Quand des facteurs humains interviennent dans une large majorité des accidents de la route, l’amélioration de la sécurité routière passe par l’introduction de systèmes d’assistance, afin d’aider le conducteur dans les tâches de conduites les plus complexes. Les systèmes de conduites « by-wire », en facilitant le partage des tâches de conduites entre le conducteur et les systèmes d’assistance, représentent une avancée majeure vers une automatisation progressive de la conduite.Cependant, le déploiement de ces systèmes est freiné pour des questions de sûreté de fonctionnement, et nécessite la mise en œuvre d’outils de diagnostique pour détecter et corriger d'éventuelles défaillances. Dans le cadre de cette thèse, nous évaluons un algorithme de détection et de localisation des fautes compatible avec une architecture « by-wire », basé sur l’approche multi-modèles interagissants. Cette méthode nécessite l'estimation probabiliste de l’état du véhicule, pour laquelle différents observateurs non linéaires sont comparés. Pour cette démarche, l’accent est mis sur la validation expérimentale des résultats qui a nécessité l’équipement d’un véhicule de test.Une fois que la faute est localisée, nous étudions les différentes stratégies de contrôle du véhicule en fonction des actionneurs encore disponibles.Cette étude montre que les effets d'une défaillance sur les directions découplées sont particulièrement difficiles à corriger. / The improvement of the road safety implies to increase the place of driving assistance systems for road vehicles. Paving the road of the fully autonomous vehicle, the drive-by-wire technology could improve the potential of the vehicle control. The implementation of these new embedded systems is still limited, mainly for reliability reasons, thus requiring the development of diagnostic mechanisms to detect occurring faults. In a first step, we evaluate a fault detection and isolation algorithm, based on the interacting multiple models approach. The method relies on a probabilistic estimation of the vehicle state, for which different non-linear observer schemes are compared. The experimental validation required the preparation of a test vehicle.Then, when a fault is identified, the optimal back-up control strategies are investigated according to the availability of actuators.Thus study points out that faults on steer-by-wire systems are particularly difficult to treat.

Information delivery on the Web often relies on general purpose Web pages that require the reader to adapt to them. This limitation is addressed by approaches such as spatial hypermedia and adaptive hypermedia. Spatial hypermedia augments the representation power of hypermedia and adaptive hypermedia explores the automatic modification of the presentation according to user needs. This dissertation merges these two approaches, combining the augmented expressiveness of spatial hypermedia with the flexibility of adaptive hypermedia. This dissertation presents the Multi-model Adaptive Spatial Hypermedia framework (MASH). This framework provides the theoretical grounding for the augmentation of spatial hypermedia with dynamic and adaptive functionality and, based on their functionality, classifies systems as generative, interactive, dynamic or adaptive spatial hypermedia. Regarding adaptive hypermedia, MASH proposes the use of multiple independent models that guide the adaptation of the presentation in response to multiple relevant factors. The framework is composed of four parts: a general system architecture, a definition of the fundamental concepts in spatial hypermedia, an ontological classification of the adaptation strategies, and the philosophy of conflict management that addresses the issue of multiple independent models providing contradicting adaptation suggestions. From a practical perspective, this dissertation produced WARP, the first MASH-based system. WARP’s novel features include spatial transclusion links as an alternative to navigational linking, behaviors supporting dynamic spatial hypermedia, and personal annotations to spatial hypermedia. WARP validates the feasibility of the multi-model adaptive spatial hypermedia and allows the exploration of other approaches such as Web-based spatial hypermedia, distributed spatial hypermedia, and interoperability issues between spatial hypermedia systems. In order to validate the approach, a user study comparing non-adaptive to adaptive spatial hypertext was conducted. The study included novice and advanced users and produced qualitative and quantitative results. Qualitative results revealed the emergence of reading behaviors intrinsic to spatial hypermedia. Users moved and modified the objects in order to compare and group objects and to keep track of what had been read. Quantitative results confirmed the benefits of adaptation and indicated a possible synergy between adaptation and expertise. In addition, the study created the largest spatial hypertext to date in terms of textual content.

No description available.

Machine tools are essential components of modern manufacturing. They are com posed of various mechanical, hydraulic, and electrical systems such as the spindle, tool changer, cooling system, and the linear and rotary feed drives. Due to their com plexity, high cost, and importance to the manufacturing process it is recommended to implement some sort of condition monitoring and predictive maintenance to ensure that they remain reliable and high performing. One way of potentially implement ing predictive maintenance and condition monitoring is digital twins. Digital twins enable the real-time, accurate, and complex modeling and monitoring of mechanical systems. They utilize data collected from the system to constantly update their mod els which can be used for monitoring of the systems state and future predictions. This work presents a digital twin workbench of a machine tool feed drive. The workbench enables the collection and analysis of large, varied, high-frequency data which can be used to construct a digital twin of the feed drive. A digital twin can enable many other useful functionalities. Some of these functionalities include condition moni toring, modeling, control, visualization, and simulation. These functionalities can enable maximum asset performance and are key in implementing effective predictive maintenance. The main contributions of this work are the following: The design and iv construction of a machine tool feed drive which implements a novel external distur bance force method. A new method of fault detection in ball screws using interacting multiple models which was shown to provide accurate estimates of levels of preloads in a ball screw driven feed drive. A digital twin based modeling strategy and analysis of the data generated by the system including system modeling and observations on modeling difficulties. / Thesis / Master of Applied Science (MASc) / Digital twins enable the real-time, accurate, and complex modeling and monitoring of mechanical systems. Machine tools are essential components of modern manufac turing. They are composed of various mechanical, hydraulic, and electrical systems such as the spindle, tool changer, cooling system, and linear and rotary feed drives. This work presents the design of a workbench of a machine tool linear feed drive, a fault detection strategy, and a digital twin modeling solution. The workbench enables the collection and analysis of large, varied, high-frequency data which can be used to construct a digital twin of the feed drive. A digital twin can enable many other useful functionalities. Some of these functionalities include condition monitoring, modeling, control, visualization, and simulation. These functionalities can enable maximum asset performance and are key in implementing effective predictive maintenance.

To develop and maintain a modern combat aircraft it is important to have simple, yet accurate, threat models to support early stages of functional development. Therefore this thesis develops and evaluates a model of an active radio frequency (RF) seeker for a missile with data-link capability. The highly parametrized MATLAB-model consists of a pulse level radar model, a tracker using either interacting multiple models (IMM) or particle filters, and a guidance law. Monte Carlo simulations with the missile model indicate that, under the given conditions, the missile performs well (hit rate>99%) with both filter types, and the model is relatively insensitive to lost data-link transmissions. It is therefore under normal conditions not worthwhile to use the more computer intense particle filter today, however when the data-link degrades the particle filter performs considerably better than the IMM filter. Analysis also indicate that the measurements generated by the radar model are neither independent, white nor Gaussian. This contradicts the assumptions made in this, and many other radar applications. However, the performance of the model suggests that the assumptions are acceptable approximations of actual conditions, but further studies within this are recommended to verify this.

Predictive control relies on predictions of the future behaviour of the system to be controlled. These predictions are calculated from a model of this system, thus making the model the cornerstone of the predictive controller. Furthermore predictive control is the only advanced control methodology that has managed to become widely used in the industry. The necessity of good models in the predictive control context can thus be motivated both from the very nature of predictive control and from its widespread use in industry. This thesis is concerned with examining the use of multiple models in the predictive controller. In order to do this the standard predictive control formulation has been extended to incorporate the use of multiple models. The most general case of this new formulation allows the use of an individual model for each prediction horizon. The models are estimated using measurements of the input and output sequences from the true system. When using this data to find a good model of the system it is important to remember the intended purpose of the model. In this case the model is going to be used in a predictive controller and the most important feature of the models is to deliver good k-step ahead predictions. The identification algorithms used to estimate the models thus strives for estimating models good at calculating these predictions. Finally this thesis presents some complete simulations of these ideas showing the potential of using multiple models in the predictive control framework.

In this thesis, a new guidance method for fighter aircrafts and a new guidance method for missiles are developed. Also, guidance and control systems of the aircraft and the missile used are designed to simulate the generic engagement scenarios between the missile and the aircraft. Suggested methods have been tested under excessive simulation studies. The aircraft guidance method developed here is a rule based missile evasion method. The main idea to develop this method stems from the maximization of the miss distance for an engagement scenario between a missile and an aircraft. To do this, an optimal control problem with state and input dependent inequality constraints is solved and the solution method is applied on different problems that represent generic scenarios. Then, the solutions of the optimal control problems are used to extract rules. Finally, a method that uses the interpolation of the extracted rules is given to guide the aircraft. The new guidance method developed for missiles is formulated by modifying the classical proportional navigation guidance method using the position estimates. The position estimation is obtained by utilization of a Kalman based filtering method, called interacting multiple models.

<p>Predictive control relies on predictions of the future behaviour of the system to be controlled. These predictions are calculated from a model of this system, thus making the model the cornerstone of the predictive controller. Furthermore predictive control is the only advanced control methodology that has managed to become widely used in the industry. The necessity of good models in the predictive control context can thus be motivated both from the very nature of predictive control and from its widespread use in industry. </p><p>This thesis is concerned with examining the use of multiple models in the predictive controller. In order to do this the standard predictive control formulation has been extended to incorporate the use of multiple models. The most general case of this new formulation allows the use of an individual model for each prediction horizon. </p><p>The models are estimated using measurements of the input and output sequences from the true system. When using this data to find a good model of the system it is important to remember the intended purpose of the model. In this case the model is going to be used in a predictive controller and the most important feature of the models is to deliver good k-step ahead predictions. The identification algorithms used to estimate the models thus strives for estimating models good at calculating these predictions. </p><p>Finally this thesis presents some complete simulations of these ideas showing the potential of using multiple models in the predictive control framework.</p>

<p>To develop and maintain a modern combat aircraft it is important to have simple, yet accurate, threat models to support early stages of functional development. Therefore this thesis develops and evaluates a model of an active radio frequency (RF) seeker for a missile with data-link capability. The highly parametrized MATLAB-model consists of a pulse level radar model, a tracker using either interacting multiple models (IMM) or particle filters, and a guidance law. </p><p>Monte Carlo simulations with the missile model indicate that, under the given conditions, the missile performs well (hit rate>99%) with both filter types, and the model is relatively insensitive to lost data-link transmissions. It is therefore under normal conditions not worthwhile to use the more computer intense particle filter today, however when the data-link degrades the particle filter performs considerably better than the IMM filter. Analysis also indicate that the measurements generated by the radar model are neither independent, white nor Gaussian. This contradicts the assumptions made in this, and many other radar applications. However, the performance of the model suggests that the assumptions are acceptable approximations of actual conditions, but further studies within this are recommended to verify this.</p>

Cette étude traite de l’identification de système dynamique non-linéaire. Une architecture multimodèle capable de surmonter certaines difficultés de l’architecture neuronale de type MLP a été étudiée. L’approche multimodèle consiste à représenter un système complexe par un ensemble de modèles de structures simples à validité limitée dans des zones bien définies. A la place de la structure affine des modèles locaux généralement utilisée, cette étude propose une structure polynômiale plus générale, capable de mieux appréhender les non-linéarités locales, réduisant ainsi le nombre de modèles locaux. L’estimation paramétrique d’une telle architecture multimodèle peut se faire suivant une optimisation linéaire, moins coûteuse en temps de calcul que l’estimation paramétrique utilisée dans une architecture neuronale. L’implantation des multimodèles récurrents, avec un algorithme d’estimation paramétrique plus souple que l’algorithme de rétro-propagation du gradient à travers le temps utilisé pour le MLP récurrent a également été effectuée. Cette architecture multimodèle permet de représenter plus facilement des modèles non-linéaires bouclés tels que les modèles NARMAX et NOE. La détermination du nombre de modèles locaux dans une architecture multimodèle nécessite la décomposition (le partitionnement) de l’espace de fonctionnement du système en plusieurs sous-espaces où sont définies les modèles locaux. Des modes de partitionnement flou (basé sur les algorithmes de« fuzzy-c-means », de « Gustafson et Kessel » et du « subtractive clustering ») ont été présentés. L’utilisation de telles méthodes nécessite l’implantation d’une architecture multimodèle où les modèles locaux peuvent être de structures différentes : polynômiales de degrés différents, neuronale ou polynômiale et neuronale. Une architecture multimodèle hétérogène répondant à ses exigences a été proposée, des algorithmes d’identification structurelles et paramétriques ont été présentés. Une étude comparative entre les architectures MLP et multimodèle a été menée. Le principal atout de l’architecture multimodèle par rapport à l’architecture neuronale de type MLP est la simplicité de l’estimation paramétrique. Par ailleurs, l’utilisation dans une architecture multimodèle d’un mode de partitionnement basé sur la classification floue permet de déterminer facilement le nombre de modèles locaux, alors que la détermination du nombre de neurones cachés pour une architecture MLP reste une tâche difficile / This work deals with non linear dynamical system identification. A multiple model architecture which overcomes certain insufficiencies of MLP neural networks is studied. Multiple model approach consists of modeling complex systems by mean of a set of simple local models whose validity are limited in well defined zones. Instead of using conventional affine models, a more general polynomial structure is proposed in this study, enabling to better apprehend local non linearities, reducing thus the number of local models. Models parameters of such a structure are estimated by linear optimization, which reduces computation time with respect to parameter estimation of a neural network architecture. The implementation of recurrent multiple models, with a more convenient learning algorithm than the back propagation through time, used in recurrent MLP models, is also studied. Such implementations facilitate representation of recurrent models like NARMAX and NOE. The determination of the number of local models in a multiple model architecture requires decomposition of system’s feature space into several sub-systems in which local models are defined. Fuzzy partitioning methods (based of « fuzzy-c-means », « Gustafson and Kessel » and « subtractive clustering »algorithms) are presented. The use of such methods requires the implementation of a multiple model architecture where local models can have different structures : polynomial with different degrees, neural or polynomial and neural. A multiple model with a heterogeneous architecture satisfying these requirements is proposed and structural and parametrical identification algorithms are presented. A comparative study between multiple model and MLP architectures is done. The main advantage of the multiple model architecture is the parameter estimation simplicity. In addition, the use of fuzzy partitioning methods in multiple model architecture enables to find easily the number of local models while the determination of hidden neurons in an MLP architecture remains a hard task