This thesis presents the development and validation of a novel interface modelling methodology integrated with a system architectural analysis framework that emphasises the need to manage the integrity of deriving and allocating requirements across multiple levels of abstraction in a structured manner. The state of the art review in this research shows that there is no shared or complete interface definition model that could integrate diverse interaction viewpoints for defining system requirements with complete information. Furthermore, while existing system modelling approaches define system architecture with functions and their allocation to subsystems to meet system requirements, they do not robustly address the importance of considering well-defined interfaces in an integrated manner at each level of systems hierarchy. This results in decomposition and integration issues across the multiple levels of systems hierarchy. Therefore, this thesis develops and validates following: -Interface Analysis Template as a systematic tool that integrates diverse interaction viewpoints for modelling system interfaces with intensive information for deriving requirements. -Coupling Matrix as an architecture analysis framework that not only allocates functions to subsystems to meet requirements but also promotes consistent consideration of well-defined interfaces at each level of design hierarchy. Insights from the validation of developed approach with engineering case studies within an automotive OEM are discussed, reflecting on the effectiveness, efficiency and usability of the methods.

This thesis presents the development and validation of a novel interface modelling methodology integrated with a system architectural analysis framework that emphasises the need to manage the integrity of deriving and allocating requirements across multiple levels of abstraction in a structured manner. The state of the art review in this research shows that there is no shared or complete interface definition model that could integrate diverse interaction viewpoints for defining system requirements with complete information. Furthermore, while existing system modelling approaches define system architecture with functions and their allocation to subsystems to meet system requirements, they do not robustly address the importance of considering well-defined interfaces in an integrated manner at each level of systems hierarchy. This results in decomposition and integration issues across the multiple levels of systems hierarchy. Therefore, this thesis develops and validates following: -Interface Analysis Template as a systematic tool that integrates diverse interaction viewpoints for modelling system interfaces with intensive information for deriving requirements. -Coupling Matrix as an architecture analysis framework that not only allocates functions to subsystems to meet requirements but also promotes consistent consideration of well-defined interfaces at each level of design hierarchy. Insights from the validation of developed approach with engineering case studies within an automotive OEM are discussed, reflecting on the effectiveness, efficiency and usability of the methods.

Le travail de recherche propose et développe un cadre de système des systèmes (SdS) pour l’analyse de risques des installations industrielles et des infrastructures critiques. Les méthodes pour la représentation, la modélisation et la simulation d’un système sont développées pour identifier les particularités du SdS quant à leur vulnérabilité et leur résilience physique à des défaillances aléatoires et risques naturels. Plusieurs techniques de représentation, telles que l’arbre de défaillances, le Muir Web, la modélisation hiérarchique, le Goal Tree Success Tree – Dynamic Master Logic Diagram, sont étudiées et approfondies depuis l’origine pour s’adapter aux objectifs de l’analyse de SdS. Une méthode de représentation est développée ex novo, à savoir, le graphe hiérarchique. Dans ces cadres de représentation, des états binaires et multiples sont utilisés pour modéliser les performances des SdS à analyser. La simulation Monte Carlo et l’analyse d’intervalle sont combinées pour évaluer quantitativement des modèles de SdS en présence d’incertitude (due à la variabilité naturelle d’un phénomène ou au manque d’information). La mise en oeuvre de ces approches est illustrée dans deux domaines d’application : l’évaluation du risque d’événements externes et la vulnérabilité d’infrastructures critiques. / This thesis propounds and develops a system-of-systems (SoS) framework for the risk analysis of industrial installations and critical infrastructures. System representation, modeling and simulation methods are developed to capture the peculiar features of SoS, with respect to their vulnerability and physical resilience to random failures and natural hazards. Several representation techniques of literature, i.e., Fault Tree, Muir Web, Hierarchical Modeling, Goal Tree Success Tree – Dynamic Master Logic Diagram, are explored and originally extended/tailored to fit the purpose of SoS analysis. One representation method is developed ex-novo, namely the Hierarchical Graph. Within these representation frameworks, binary and multiple states are used to model the performances of the SoS under analysis. Monte Carlo simulation and interval analysis are combined for the quantitative evaluation of the SoS models in presence of uncertainty (due to both randomness and lack of knowledge). Examples of analyses are carried out within two application areas: external event risk assessment and vulnerability of critical infrastructures.

The current technological environment is evolving increasingly fast, and the development of new devices, technologies, and architectures has opened an emergent era where the digital and physical world work together. The implementation and use of systems based on a service-oriented architecture (SOA) in conjunction with the Internet of Things (IoT) and cyberphysical systems (CPS) have been extended during the last decades in numerous scenarios in industry and other domains. However, some of the major barriers to this approach are the lack of interoperability and the amount of engineering effort required for their integration.   The research presented in this thesis targets issues related to digitalization and automation. It is framed by the Industry 4.0 paradigm, which promotes the rise of efficiency and sustainability on industrial production. The interoperability between heterogeneous systems and different domains is one of the main challenges of Industry 4.0. The quest for solutions that help to increase interoperability is an important part of this research.   This thesis proposes a set of architectural design principles and tools in order to reduce engineering effort by means of finding solutions that enable autonomous integration and increase interoperability without human intervention. The research is focused on the IoT field, taking into account resource-constrained devices, system of systems integration, and data models.   A detailed investigation of various interoperability mismatch problems is presented in this thesis. The proposed solution is an adapter system that can aid in the generation of new service consumer interfaces at both compile-time and run-time. The proposed approach requires a new point of view in the service description field that can provide a holistic description of the information required for the generation of consumer interfaces. In addition, aspects related to interoperability, such as the multiple IoT frameworks in the current market, naming conventions, syntactic modeling and translation, and security, are also partially analyzed.   On a separate track, service composition in resource-constrained devices  is analyzed in terms of latency, using the orchestration provided by the Arrowhead Framework.

Un Système de Systèmes (SdS ou System of Systems - SoS) est un système complexe résultant de l'assemblage de composants existant ou à créer, de nature hétérogène (e.g. des systèmes techniques ou socio techniques appelés sous-systèmes, dispositifs techniques, acteurs ou organisations, ou encore des infrastructures plus ou moins complexes pouvant être perçues comme des SdS). Cet assemblage est nécessaire à ces composants pour agir et interagir avec d'autres composants afin de réaliser une mission commune, éventuellement limitée dans le temps et qu'aucun de ces composants ne pourrait réaliser seul. De fait, un SdS possède des caractéristiques particulières comme l'hétérogénéité, la possible émergence de propriétés et de comportements durant les interactions entre les composants et à leurs interfaces, la préservation de l'autonomie managériale et opérationnelle de ces composants, la répartition géographique de ces composants, un cycle de vie particulier, etc. L'Ingénierie Système (IS ou Systems Engineering - SE) propose et promeut un ensemble de concepts, de processus maintenant standardisés, l'usage incontournable de modèles (on parle alors de Model Based Systems Engineering – MBSE) et de bonnes pratiques pour concevoir et réaliser des systèmes complexes. Du fait de ses caractéristiques particulières, la conception et le développement d'un SdS (SoS Engineering - SoSE) est elle-même particulière même si elle emprunte à l'IS nombre de traits communs. En effet, le choix et l'assemblage des composants, leurs besoins en termes d'interfaces pour faciliter leurs interactions entre eux et avec l'environnement du SdS, les propriétés et comportements émergents entres autres caractéristiques, impliquent des efforts de la part des personnes en charge d'un SdS. Il faut alors, pour les aider dans leurs tâches, conceptualiser et développer des langages, méthodes et outils supports. Le SoSE a en effet des besoins particuliers de modélisation, de vérification, de validation de modèles. Il nécessite également de disposer de moyens de simulation et d'évaluation du comportement global du SdS et de ses propriétés, par exemple, lorsqu'il doit faire face à des événements redoutés (e.g. ajout, modification ou retrait d'un composants, évolution de la mission, etc.). Le but est que ces personnes puissent progresser en confiance et leur donner les moyens de fournir des modèles de SdS avec lesquels l'analyse des propriétés du SdS devient possible, avant même d'alimenter les activités de décision et d'optimisation en cours de conception du SdS. Ce travail s'intéresse à une propriété importante pour les SdS et leurs composants : l'interopérabilité. Elle est vue ici comme une exigence sommative des capacités et des capabilités des composants à être et rester compatibles, à inter opérer efficacement, à rester autonome pendant l'interaction et à la réversibilité de la relation d'interaction lorsque celle-ci s'achève. L'interopérabilité garantit donc ou, à défaut, maximise la capacité d'un composant à travailler sans perte et harmonieusement avec un autre composant, dans différentes situations et avec un niveau de performance attendu, tout en respectant un ensemble d'autres exigences venant des parties prenantes impliquées ou concernées par le SdS visé.Cette thèse consiste à formaliser et à développer une méthode pour accompagner la modélisation, la vérification de modèles et l'analyse de l'interopérabilité dans un SdS. En conséquence elle repose sur 1) un ensemble de concepts et de relations entre ces concepts pour décrire un SdS et la propriété d'interopérabilité, 2) des langages spécifiques de modélisation (DSML) pour manipuler ces concepts et relations et donc créer des « modèles » de SdS, 3) d'un processus opératoire et 4) d'outils de modélisation, de vérification des modèles, de simulation du comportement et d'évaluation de l'interopérabilité et de son influence sur la performance, la stabilité et l'intégrité du SdS en cours de fonctionnement. / A System of Systems (SdS) is a complex system which is seen as a group of, in most cases, existing and heterogeneous entities (e.g. technical systems or socio-technical called subsystems, actors or organizations or even complex infrastructures that can be considered as SoS) assembled together in order to interact, during a timeframe to produce some kind of capabilities, products or services and to achieve a global mission that a system alone cannot fulfill. Moreover, the SoS has some particular characteristics such as: Operational Independence and Managerial Independence (autonomy), Evolutionary Development, Emergent Behavior, Geographic distribution, Connectivity and Diversity etc. The systems engineering (SE) provides and promotes a set of concepts, principles, processes, standards, an essential use of models (Model Based Systems Engineering - MBSE)and a good practice to design and conduct complex systems. However, even if the System of Systems Engineering (SoSE) shares some common features with the SE, SoS characteristics, assembling, interfacing and interactions between its entities, induce an additional effort, required from the persons responsible of the SoS, over the SE. Therefore, and in order to help these persons in their tasks, it is necessary to conceptualize and develop languages, methods and tolls supports. The SoSE has special needs in terms of modeling and models' verification and validation. Moreover, it requires to have means to simulate and evaluate the global behavior of the SoS and its properties, for example, when it has to face dangerous events (e.g. adding, removing or modifying a component, mission's evolution etc.). The aim is to help designers and engineers to progress in confidence by giving them the means to have SoS models with which the analysis of the SoS properties becomes possible. In this work, a particular attention is given to an important property of the SoS and its components: the interoperability. It is seen here as a summative requirement of components capacities and capabilities to remain compatible, to interoperate and to remain autonomous during the interactions and reversible after it. The interoperability guarantees or, by default, maximizes the capacity of a component to work, harmoniously and without any loss, with another component, in various situations and with an expected level of performance while respecting a set of requirements (stakeholders involved or concerned by the SoS).This thesis consists in formalizing and developing a method to support modeling, model's verification and the analysis of the interoperability in a SoS. Therefore, it is based on 1) a set of concepts and relationships between these concepts to describe a SoS as well as the interoperability property, 2) Domain Specific Modeling Languages (DSMLs) to manipulate these concepts and relationships and thus creating a SoS' model, 3) an operating process and 4) a modeling and verification tools, simulating behavior and evaluation of the interoperability and its impact on the SoS performance, stability and integrity while it is operating.