Universiẗat, Diss., 2005--Düsseldorf.

Le 21_eme siècle est le siècle de l'explosion informatique, des milliards de Données sont produites, collectées et stockées dans notre vie quotidienne. Les façons de collecter les ensembles de données sont multiples mais toujours en essayant d'optimiser le critère qui consiste _a avoir le maximum d'information dans le minimum de données numérique. Il est préférable de collecter directement l'information, car les informations étant contraintes sont dans un espace plus faible que celui où évolues les données (signaux ou états). Cette méthode est donc appelée \la collecte de l'information", et conceptuellement peut ^être résumée dans les trois étapes suivantes : (1) la modélisation, ceci consiste _a condenser l'information pertinente pour les signaux _a un sous-espace plus petit; (2) l'acquisition, ceci consiste _a collecter et préserver l'information dans un espace inferieur _a la dimension des données et (3) la restauration, ceci consiste _a reconstituer l'information dans son espace d'origine. En suivant cette pensée, les principales contributions de cette thèse, concernant les observateurs et le \Compressive Sensing" (CS) basé sur des modèles bay_esiens peuvent ^être unies dans le cadre de la collecte de l'information : les principaux problèmes concernés par ces deux applications peuvent ^être de façon analogue, scindés en les trois étapes sus- mentionnées. Dans la première partie de la th_ese, le problème réside dans le domaine des systèmes dynamiques où l'objectif est de retrouver l'état du système _a partir de la mesure de la sortie. Il nous faut donc déterminer si les états du système sont récupérables _a partir des mesures de la sortie et de la connaissance partielle ou totale du modèle dynamique, c'est le problème de l'observabilité. Ensuite de transposer notre problème dans une représentation plus appropriée, c'est l'écriture sous forme normale et en récupérer l'information, c'est la phase de synthèse d'observateur. Plus précisément dans cette partie, nous avons considéré une classe de systèmes à commutation haute fréquence allant jusqu'au phénomène de Zénon. Pour ces deux types de commutation les transitions de l'état discret sont considérées trop élevées pour ^être mesurées. Toutefois, la valeur moyenne obtenue par filtrage des transitions peut ^être acquise ce qui donne une connaissance partielle des états discrets. Ici, avec ces seuls informations partielles, nous avons discuté de l'observabilité et ceci par les approches géométrie différentielle et algébrique. Aussi, des observateurs ont été proposes par la suite. Dans la deuxième partie de cette thèse, nous avons abordé de la même manière le thème du CS qui est une alternative efficace à l'acquisition abondante de données faiblement informatives pour ensuite les compresser. Le CS se propose de collecter l'information directement de façon compressée, ici les points clés sont la modélisation du signal en fonction des connaissances a priori dont on dispose, ainsi que la construction d'une matrice de mesure satisfaisant la \restricted isometry property" et finalement la restauration des signaux originaux clairsemés en utilisant des algorithmes d'éparpillement régularisé et d'inversion linéaire. Plus précisément, dans cette seconde partie, en considérant les propriétés du CS liées _a la modélisation, la capture et la restauration, il est proposé : (1) d'exploiter les séquences chaotiques pour construire la matrice de mesure qui est appelée la matrice chaotique de mesure, (2) considérer des types de modèle de signal clairsemé et reconstruire le modèle du signal à partir de ces structures sous-jacentes des modèles clairsemés, et (3) proposer trois algorithmes non paramétriques pour la méthode bayesienne hiérarchique. Dans cette dernière partie, des résultats expérimentaux prouvent d'une part que la matrice chaotique de mesure a des propriétés semblables aux matrices aléatoires sous-gaussienne et d'autre part que des informations supplémentaires sur les structures sous-jacentes clairsemés améliorent grandement les performances de reconstruction du signal et sa robustesse vis-a-vis du bruit. / This is the era of information-explosion, billions of data are produced, collected and then stored in our daily life. The manners of collecting the data sets are various but always following the criteria { the less data while the more information. Thus the most favorite way is to directly measure the information, which, commonly, resides in a lower dimensional space than its carrier, namely, the data (signals or states). This method is thus called information measuring, and conceptually can be concluded in a framework with the following three steps: (1) modeling, to condense the information relevant to signals to a small subspace; (2) measuring, to preserve the information in lower dimensional measurement space; and (3) restoring, to reconstruct signals from the lower dimensional measurements. From this vein, the main contributions of this thesis, saying observer and model based Bayesian compressive sensing can be well uni_ed in the framework of information measuring: the main concerned problems of both applications can be decomposed into the above three aspects. In the _rst part, the problem is resided in the domain of control systems where the objective of observer design is located in the observability to determine whether the system states are recoverable and observation of the system states from the lower dimensional measurements (commonly but not restrictively). Speci_cally, we considered a class of switched systems with high switching frequency, or even with Zeno phenomenon, where the transitions of the discrete state are too high to be captured. However, the averaged value obtained through filtering the transitions can be easily sensed as the partial knowledge. Consequently, only with this partial knowledge, we discussed the observability respectively from differential geometric approach and algebraic approach and the corresponding observers are designed as well. At the second part, we switched to the topic of compressive sensing which is objected to sampling the sparse signals directly in a compressed manner, where the central fundamentals are resided in signal modeling according to available priors, constructing sensing matrix satisfying the so-called restricted isometry property and restoring the original sparse signals using sparse regularized linear inversion algorithms. Respectively, considering the properties of CS related to modeling, measuring and restoring, we propose to (1) exploit the chaotic sequences to construct the sensing matrix (or measuring operator) which is called chaotic sensing matrix, (2) further consider the sparsity model and then rebuild the signal model to consider structures underlying the sparsity patterns, and (3) propose three non-parametric algorithms through the hierarchical Bayesian method. And the experimental results prove that the chaotic sensing matrix is with the similar property to sub-Gaussian random matrix and the additional consideration on structures underlying sparsity patterns largely improves the performances of reconstruction and robustness.

The motivation of this research project was in response to problems of re-condensation in drying, reduced drying rate encountered by the food and beverage packaging industry which led to the aim of developing a better performing drying system as well as achieving high energy efficiency. A hybrid dryer suited for rapid drying applications is designed, constructed and experimentally tested and considered in atmospheric environment only. The system employs a heat pump in conjunction with a heat reactivated desiccant wheel to provide an efficient drying capability and supply low dew point temperature (DPT) conditions. The combined system utilises the heat dissipated by the condenser in regenerating the desiccant wheel, to increase the economic feasibility of such a hybrid system. Up to 60% heat energy can be saved by using the hybrid system in the rapid surface drying applications. Mathematical models are developed to obtain the correlations among the design operating and performance parameters of the dehumidification systems. The mathematical models can be used to estimate the performance of the hybrid system as well as the performance of the individual components of the system. A prototype model was designed, fabricated and tested. The experimental facility consisted of a heat pump desiccant dehumidifier with the new ecological R134a as a refrigerant which used the heat dissipated by the condenser. An analysis of the experimental data was conducted to determine the practical relationship between the operational parameters (COP, ma and TR) and performance parameters (SMER, DPT and ε) of the system. The observed behaviours of the test cases are suggested to be governed by a specific combination of the operation parameters. The analysis shows that the proposed hybrid system can deliver supply air at a much lower DPT compared with the single refrigerant circuit and a desiccant wheel. It is shown that the specific moisture extraction rate (SMER) for conventional dryers is 0.5 - 1 kg/kWh and SMER for heat pump based system is 3 - 4 kg/kWh whereas the hybrid system achieves SMER >5 kg/kWh. By operating the combined system in tandem, a greater amount of dehumidification could be realised due to the improved ratio of latent to the total load. The present research also confirms the importance of improving heat recovery to improve the performance of a heat-pump-assisted drying system.

In this thesis, a methodology for the detection of anomalies in the cardiovascular system is presented. The cardiovascular system is one of the most fascinating and complex physiological systems. Nowadays, cardiovascular diseases constitute one of the most important causes of mortality in the world. For instance, an estimate of 17.3 million people died in 2008 from cardiovascular diseases. Therefore, many studies have been devoted to modeling the cardiovascular system in order to better understand its behavior and find new reliable diagnosis techniques. The lumped parameter model of the cardiovascular system proposed in [1] is restructured using a hybrid systems approach in order to include a discrete input vector that represents the influence of the mitral and aortic valves in the different phases of the cardiac cycle. Parting from this model, a Taylor expansion around the nominal values of a vector of parameters is conducted. This expansion serves as the foundation for a component fault detection process to detect changes in the physiological parameters of the cardiovascular system which could be associated with cardiovascular anomalies such as atherosclerosis, aneurysm, high blood pressure, etc. An Extended Kalman Filter is used in order to achieve a joint estimation of the state vector and the changes in the considered parameters. Finally, a bank of filters is, as in [2], used in order to detect the appearance of heart valve diseases, particularly stenosis and regurgitation. The first numerical results obtained are presented.

The simultaneous treatment of different produced water streams with the forward osmosis membrane distillation hybrid system (FO-MD) has been suggested recently. This work investigates the need for pretreatment of produced water prior to filtration with FO-MD in order to reduce the level of fouling and scaling in the system. The desalter effluent (DE) stream was selected as FO feed solution, and the water oil separator (WO) stream was used as FO draw solution/MD feed solution, and a significant flux decline was observed in FO and MD within the first 5 hours of operations. SEM and EDX analysis indicated that the formation of scale layer on both membranes was the main reason for the sharp flux decline. Silica was the major contributor to the scaling of the support layer of the FO membrane. While the scaling layer on MD membrane consisted mainly of CaSO4 crystals with some deposition of Silica. Therefore, electrocoagulation (EC) was selected for the pretreatment of produced water to target the removal of Ca, SiO2 and SO4 ions in order to reduce the likelihood of inorganic fouling in FO-MD. The different parameters of EC, namely, the current density, electrolysis time, and initial pH were tested at a wide range of values of 7-70 mA/cm2 , 10-60 minutes, 5-9, respectively. calcium and sulfate ions were not effectively removed at the relatively high applied current density of 70 mA/cm2 , while high removal of silica was achieved even at low applied current densities. The optimum conditions of EC for silica removal were found to be 7 mA/cm2 for the current density and 10 minutes for the electrolysis time which resulted in a 97% removal of silica. it was found that due to pretreatment, the average FO and MD fluxes increased by 49% (9.93 LMH) and 39% (8.55 LMH), respectively. Therefore, even though EC did not show promising results in terms of the removal of calcium and sulfate, efficient silica removal was achieved with minimum energy requirements which suggests that it could have a potential to be integrated with the FO-MD hybrid system for the treatment and reclamation of produced water.

No description available.

No description available.

No description available.

The goal of an optimal control algorithm is to improve the performance of a system. For a stochastic system, a typical optimal control method minimizes the mean (first cumulant) of the cost function. However, there are other statistical properties of the cost function, such as variance (second cumulant) and skewness (third cumulant), which will affect the system performance. In this dissertation, the work on the statistical optimal control are presented, which extends the traditional optimal control method using cost cumulants to shape the system performance. Statistical optimal control will allow more design freedom to achieve better performance. The solutions of statistical control involve solving partial differential equations known as Hamilton-Jacobi-Bellman equation. A numerical method based on neural networks is employed to find the solutions of the Hamilton-Jacobi-Bellman partial differential equation. Furthermore, a complex problem such as multiple satellite control, has both continuous and discrete dynamics. Thus, a hierarchical hybrid architecture is developed in this dissertation where the discrete event system is applied to discrete dynamics, and the statistical control is applied to continuous dynamics. Then, the application of a multiple satellite navigation system is analyzed using the hierarchical hybrid architecture. Through this dissertation, it is shown that statistical control theory is a flexible optimal control method which improves the performance; and hierarchical hybrid architecture allows control and navigation of a complex system which contains continuous and discrete dynamics. / Electrical and Computer Engineering

Yes / The coupling of thermal (Multi Stage Flash, MSF) and membrane processes (Reverse Osmosis, RO) in desalination systems has been widely presented in the literature to achieve an improvement of performance compared to an individual process. However, very little study has been made to the combined Multi Effect Distillation (MED) and Reverse Osmosis (RO) processes. Therefore, this research investigates several design options of MED with thermal vapor compression (MED_TVC) coupled with RO system. To achieve this aim, detailed mathematical models for the two processes are developed, which are independently validated against the literature. Then, the integrated model is used to investigate the performance of several configurations of the MED_TVC and RO processes in the hybrid system. The performance indicators include the fresh water productivity, energy consumption, fresh water purity, and recovery ratio. Basically, the sensitivity analysis for each configuration is conducted with respect to seawater conditions and steam supply variation. Most importantly, placing the RO membrane process upstream in the hybrid system generates the overall best configuration in terms of the quantity and quality of fresh water produced. This is attributed to acquiring the best recovery ratio and lower energy consumption over a wide range of seawater salinity.