivas das características da malha. Conseqüentemente, a análise muitas vezes é confusa e difícil, acarretando um diagnóstico difícil. Reduzir a distância entre a auditoria e o diagnóstico, propondo novas métricas para ajudar o engenheiro no gerenciamento de malhas é o foco central deste trabalho. A presente tese é segmentada em duas partes: auditoria e diagnóstico, dentro das quais se inserem as principais contribuições deste trabalho. A primeira contribuição deste trabalho é a proposição de uma metodologia para decompor o impacto da velocidade do controlador, tempo morto e ruído branco sobre a variância total da malha, auxiliando o engenheiro de processos a tomar uma ação: trocar os parâmetros de ajuste, aumentar a ordem do controlador, substituir o medidor, reduzir o tempo morto, entre outros. O método proposto requer apenas dados de operação normal e o tempo morto da malha, não sendo necessários testes intrusivos, sendo possível a aplicação industrial. O conjunto de métricas propostas foi aplicado em três casos de estudo, fornecendo resultados promissores. Dentro do campo de auditoria, se insere a segunda contribuição: a proposição de uma metodologia para estimação de parâmetros conclusivos (ou determinísticos) para avaliação da performance e robustez de controladores (Máxima Sensibilidade, razão entre o tempo de subida de malha aberta e fechada, razão entre o tempo de assentamento de malha aberta e fechada, entre outros), baseado em índices estocásticos (i.e. métricas que podem ser computadas em tempo real, sem testes intrusivos) e parâmetros da malha (tempo morto, constante de tempo). O modelo de inferência para desempenho e robustez, chamado PRIM, é proposto. Este modelo fornece uma clara indicação do real desempenho e robustez da malha, facilitando a análise e tornando o diagnóstico direto. Além disso, visando elucidar a vantagem dos índices determinísticos sobre os estocásticos, ambos serão aplicados a um conjunto de casos de estudo. A última contribuição no campo de auditoria está na estimação do potencial impacto econômico de cada malha, como ferramenta para hierarquizar sua manutenção. Considerando que uma típica planta química possui centenas ou milhares de malhas e que a maioria delas possui significativo potencial de melhora, a priorização da manutenção é essencial. O impacto econômico é baseado no conceito de Matriz de Variabilidade (VM), que é uma matriz que mostra o impacto da melhora da performance de um controlador sobre a variabilidade de toda a planta. A segunda parte deste trabalho aborda o diagnóstico de algumas avarias comuns em controladores. A primeira metodologia visa quantificar a banda de agarramento em válvulas de controle, que apresentam este fenômeno, através de um modelo de inferência chamado SIM. A metodologia proposta requer apenas dados de operação normal do controlador e planta. A contribuição final deste trabalho é um método simples para avaliação do modelo em controladores preditivos MPCs, baseado na Análise de Componentes Independentes (ICA). Esta análise é bastante útil para indicar os canais que possuem modelo pobre em relação à planta. O trabalho finaliza com as conclusões finais e os trabalhos futuros relativos a esta tese. / Many works available in the literature support the importance of control loop performance assessment (CLPA) tools. Industrially, there is an increasing interest in this area and the reason is trivial: ensuring the exact performance for each loop means allowing it to operate in a high profitable operating point. However, the available methodologies in commercial tools and literature do not provide clear and conclusive metrics of the actual loop performance. Consequently, the diagnostics is not easy, making the analysis sometimes difficult and confusing. The aim of this thesis is to reduce the gap between assessment and diagnostics by proposing a set of metrics to help the engineer in control loop performance management. The thesis is divided into two sections: Assessment and Diagnostic. The first contribution of this work is the proposition of a new methodology to decompose the impact of control loop performance, time delay, and white noise in the total control loop variance, helping the engineer to diagnose the loop performance problem and take the right action to achieve the desired product variability (by changing tuning parameters, changing controller type, replacing instrument, or changing the process, among others). The proposed method does not require any invasive tests, only control loop routine operating data and process time delay, allowing the industrial application of the proposed indices in real time. The methodology was applied in three case studies, providing very good results. The second contribution is propose a methodology to estimate conclusive indices (also called deterministic indices) to evaluate loop performance and robustness (maximal sensitivity, ratio between open and closed loop rise time, settling time, among others), based on stochastic indices (i.e. indices that can be computed using only normal operating data) and process parameters (time delay and time constant). A performance and robustness inferential model (PRIM) for loop assessment will be shown. The PRIM provides a clear picture of loop performance and robustness, making the analysis easier and the diagnostics straightforward. Moreover, this section highlights the limitation and drawbacks of stochastic metrics for CLPA. The last contribution of the Assessment section proposes a method to prioritize loop maintenance, based on the economic impact of each loop. Assuming that a typical process plant has hundreds or thousands of loops and most of them have significant impact over loop variability, prioritizing loop maintenance is essential. The economic impact is based on the concept of Variability Matrix (VM), which is an array that shows the impact of performance improvement of a given loop on the whole plant. The second section of this work is the Diagnostics of some controller problems. The first methodology aims to quantify the stickband in a sticky valve, using only normal operating data of controller output and process variable. An analogous inference model, called Stiction Inference Model (SIM), is proposed to estimate the stickband. In the final contribution, it is proposed a simple method to evaluate the Model Plant Mismatch (MPM) based on Independent Component Analysis (ICA). This analysis is very useful to point out the poor channels models in Model Predictive Controllers. The work ends with concluding remarks and further work.

Na literatura há vários trabalhos elucidando a importância de ferramentas para auditar malhas de controle. Na indústria, o interesse por esse tipo de software é crescente, devido ao benefício trazido: garantir o exato desempenho para cada controlador significa atingir pontos de operação mais lucrativos. Todavia, as metodologias disponíveis na literatura e nas ferramentas comerciais não provêm medidas conclusivas das características da malha. Conseqüentemente, a análise muitas vezes é confusa e difícil, acarretando um diagnóstico difícil. Reduzir a distância entre a auditoria e o diagnóstico, propondo novas métricas para ajudar o engenheiro no gerenciamento de malhas é o foco central deste trabalho. A presente tese é segmentada em duas partes: auditoria e diagnóstico, dentro das quais se inserem as principais contribuições deste trabalho. A primeira contribuição deste trabalho é a proposição de uma metodologia para decompor o impacto da velocidade do controlador, tempo morto e ruído branco sobre a variância total da malha, auxiliando o engenheiro de processos a tomar uma ação: trocar os parâmetros de ajuste, aumentar a ordem do controlador, substituir o medidor, reduzir o tempo morto, entre outros. O método proposto requer apenas dados de operação normal e o tempo morto da malha, não sendo necessários testes intrusivos, sendo possível a aplicação industrial. O conjunto de métricas propostas foi aplicado em três casos de estudo, fornecendo resultados promissores. Dentro do campo de auditoria, se insere a segunda contribuição: a proposição de uma metodologia para estimação de parâmetros conclusivos (ou determinísticos) para avaliação da performance e robustez de controladores (Máxima Sensibilidade, razão entre o tempo de subida de malha aberta e fechada, razão entre o tempo de assentamento de malha aberta e fechada, entre outros), baseado em índices estocásticos (i.e. métricas que podem ser computadas em tempo real, sem testes intrusivos) e parâmetros da malha (tempo morto, constante de tempo). O modelo de inferência para desempenho e robustez, chamado PRIM, é proposto. Este modelo fornece uma clara indicação do real desempenho e robustez da malha, facilitando a análise e tornando o diagnóstico direto. Além disso, visando elucidar a vantagem dos índices determinísticos sobre os estocásticos, ambos serão aplicados a um conjunto de casos de estudo. A última contribuição no campo de auditoria está na estimação do potencial impacto econômico de cada malha, como ferramenta para hierarquizar sua manutenção. Considerando que uma típica planta química possui centenas ou milhares de malhas e que a maioria delas possui significativo potencial de melhora, a priorização da manutenção é essencial. O impacto econômico é baseado no conceito de Matriz de Variabilidade (VM), que é uma matriz que mostra o impacto da melhora da performance de um controlador sobre a variabilidade de toda a planta. A segunda parte deste trabalho aborda o diagnóstico de algumas avarias comuns em controladores. A primeira metodologia visa quantificar a banda de agarramento em válvulas de controle, que apresentam este fenômeno, através de um modelo de inferência chamado SIM. A metodologia proposta requer apenas dados de operação normal do controlador e planta. A contribuição final deste trabalho é um método simples para avaliação do modelo em controladores preditivos MPCs, baseado na Análise de Componentes Independentes (ICA). Esta análise é bastante útil para indicar os canais que possuem modelo pobre em relação à planta. O trabalho finaliza com as conclusões finais e os trabalhos futuros relativos a esta tese. / Many works available in the literature support the importance of control loop performance assessment (CLPA) tools. Industrially, there is an increasing interest in this area and the reason is trivial: ensuring the exact performance for each loop means allowing it to operate in a high profitable operating point. However, the available methodologies in commercial tools and literature do not provide clear and conclusive metrics of the actual loop performance. Consequently, the diagnostics is not easy, making the analysis sometimes difficult and confusing. The aim of this thesis is to reduce the gap between assessment and diagnostics by proposing a set of metrics to help the engineer in control loop performance management. The thesis is divided into two sections: Assessment and Diagnostic. The first contribution of this work is the proposition of a new methodology to decompose the impact of control loop performance, time delay, and white noise in the total control loop variance, helping the engineer to diagnose the loop performance problem and take the right action to achieve the desired product variability (by changing tuning parameters, changing controller type, replacing instrument, or changing the process, among others). The proposed method does not require any invasive tests, only control loop routine operating data and process time delay, allowing the industrial application of the proposed indices in real time. The methodology was applied in three case studies, providing very good results. The second contribution is propose a methodology to estimate conclusive indices (also called deterministic indices) to evaluate loop performance and robustness (maximal sensitivity, ratio between open and closed loop rise time, settling time, among others), based on stochastic indices (i.e. indices that can be computed using only normal operating data) and process parameters (time delay and time constant). A performance and robustness inferential model (PRIM) for loop assessment will be shown. The PRIM provides a clear picture of loop performance and robustness, making the analysis easier and the diagnostics straightforward. Moreover, this section highlights the limitation and drawbacks of stochastic metrics for CLPA. The last contribution of the Assessment section proposes a method to prioritize loop maintenance, based on the economic impact of each loop. Assuming that a typical process plant has hundreds or thousands of loops and most of them have significant impact over loop variability, prioritizing loop maintenance is essential. The economic impact is based on the concept of Variability Matrix (VM), which is an array that shows the impact of performance improvement of a given loop on the whole plant. The second section of this work is the Diagnostics of some controller problems. The first methodology aims to quantify the stickband in a sticky valve, using only normal operating data of controller output and process variable. An analogous inference model, called Stiction Inference Model (SIM), is proposed to estimate the stickband. In the final contribution, it is proposed a simple method to evaluate the Model Plant Mismatch (MPM) based on Independent Component Analysis (ICA). This analysis is very useful to point out the poor channels models in Model Predictive Controllers. The work ends with concluding remarks and further work.

Na literatura há vários trabalhos elucidando a importância de ferramentas para auditar malhas de controle. Na indústria, o interesse por esse tipo de software é crescente, devido ao benefício trazido: garantir o exato desempenho para cada controlador significa atingir pontos de operação mais lucrativos. Todavia, as metodologias disponíveis na literatura e nas ferramentas comerciais não provêm medidas conclusivas das características da malha. Conseqüentemente, a análise muitas vezes é confusa e difícil, acarretando um diagnóstico difícil. Reduzir a distância entre a auditoria e o diagnóstico, propondo novas métricas para ajudar o engenheiro no gerenciamento de malhas é o foco central deste trabalho. A presente tese é segmentada em duas partes: auditoria e diagnóstico, dentro das quais se inserem as principais contribuições deste trabalho. A primeira contribuição deste trabalho é a proposição de uma metodologia para decompor o impacto da velocidade do controlador, tempo morto e ruído branco sobre a variância total da malha, auxiliando o engenheiro de processos a tomar uma ação: trocar os parâmetros de ajuste, aumentar a ordem do controlador, substituir o medidor, reduzir o tempo morto, entre outros. O método proposto requer apenas dados de operação normal e o tempo morto da malha, não sendo necessários testes intrusivos, sendo possível a aplicação industrial. O conjunto de métricas propostas foi aplicado em três casos de estudo, fornecendo resultados promissores. Dentro do campo de auditoria, se insere a segunda contribuição: a proposição de uma metodologia para estimação de parâmetros conclusivos (ou determinísticos) para avaliação da performance e robustez de controladores (Máxima Sensibilidade, razão entre o tempo de subida de malha aberta e fechada, razão entre o tempo de assentamento de malha aberta e fechada, entre outros), baseado em índices estocásticos (i.e. métricas que podem ser computadas em tempo real, sem testes intrusivos) e parâmetros da malha (tempo morto, constante de tempo). O modelo de inferência para desempenho e robustez, chamado PRIM, é proposto. Este modelo fornece uma clara indicação do real desempenho e robustez da malha, facilitando a análise e tornando o diagnóstico direto. Além disso, visando elucidar a vantagem dos índices determinísticos sobre os estocásticos, ambos serão aplicados a um conjunto de casos de estudo. A última contribuição no campo de auditoria está na estimação do potencial impacto econômico de cada malha, como ferramenta para hierarquizar sua manutenção. Considerando que uma típica planta química possui centenas ou milhares de malhas e que a maioria delas possui significativo potencial de melhora, a priorização da manutenção é essencial. O impacto econômico é baseado no conceito de Matriz de Variabilidade (VM), que é uma matriz que mostra o impacto da melhora da performance de um controlador sobre a variabilidade de toda a planta. A segunda parte deste trabalho aborda o diagnóstico de algumas avarias comuns em controladores. A primeira metodologia visa quantificar a banda de agarramento em válvulas de controle, que apresentam este fenômeno, através de um modelo de inferência chamado SIM. A metodologia proposta requer apenas dados de operação normal do controlador e planta. A contribuição final deste trabalho é um método simples para avaliação do modelo em controladores preditivos MPCs, baseado na Análise de Componentes Independentes (ICA). Esta análise é bastante útil para indicar os canais que possuem modelo pobre em relação à planta. O trabalho finaliza com as conclusões finais e os trabalhos futuros relativos a esta tese. / Many works available in the literature support the importance of control loop performance assessment (CLPA) tools. Industrially, there is an increasing interest in this area and the reason is trivial: ensuring the exact performance for each loop means allowing it to operate in a high profitable operating point. However, the available methodologies in commercial tools and literature do not provide clear and conclusive metrics of the actual loop performance. Consequently, the diagnostics is not easy, making the analysis sometimes difficult and confusing. The aim of this thesis is to reduce the gap between assessment and diagnostics by proposing a set of metrics to help the engineer in control loop performance management. The thesis is divided into two sections: Assessment and Diagnostic. The first contribution of this work is the proposition of a new methodology to decompose the impact of control loop performance, time delay, and white noise in the total control loop variance, helping the engineer to diagnose the loop performance problem and take the right action to achieve the desired product variability (by changing tuning parameters, changing controller type, replacing instrument, or changing the process, among others). The proposed method does not require any invasive tests, only control loop routine operating data and process time delay, allowing the industrial application of the proposed indices in real time. The methodology was applied in three case studies, providing very good results. The second contribution is propose a methodology to estimate conclusive indices (also called deterministic indices) to evaluate loop performance and robustness (maximal sensitivity, ratio between open and closed loop rise time, settling time, among others), based on stochastic indices (i.e. indices that can be computed using only normal operating data) and process parameters (time delay and time constant). A performance and robustness inferential model (PRIM) for loop assessment will be shown. The PRIM provides a clear picture of loop performance and robustness, making the analysis easier and the diagnostics straightforward. Moreover, this section highlights the limitation and drawbacks of stochastic metrics for CLPA. The last contribution of the Assessment section proposes a method to prioritize loop maintenance, based on the economic impact of each loop. Assuming that a typical process plant has hundreds or thousands of loops and most of them have significant impact over loop variability, prioritizing loop maintenance is essential. The economic impact is based on the concept of Variability Matrix (VM), which is an array that shows the impact of performance improvement of a given loop on the whole plant. The second section of this work is the Diagnostics of some controller problems. The first methodology aims to quantify the stickband in a sticky valve, using only normal operating data of controller output and process variable. An analogous inference model, called Stiction Inference Model (SIM), is proposed to estimate the stickband. In the final contribution, it is proposed a simple method to evaluate the Model Plant Mismatch (MPM) based on Independent Component Analysis (ICA). This analysis is very useful to point out the poor channels models in Model Predictive Controllers. The work ends with concluding remarks and further work.

Yes / In this paper, a performance enhancement scheme has been investigated for a class of stochastic nonlinear systems via set-point adjustment. Considering the practical industrial processes, the multi-layer systematic structure has been adopted to achieve the control design requirements subjected to random noise. The basic loop control is given by PID design while the parameters have been fixed after the design phase. Alternatively, we can consider that there exists an unadjustable loop control. Then, the additional loop is designed for performance enhancement in terms of the tracking accuracy. In particular, a novel approach has been presented to dynamically adjust the set-points using the estimated states of the systems through extended Kalman filter (EKF). Minimising the entropy criterion, the parameters of the set-point adjustment controller can be optimised which will enhance the performance of the entire closed-loop systems. Based upon the presented scheme, the stochastic stability analysis has been given to demonstrate that the closed-loop tracking errors are bounded in probability one. To indicate the effectiveness of the presented control scheme, the numerical examples have been given and the simulation results imply that the designed systems are bounded and the tracking performance can be enhanced simultaneously. In summary, a new framework for system performance enhancement has been presented even if the loop control is unadjustable which forms the main contribution of this paper.

In this thesis, we report the observations of optical squeezing from second harmonic generation (SHG), optical parametric oscillation (OPO) and optical parametric amplification (OPA). Demonstrations and proposals of applications involving the squeezed light and electro-optic control loops are presented. ¶ In our SHG setup, we report the observation of 2.1 dB of intensity squeezing on the second harmonic (SH) output. Investigations into the system show that the squeezing performance of a SHG system is critically affected by the pump noise and a modular theory of noise propagation is developed to describe and quantify this effect. Our experimental data has also shown that in a low-loss SHG system, intra-cavity nondegenerate OPO modes can simultaneously occur. This competition of nonlinear processes leads to the optical clamping of the SH output power and in general can degrade the SH squeezing. We model this competition and show that it imposes a limit to the observable SH squeezing. Proposals for minimizing the effect of competition are presented. ¶ In our OPO setup, we report the observation of 7.1 dB of vacuum squeezing and more than 4 dB of intensity squeezing when the OPO is operating as a parametric amplifier. We present the design criteria and discuss the limits to the observable squeezing from the OPO.We attribute the large amount of squeezing obtained in our experiment to the high escape efficiency of the OPO. The effect of phase jitter on the squeezing of the vacuum state is modeled. ¶ The quantum noise performance of an electro-optic feedforward control loop is investigated. With classical coherent inputs, we demonstrate that vacuum fluctuations introduced at the beam splitter of the control loop can be completely cancelled by an optimum amount of positive feedforward. The cancellation of vacuum fluctuations leads to the possibility of noiseless signal amplification with the feedforward loop. Comparison shows that the feedforward amplifier is superior or at least comparable in performance with other noiseless amplification schemes. When combined with an injection-locked non-planar ring Nd:YAG laser, we demonstrate that signal and power amplifications can both be noiseless and independently variable. ¶ Using squeezed inputs to the feedforward control loop, we demonstrate that information carrying squeezed states can be made robust to large downstream transmission losses via a noiseless signal amplification. We show that the combination of a squeezed vacuum meter input and a feedforward loop is a quantum nondemolition (QND) device, with the feedforward loop providing an additional improvement on the transfer of signal. In general, the use of a squeezed vacuum meter input and an electro-optic feedforward loop can provide pre- and post- enhancements to many existing QND schemes. ¶ Finally, we proposed that the quantum teleportation of a continuous-wave optical state can be achieved using a pair of phase and amplitude electro-optic feedforward loops with two orthogonal quadrature squeezed inputs. The signal transfer and quantum correlation of the teleported optical state are analysed. We show that a two dimensional diagram, similar to the QND figures of merits, can be used to quantify the performance of a teleporter.

This thesis proposes a method and develops a model for the participation of a combined cycle power plant to power system frequency control. Through the period of integration to the UCTE system, (Union for Coordination of Transmission of Electricity in Europe) frequency behavior of Turkey&rsquo / s grid and studies related to its improvement had been a great concern, so is the reason that main subject of my thesis became as &ldquo / Power System Frequency Control&rdquo / . Apart from system-wide global control action (secondary control) / load control loops at power plants, reserve power and its provision even at the minimum capacity generation stage, (primary control) are the fundamental concerns of this subject. The adjustment of proper amount of reserve at the power plants, and correct system response to any kind of disturbance, in the overall, are measured by the quality of the frequency behaviour of the system. A simulator that will simulate a dynamic gas turbine and its control system model, together with a combined cycle power plant load controller is the outcome of this thesis.

Submitted by Izabel Franco (izabel-franco@ufscar.br) on 2016-09-06T18:00:39Z No. of bitstreams: 1 DissMHP.pdf: 3604162 bytes, checksum: 5844b74d634a30ad629fc36c26706ee1 (MD5) / Approved for entry into archive by Marina Freitas (marinapf@ufscar.br) on 2016-09-12T13:56:27Z (GMT) No. of bitstreams: 1 DissMHP.pdf: 3604162 bytes, checksum: 5844b74d634a30ad629fc36c26706ee1 (MD5) / Approved for entry into archive by Marina Freitas (marinapf@ufscar.br) on 2016-09-12T13:56:40Z (GMT) No. of bitstreams: 1 DissMHP.pdf: 3604162 bytes, checksum: 5844b74d634a30ad629fc36c26706ee1 (MD5) / Made available in DSpace on 2016-09-12T13:57:22Z (GMT). No. of bitstreams: 1 DissMHP.pdf: 3604162 bytes, checksum: 5844b74d634a30ad629fc36c26706ee1 (MD5) Previous issue date: 2015-06-08 / Não recebi financiamento / Autonomous mobile robots are a special category of robots designed for performing tasks without the intervention of human beings. Some robots are designed to perform tasks in completely inhospitable environments such as the earth´s subsurface, the ocean depths or spatial exploration. In order to consider a robot as autonomous, a fundamental premise is to have self-adaptation capabilities. Over the last years, the advances in technology allow the development of self-adaptive systems, which are able to manage themselves to recuperate from faults or even change their behavior and structure in order to improve the quality of the delivered service. A critical point when building any software is its architecture, i.e., its structural organization in a set of interacting components. In this context, reference architecture is a technique that is well known for combining the best practices, patterns and strategies for building and standardizing domain specific software. Nowadays, there is a lack of studies presenting reference architectures for structuring self-adaptive software of mobile robots in order to decrease maintenance efforts. A number of studies claim that self-adaptive systems are based on the control theory and, more specifically, on the use of control loops in their architecture to perform adaptations. Therefore, this master thesis proposes SARAMR, a control loop-based reference architecture whose goal is to make maintenance activities a more productive task. The employment of the architecture divides the whole system in two modules; base application and adaptation module. The adaptation module encompasses the control loops and the base application is further divided into three other components: environment, behaviors and the electromechanical part. SARAMR was qualitatively evaluated by means of the development of two applications: a self-adaptive wall follower mobile robot and another conventional one to performing monitoring in in-door environments. Next, some maintenance activities were created to investigate the effort of applying them. We have observed that the separation of concerns of our architecture allows new components to be added causing less impacts than in systems developed in an adhoc way. / Robôs móveis autônomos fazem parte de uma categoria especial de robôs projetados para realizar tarefas sem a intervenção de seres humanos. Alguns robôs são projetados para realizar tarefas em ambientes completamente inóspitos à vida humana como no subsolo terrestre, nas profundezas de oceanos ou na exploração espacial. Para que um robô seja considerado autônomo, uma premissa fundamental é possuir capacidades de autoadaptação. Nos últimos anos, os avanços da tecnologia possibilitaram o desenvolvimento de sistemas robóticos autoadaptativos, que são capazes de gerenciarem a si próprios, se recuperarem de falhas e também de alterarem seu comportamento e estrutura com o objetivo de otimizar e/ou manter a qualidade do serviço (QoS) oferecido. Uma questão crítica para a concepção e construção de qualquer sistema de software é sua arquitetura, isto é, sua organização estrutural em um conjunto de componentes que interagem. Nesse contexto, a utilização de arquiteturas de referência é uma abordagem conhecida atualmente por combinar as melhores práticas, padrões e estratégias para a construção e padronização de sistemas de software para um determinado domínio. Atualmente, nota-se uma carência de estudos que apresentem arquiteturas de referência para estruturar o software de robôs móveis autoadaptativos de forma a facilitar atividades de manutenção nesses sistemas. Muitos estudos apontam que sistemas autoadaptativos são baseados na teoria do controle e mais especificamente na utilização de loops de controle em sua arquitetura para realizar as adaptações. Diante disso, este trabalho propõe a arquitetura de referência SARAMR, uma arquitetura de referência baseada em loops de controle cujo objetivo é facilitar atividades de manutenção no software desses sistemas. A utilização da arquitetura divide o sistema em dois módulos: aplicação base e módulo de adaptação. O módulo de adaptação envolve os loops de controle e a aplicação base ainda é subdividida em três componentes: ambiente, comportamentos e a parte eletromecânica. SARAMR foi avaliada de forma qualitativa mediante o desenvolvimento duas aplicações: um robô autoadaptativo seguidor de paredes e um outro convencional de patrulhamento. Depois disso, algumas manutenções evolutivas foram idealizadas no sentido de averiguar o esforço de aplicá-las. Constatou-se que a separação de interesses existente na arquitetura permite que novos componentes possam ser adicionados com impacto menor do que em sistemas que não usam essa arquitetura.

L'observation des exoplanètes pose deux problèmes : le contraste entre la planète et l'étoile et leur très faible séparation. L'une des techniques permettant de résoudre ces difficultés est l'interférométrie en frange noire : deux pupilles sont recombinés pour faire une interférence destructive sur l'étoile, et leur base est réglée pour que l'interférence soit constructive sur la planète. Cependant, pour garantir une extinction suffisante de l'étoile, la différence de trajet optique entre les faisceaux doit être de l'ordre du nanomètre, et le pointage meilleur que le centième de tache d'Airy, malgré les perturbations extérieures.Pour valider les points critiques d'une telle mission spatiale, un démonstrateur de laboratoire, PERSÉE, a été défini par un consortium dirigé par le CNES et incluant l'IAS, le LESIA, l'ONERA, l'OCA et Thales Alenia Space puis intégré à l'Observatoire de Meudon. Ce banc simule une mission spatiale dans son ensemble (interféromètre et cophasage nanométrique). Son objectif est de délivrer et maintenir une extinction de 10^-4 stabilisé à mieux que 10^-5 sur plusieurs heures, en présence de perturbations typiques que l'on injecte.Mon travail de thèse a consisté à intégrer le banc en étapes successives et à développer des procédures d'étalonnage. Ceci m'a aidé à caractériser les différents éléments critiques séparément avant de les regrouper. Après avoir mis en œuvre les boucles de contrôle du cophasage, leur analyse précise m'a permis de réduire à 0,3 nm rms le résidu de différence de marche, et à 0,4 % de la tache d'Airy le résidu de tip/tilt, malgré la présence de perturbations d'une dizaine de nanomètres d'amplitude, constituées de plusieurs dizaines de fréquences vibratoires entre 1 et 100 Hz. Cela a été possible grâce à l'implémentation d'un contrôleur linéaire quadratique gaussien, paramétré par la mesure préalable de la perturbation pour la réduire au maximum. Grâce à ces très bons résultats, j'ai pu obtenir un taux d'extinction record sur la bande [1,65 – 2,45] µm de 8,8x10^-6 stabilisé à 9x10^-7 sur quelques heures, soit une décade meilleure que les spécifications initiales. L'extrapolation de ces résultats au cas d'une mission spatiale montre que les performances attendues sont atteignables si le flux disponible est suffisamment important. Avec des télescopes de 40 cm et une fréquence d'asservissement de l'ordre de 100 Hz, des étoiles de magnitude inférieure à 9 devraient être observables. / There are two problems with the observation of exoplanets: the contrast between the planet and the star and their very low separation. One technique solving these problems is nulling interferometry: two pupils are recombined to make a destructive interference on the star, and their base is adjusted to create a constructive interference on the planet. However, to ensure a sufficient extinction of the star, the optical path difference between the beams must be around the nanometer, and the pointing must be better than one hundredth of Airy disk, despite the external disturbances.To validate the critical points of such a space mission, a laboratory demonstrator, PERSÉE, was defined by a consortium led by CNES, including IAS, LESIA, ONERA, OCA and Thales Alenia Space and integrated in Meudon Observatory. This bench simulates the entire space mission (interferometer and nanometric cophasing system). Its goal is to deliver and maintain an extinction of 10^-4 stable at better than 10^-5 over a few hours in the presence of typical injected disturbances.My thesis work consisted in integrating the bench in successive stages and to develop calibration procedures. This helped me to characterize the critical elements separately before grouping them. After having implemented the control loops of the cophasing system, their precise analysis helped me to reduce down to 0.3 nm rms the residual OPD, and 0.4 % of the Airy disk the residual tip/tilt, despite disturbances of tens of nanometers, consisting of several tens of vibrational frequencies between 1 and 100 Hz. This has been achieved by the implementation of a linear quadratic Gaussian controller, parameterized by the preliminary measurement of the disturbance to minimize. Thanks to these excellent results, I obtained on the band [1.65 – 2.45] µm a record null rate of 8.8x10^-6 stabilized at 9x10^-7 over a few hours, a decade better than the original specifications. An extrapolation of these results to the case of a space mission shows that the expected performance is achievable if the available flux is sufficiently important. With telescopes of 40 cm and a control frequency around 100 Hz, stars brighter than magnitude 9 should be observable.