O desenvolvimento de sistemas de controle em plataformas de tempo real é uma tarefa que envolve Engenharia de Controle e Ciência da Computação. Nas últimas décadas, estas áreas se desenvolveram como áreas independentes. Este trabalho busca diminuir distância entre as áreas propondo a utilização de métodos de Engenharia de Software em uma fase de modelagem do software de controle. Uma das propostas apresentadas no trabalho é a utilização de frameworks orientados a objetos no processo de Rapid Control Prototyping (RCP) para substituir a geração automática de código, eliminar os problemas de integração com código legado e tornar o processo RCP mais interativo. Outra proposta é a utilização de plataformas para RCP composta por uma camada de hardware real, uma camada SOTR e uma camada de aplicação formada pelo framework para análise e desenvolvimento de sistemas de controle centralizado ou distribuído. / The development of real-time platform control systems is a task that involves Control Engineering and Computer Science. In the last decade, these areas have developed independent from each other. This paper seeks to decrease the distance between this areas, by proposing the use of Software Engineering methods in a software control modeling phase. One of the propositions in this paper is the use of object orientated frameworks in the Rapid Control Prototyping (RCP) process to substitute the automatic code generation, thus eliminating the problems with the legacy code and making the RCP process more interactive. Another proposition is the use of RCP directed platforms composed by a real hardware layer, a RTOS layer and an application layer formed by the framework for the analysis and development of the centralized or distributed control systems.

O desenvolvimento de sistemas de controle em plataformas de tempo real é uma tarefa que envolve Engenharia de Controle e Ciência da Computação. Nas últimas décadas, estas áreas se desenvolveram como áreas independentes. Este trabalho busca diminuir distância entre as áreas propondo a utilização de métodos de Engenharia de Software em uma fase de modelagem do software de controle. Uma das propostas apresentadas no trabalho é a utilização de frameworks orientados a objetos no processo de Rapid Control Prototyping (RCP) para substituir a geração automática de código, eliminar os problemas de integração com código legado e tornar o processo RCP mais interativo. Outra proposta é a utilização de plataformas para RCP composta por uma camada de hardware real, uma camada SOTR e uma camada de aplicação formada pelo framework para análise e desenvolvimento de sistemas de controle centralizado ou distribuído. / The development of real-time platform control systems is a task that involves Control Engineering and Computer Science. In the last decade, these areas have developed independent from each other. This paper seeks to decrease the distance between this areas, by proposing the use of Software Engineering methods in a software control modeling phase. One of the propositions in this paper is the use of object orientated frameworks in the Rapid Control Prototyping (RCP) process to substitute the automatic code generation, thus eliminating the problems with the legacy code and making the RCP process more interactive. Another proposition is the use of RCP directed platforms composed by a real hardware layer, a RTOS layer and an application layer formed by the framework for the analysis and development of the centralized or distributed control systems.

Les travaux de cette thèse s'inscrivent dans l'objectif général de fournir des recommandations sur la façon d'intégrer les effets du site dans l'évaluation probabiliste des risques sismiques, mieux connue sous le nom de PSHA, une méthodologie connue et utilisée à l'échelle mondiale pour estimer l'aléa et le risque sismiques à l'échelle régionale et locale. Nous passons donc en revue les méthodes disponibles dans la littérature pour obtenir la courbe d'aléa sismique en surface d'un site non-rocheux, en commençant par les méthodes les plus simples et plus génériques (partiellement probabiliste), jusqu'aux méthodes site-spécifiques (partiellement et entièrement probabilistes) qui nécessitent une caractérisation du site de plus en plus poussée, rarement disponible sauf cas exceptionnel comme par exemple le site test d'Euroseistest. C'est justement sur l'exemple de ce site que sont donc comparées un certain nombre de ces méthodes, ainsi qu'une nouvelle.La spécificité et la difficulté de ces études PSHA "site-spécifiques" vient du caractère non-linéaire de la réponse des sites peu rigides, ainsi que du fait que le rocher de référence contrôlant cette réponse est souvent très rigide. Les aspects "ajustement rocher dur" et "convolution" de l'aléa sismique au rocher avec la fonction d'amplification ou la fonction transfert (empirique ou numérique) d’un site font donc l'objet d'une attention particulière dans ces études comparatives. Un cadre général est présenté sur la façon de prendre en compte simultanément les caractéristiques spécifiques au site, la variabilité aléatoire complète ou réduite ("single station sigma"), les ajustements hôte-cible et le comportement linéaire / non linéaire d'un site, où nous expliquons toutes les étapes, corrections, avantages et difficultés que nous avons trouvés dans le processus et les différentes façons de les mettre en oeuvre.Cette étude comparative est divisée en deux parties: la première porte sur les méthodes non site-spécifiques et les méthodes hybrides site-spécifique (évaluation probabiliste de l'aléa au rocher et déterministe de la réponse de site), la seconde porte sur deux approches prenant en compte la convolution aléa rocher / réponse de site de façon probabiliste. Un des résultats majeurs de la première est l'augmentation de l'incertitude épistémique sur l'aléa en site meuble comparé à l'aléa au rocher, en raison du cumul des incertitudes associées à chaque étape. Un autre résultat majeur commun aux deux études est l'impact très important de la non-linéarité du sol dans les sites souples, ainsi que de la façon de les prendre en compte: la variabilité liée à l'utilisation de différents codes de simulation NL apparaît plus importante que la variabilité liée à différentes méthodes de convolution 100% probabilistes. Nous soulignons l'importance d'améliorer la manier d’inclure les effets du site dans les méthodes de l’estimation de l’aléa sismique probabiliste ou PSHA, et nous soulignons aussi l'importance d'instrumenter des sites actifs avec des sédiments meubles, comme l'Euroseistest, afin de tester et valider les modèles numériques.Finalement, on présente un résumé des résultats, des conclusions générales, de la discussion sur les principaux problèmes méthodologiques et des perspectives d'amélioration et de travail futur.Mots-clés: Effets du site, incertitude épistémique, PSHA, single station sigma, ajustements hôte-cible, effets linéaires et non linéaires, réponse de site / The overall goal of this research work is of provide recommendations on how to integrate site effects into Probabilistic Seismic Hazard Assessment, better known as PSHA, a well-known and widely used methodology. Globally used to estimate seismic hazard and risk at regional and local scales. We therefore review the methods available in the literature to obtain the seismic hazard curve at the surface of a soft soil site, starting with the simplest and most generic methods (partially probabilistic), up to the full site-specific methods (partially and fully probabilistic), requiring an excellent site-specific characterization, rarely available except exceptional cases such as the case of Euroseistest site. It is precisely on the example of this site that are compared a number of these methods, as well as a new one. And it is precisely at the Euroseistest that we performed an example of application of the different methods as well as a new one that we propose as a result of this work.The specificity and difficulty of these "site-specific" PSHA studies comes from the non-linear nature of the response of the soft sites, as well as from the fact that the reference rock controlling this response is often very rigid. The "rock to hard rock adjustment" and "convolution" aspects of the rock seismic hazard, together with the amplification function or the transfer function (empirical or numerical) of a site are therefore the subject of particular attention in these studies. comparative studies. A general framework is presented on how to simultaneously take into account the site-specific characteristics, such as the complete or reduced random variability ("single station sigma"), host-to -target adjustments and the linear / nonlinear behavior of a site, where we explain all the followed steps, the different corrections performed, the benefits and difficulties that we found in the process and the ways we sort them and discussing them when the answer was not straight forward.This comparative study is divided into two parts: the first deals with non-site-specific methods and site-specific hybrid methods (probabilistic evaluation of rock hazard and deterministic of the site response). The second deals with two approaches taking into account the convolution of rock hazard and the site response in a probabilistically way. One of the major results of the first is the increase of the epistemic uncertainty on the soft site hazard compared to the rock hazard, due to acumulation of uncertainties associated to each step. Another major common result to both studies is the very important impact of non-linearity on soft sites, as well as the complexity on how to account for them: the variability associated with the use of different non-linear simulation codes appears to be greater than the method-to-method variability associated with the two different full convolution probabilistic methods. We emphasize on the importance of improving the way in which the site effects are included into probabilistic seismic hazard methods, PSHA. And we also emphasize on the importance of instrumenting active sites with soft sediments, such as the Euroseistest, to test and validate numerical models.Finally, a summary of the results, the general conclusions, discussion of key methodological issues, and perspectives for improvement and future work are presented.Keywords: Site Effects, Epistemic Uncertainty, PSHA, single station sigma, host to target adjustments, linear and nonlinear site effects, soil site response.