Les travaux présentés dans ce mémoire s’inscrivent dans un projet visant à fabriquer des valves aortiques bioartificielles de remplacement pour des patients atteints de maladies cardiaques. La méthode globale étudiée consiste à produire un moule sacrificiel en sucre vitrifié produit par fabrication additive, prenant la forme d’une valve aortique et injecté avec un échafaudage cellulaire. En soumettant la valve moulée aux conditions physiologiques ressenties par une valve aortique réelle, il est espéré qu’une valve aortique fonctionnelle sera développée. Un des éléments importants dans ce procédé est l’échafaudage cellulaire. Puisque ce biomatériau contient des cellules vivantes, il doit être à l’abri de toutes sources de contamination. De plus, il doit permettre aux cellules de survivre et de sécréter de la matrice extracellulaire, dans le but d’éventuellement transformer l’échafaudage cellulaire en un tissu biologique fiable. Ce mémoire présente une technique de fabrication d’échafaudages cellulaires qui tient compte des enjeux liés à l’utilisation de cellules vivantes. Il s’agit d’une preuve de concept visant à s’intégrer au projet de valves aortiques bioartificielles. Afin de tester la méthode, une expérience in vitro de fabrication et de culture dynamique fut menée. Celle-ci démontra que cette méthode de fabrication fut adaptée au contexte de travail en environnement stérile, que les cellules ensemencées dans les spécimens furent distribuées de manière homogène, et que les moules en sucre vitrifié fabriqués par impression 3D ne causèrent pas de mortalité cellulaire dans ce contexte. Toutefois, des dommages mineurs furent observés après plusieurs semaines de culture, et les taux de viabilité cellulaire furent plus bas qu’attendu à cause d’un défaut au niveau de la perfusion des spécimens. Ainsi, la technique développée est prometteuse pour le projet de fabrication de valves aortiques, mais des améliorations doivent être apportées au niveau de la perfusion et du maintien de l’intégrité physique des tissus. / The work presented in this thesis is part of a project which aims at fabricating bioartificial aortic replacement valves for patients suffering from cardiac diseases. The global method studied to achieve this consists of fabricating sacrificial molds made of carbohydrate glass, produced by additive manufacturing, replicating the geometry of an aortic valve, and injected with a cellular scaffold. By exposing the molded valve to the physiological conditions a real aortic valve would experience, it is hoped that a functional aortic valve will be developed. One important aspect of this process is the cellular scaffold. Since this biomaterial contains live cells, it has to be isolated from all possible sources of contamination. Moreover, it has to favor cell survival, as well as extracellular matrix secretion, in order to eventually transform the scaffold into a reliable biological tissue. This thesis presents a fabrication technique for cellular scaffold that takes into account all the challenges linked to the use of live cells. It is a proof of concept with the aim of being included to the artificial aortic valve project. To validate this process and its aspects, an in vitro experiment of fabrication and dynamic culture was conducted. The results of this experiment showed that this method is adapted to the sterile work environment context, and that the cells seeded in the specimens were distributed homogeneously. This experience also demonstrated that the carbohydrate molds fabricated by additive manufacturing did not cause cell mortality in this context. However, minor damage was observed after several weeks of dynamic culture, and the cell viability rates were lower than expected because of suboptimal perfusion rates. This fabrication technique for cellular scaffolds is promising for the artificial aortic valves project, but improvements in terms of perfusion and preservation of physical integrity should be made.

Une structure dénombrable du premier ordre est dite homogène si tout isomorphisme entre deux sous-Structures finiment engendrées s’étend en un automorphisme de la structure globale.C’est équivalent à une propriété d’amalgamation des sous-Structures finiment engendrées, et les structures homogènes dénombrables sont aussi appelées limites de Fraïssé, en lien avec les travaux de Roland Fraïssé sur l’ordre des rationnels. Cette thèse concerne les groupes d’automorphismesdes structures homogènes, avec la question centrale suivante: est-Ce que le groupe automorphismes d’une structure homogène est universel pour la classe des groupes d’automorphismes de ces sous-Structures ? Nous répondons positivement à cette question pour les structures homogènesdans un langage relationnel et avec la propriété d’amalgamation libre, à l’aide d’une construction par tour assez similaire à une construction de Katetov et Uspenskij dans le cas de l’espace d’Urysohn. Avec des techniques similaires, nous obtenons toute sous-Structure dénombrable comme points fixes d’un automorphisme d’ordre fini pré-Déterminé. Cela nous permet par ailleurs d’étudier la complexité de la relation d’isomorphisme entre sous-Structures dénombrables, et de montrer qu’elle se réduit boreliennement à la relation de conjugaison dans le groupe d’automorphismes. Nous continuons avec les éléments d’ordre fini, en supposant de plus que les sous-Structures finies satisfont une version forte de la propriété d’extension de Hrushovski-Lascar-Herwig, et des arguments topologiques nous permettent alors de montrer que dans le groupe d’automorphismes tout élément est produit de quatre conjugués de certains éléments d’ordre fini. Nous montrons aussi des résultats similaires pour le groupe d’isométries de l’espace d’Urysohn,ou sa version bornée, la sphère d’Urysohn, en utilisant le fait que ces derniers sont très bien approximés par des espaces métriques rationnels. Enfin, revenant à la question de l’universalité du groupe automorphismes de la limite de Fraïssé, nous considérons la question plus fine de savoirsi toute sous-Structure dénombrable s’injecte de manière rigide, c’est-À-Dire de sorte chacun de ces automorphismes s’étende en un unique automorphisme de la limite de Fraïssé. D’abord, nous introduisons une construction de telle injections rigides dans le cas des graphes homogènes. Ensuite, nous modifions cette construction dans diverses classes de graphes orientés et de structures relationnelles homogènes, pour enfin la faire fonctionner dans un contexte très general de structures dans un langage relationnel fini et avec la propriété d’amalgamation libre. / A countable first-Order structure is called homogneous when each isomorphism between twofinitely generated substructures extends to an automorphism of the whole structure. This is equivalentto an amalgamation property of finitely generated substructures, and countable homogeneousstructures are also called Fraïssé limits, in connection to the work of Roland Fraïssé on theorder of rational numbers. The present thesis concerns automorphism groups of homogeneousstructures, with the following central question: is it the case that the automorphism group of a homogeneousstructure is universal for the class of automorphism groups of its substructures? Weanswer positively this question for homogeneous structures in a relational langage and with thefree amalgamation property, by using a construction rather similar to a construction of Katetov andUspenskij in the case of the Urysohn space.With similar techniques, we obtain any countable substructureas the set of fixed points of an automorphism of a given finite order. Besides, this allowsus to study the complexity of the isomorphism relation between countable substructures, and toshow that it Borel reduces to the conjugacy relation in the automorphism group. We continue withelements of finite order, assuming further that finite substructures satisfy a strong version of theHrushovski-Lascar-Herwig extension property, and topological arguments then allow us to showthat in the automorphism group any element is the product of four conjugates of certain elementsof finite order. We also show similar results for the isometry group of the Urysohn space, or itsbounded version, the Urysohn sphere, by using the fact that they are well approximated by rationalmetric spaces. Finally, concerning the question of the universality of the automorphism groupof a Fraïssé limit, we consider the finer question to know whether any countable substructure embedsin a rigid way, that is, in such a way that each of its automorphisms extends in a uniqueautomorphism of the Fraïssé limit. First, we introduce a construction of such rigid embeddings inthe case of homogeneous graphs. Then, we modify this construction in various classes of orientedgraphs and of homogeneous relational structures, ultimately to make it work in a very generalcontext of structures in a finite relational langage and with the free amalgamation property.

Abstract not available

Thesis (Ph. D.)--University of Missouri-Columbia, 2002. / Typescript. Vita. Includes bibliographical references (leaves 207-211). Also available on the Internet.

Thesis (Ph. D.)--University of Missouri-Columbia, 2002. / Typescript. Vita. Includes bibliographical references (leaves 207-211). Also available on the Internet.

Thesis (Ph. D.)--University of Texas at Austin, 2008. / Vita. Includes bibliographical references.

Thesis (Ph. D.)--University of Texas at Austin, 2004. / Supervisor: Katherine A. Liapi. Vita. Includes bibliographical references. Also available from UMI.

Actuellement, la majorité des architectures de processeurs sont fondées sur une mémoire partagée avec cohérence de caches. Des prototypes intégrant de grandes quantités de cœurs, reliés par une infrastructure de transmission de messages, indiquent que, dans un proche avenir, les architectures de processeurs vont probablement avoir ces caractéristiques. Ces deux tendances exigent que les processus s'exécutent en parallèle et rendent la programmation concurrente nécessaire. Cependant, la difficulté inhérente du raisonnement sur la concurrence peut rendre ces nouvelles machines difficiles à programmer. Nous explorons trois approches ayant pour but de faciliter la programmation concurrente. Nous proposons WFR-TM, une approche fondé sur la mémoire transactionnelle (TM), un paradigme de programmation concurrente qui utilise des transactions afin de synchroniser l'accès aux données partagées. Une transaction peut soit terminer (commit), rendant visibles ses modifications, soit échouer (abort), annulant toutes ses modifications. WFR-TM tente de combiner des caractéristiques désirables des TM optimistes et pessimistes. Une TM pessimiste n'échoue jamais aucune transaction; néanmoins les algorithmes existants utilisent des verrous pour exécuter de manière séquentielle les transactions qui contiennent des opérations d'écriture. Les algorithmes TM optimistes exécutent toutes les transactions en parallèle mais les terminent seulement si elles n'ont pas rencontré de conflit au cours de leur exécution. WFR-TM fournit des transactions en lecture seule qui sont wait-free, sans jamais exécuter d'opérations de synchronisation coûteuse (par ex. CAS, LL\SC, etc) ou sacrifier le parallélisme entre les transactions d'écriture. Nous présentons également Dense, une implémentation concurrente de graphe. Les graphes sont des structures de données polyvalentes qui permettent la mise en oeuvre d'une variété d'applications. Cependant, des applications multi-processus qui utilisent des graphes utilisent encore largement des versions séquentielles. Nous introduisons un nouveau modèle de graphes concurrents, permettant l'ajout ou la suppression de n'importe quel arc du graphe, ainsi que la traversée atomique d'une partie (ou de l'intégralité) du graphe. Dense offre la possibilité d'effectuer un snapshot partiel d'un sous-ensemble du graphe défini dynamiquement. Enfin, nous ciblons les futures architectures. Dans l'intérêt de la réutilisation du code il existe depuis quelques temps une tentative d'adaptation des environnements d'exécution de logiciel - comme par ex. JVM, l'environnement d'exécution de Java - initialement prévus pour mémoire partagée, à des machines sans cohérence de caches. Nous étudions des techniques générales pour implémenter des structures de données distribuées en supposant qu'elles vont être utilisées sur des architectures many-core, qui n'offrent qu'une cohérence partielle de caches, voir pas de cohérence du tout. / Though a majority of current processor architectures relies on shared, cache-coherent memory, current prototypes that integrate large amounts of cores, connected through a message-passing substrate, indicate that architectures of the near future may have these characteristics. Either of those tendencies requires that processes execute in parallel, making concurrent programming a necessary tool. The inherent difficulty of reasoning about concurrency, however, may make the new processor architectures hard to program. In order to deal with issues such as this, we explore approaches for providing ease of programmability. We propose WFR-TM, an approach based on transactional memory (TM), which is a concurrent programming paradigm that employs transactions in order to synchronize the access to shared data. A transaction may either commit, making its updates visible, or abort, discarding its updates. WFR-TM combines desirable characteristics of pessimistic and optimistic TM. In a pessimistic TM, no transaction ever aborts; however, in order to achieve that, existing TM algorithms employ locks in order to execute update transactions sequentially, decreasing the degree of achieved parallelism. Optimistic TMs execute all transactions concurrently but commit them only if they have encountered no conflict during their execution. WFR-TM provides read-only transactions that are wait-free, without ever executing expensive synchronization operations (like CAS, LL/SC, etc), or sacrificing the parallelism between update transactions. We further present Dense, a concurrent graph implementation. Graphs are versatile data structures that allow the implementation of a variety of applications. However, multi-process applications that rely on graphs still largely use a sequential implementation. We introduce an innovative concurrent graph model that provides addition and removal of any edge of the graph, as well as atomic traversals of a part (or the entirety) of the graph. Dense achieves wait-freedom by relying on light-weight helping and provides the inbuilt capability of performing a partial snapshot on a dynamically determined subset of the graph. We finally aim at predicted future architectures. In the interest of ode reuse and of a common paradigm, there is recent momentum towards porting software runtime environments, originally intended for shared-memory settings, onto non-cache-coherent machines. JVM, the runtime environment of the high-productivity language Java, is a notable example. Concurrent data structure implementations are important components of the libraries that environments like these incorporate. With the goal of contributing to this effort, we study general techniques for implementing distributed data structures assuming they have to run on many-core architectures that offer either partially cache-coherent memory or no cache coherence at all and present implementations of stacks, queues, and lists.

In broadband antenna applications, the antenna's cavity is usually loaded with absorbers to eliminate the backward radiation, but in doing so the radiation efficiency of the antenna is decreased. To enhance the radiation efficiency of the antennas EBG structures are used, but they operate over a narrow band. Uniform electromagnetic band gap (EBG) structures are usually periodic structures consisting of metal patches that are separated by small gaps and vias that connect the patches to the ground plane. The electrical equivalent circuit consists of a resonant tank circuit, whose capacitance is represented by the gap between the patches and inductance represented by the via. EBG structures are equivalent to a magnetic surface at the frequency of resonance and thus have very high surface impedance; this makes the EBG structures useful when mounting an antenna close to conducting ground plane, provided the antenna's currents are parallel to the EBG structure. Because EBG structures are known to operate over a very narrow band, they are not useful when used with a broadband antenna. Mushroom-like uniform EBG structures (that use vias) are compact in size have low loss, can be integrated into an antenna to minimize coupling effects of ground planes and increase radiation efficiency of the antenna. The bandwidth of an EBG structure is defined as the band where the reflection-phase from the structure is between +900 to -900. In this dissertation analysis of EBG structures is established using circuit analysis and transmission line analysis. Methods of increasing the bandwidth of EBG structures are explored, by cascading uniform EBG structures of different sizes progressively and vertically (stacked), and applications with different types of antennas are presented. Analyses in this dissertation are compared with previously published results and with simulated results using 3D electromagnetic tools. Validation of applications with antennas is carried by manufacturing prototypes and comparing measured performance with analysis and 3D electromagnetic simulations. The improvements in performance by using wideband progressive EBG and wideband stacked EBG structures are noted. / Ph. D.

Composite structures provide opportunities for weight reduction, material tailoring and integrating control surfaces with embedded transducers, which are not possible in conventional metallic structures. As a result there is a substantial increase in the use of composite materials in aerospace and other major industries, which has necessitated the need for structural health monitoring(SHM) of aerospace structures. In the context of SHM of aircraft structures, there are many areas, which are still not explored and need deep investigation. Among these, one of the major areas is the development of efficient damage models for complex composite structures, like stiffened structures, box-type structures, which are the building blocks of an aircraft wing structure. Quantification of the defect due to porosity and especially the methods for identifying the porous regions in a composite structure is another such area, which demands extensive research. In aircraft structures, it is not advisable for the structures, to have high porosity content, since it can initiate common defects in composites such as, delamination, matrix cracks etc.. In fact, there is need for a high frequency analysis to detect defects in such complex structures and also to detect damages, where the change in the stiffness due to the damage is very small. Lamb wave propagation based method is one of the efficient high frequency wave based method for damage detection and are extensively used for detecting small damages, which is essentially needed in aircraft industry. However, in order, to develop an efficient Lamb wave based SHM system, we also need an efficient computational wave propagation model. Developing an efficient computational wave propagation model for complex structures is still a challenging area. One of the major difficulty is its computational expense, when the analysis is performed using conventional FEM. However, for 1D And 2D composite structures, frequency domain spectral finite element method (SFEM), which are very effective in sensing small stiffness changes due to a defect in a structure, is one of the efficient tool for developing computationally efficient and accurate wave based damage models. In this work, we extend the efficiency of SFEM in developing damage models, for detecting damages in built-up composite structures and porous composite structure. Finally, in reality, the nature of variability of the material properties in a composite structure, created a variety of structural problems, in which the uncertainties in different parameters play a major part. Uncertainties can be due to the lack of good knowledge of material properties or due to the change in the load and support condition with the change in environmental variables such as temperature, humidity and pressure. The modeling technique is also one of the major sources of uncertainty, in the analysis of composites. In fact, when the variations are large, we can find in the literatures available that the probabilistic models are advantageous than the deterministic ones. Further, without performing a proper uncertain wave propagation analysis, to characterize the effect of uncertainty in different parameters, it is difficult to maintain the reliability of the results predicted by SFEM based damage models. Hence, in this work, we also study the effect of uncertainty in different structural parameters on the performance of the damage models, based on the models developed in the present work. First, two SFEM based models, one based on the method of assembling 2D spectral elements and the other based on the concept of coupling 2D and 1D spectral elements, are developed to perform high frequency wave propagation analysis of some of the commonly used built-up composite structures. The SFEM model developed using the plate-beam coupling approach is then used to model wave propagation in a multiple stiffened structure and also to model the stiffened structures with different cross sections such as T-section, I-section and hat section. Next, the wave propagation in a porous laminated composite beam is modeled using SFEM, based on the modified rule of mixture approach. Here, the material properties of the composite is obtained from the modified rule of mixture model, which are then used in SFEM to develop a new model for solving wave propagation problems in porous laminated composite beam. The influence of the porosity content on the parameters such as wave number, group speed and also the effect of variation in theses parameters on the time responses are studied first. Next, the effect of the length of the porous region (in the propagation direction) and the frequency of loading, on the time responses, is studied. The change in the time responses with the change in the porosity of the structure is used as a parameter to find the porosity content in a composite beam. The SFEM models developed in this study is then used in the context of wave based damage detection, in the next study. First ,the actual measured response from a structure and the numerically obtained response from a SFEM model for porous laminated composite beam are used for the estimation of porosity, by solving a nonlinear optimization problem. The damage force indicator (DFI) technique is used to locate the porous region in a beam and also to find its length, using the measured wave propagation responses. DFI is derived from the dynamic stiffness matrix of the healthy structure along with the nodal displacements of the damaged structure. Next, a wave propagation based method is developed for modeling damage in stiffened composite structures, using SFEM, to locate and quantify the damage due to a crack and skin-stiffener debonding. The method of wave scattering and DFI technique are used to quantify the damage in the stiffened structure. In the uncertain wave propagation analysis, a study on the uncertainty in material parameters on the wave propagation responses in a healthy metallic beam structure is performed first. Both modulus of elasticity and density are considered uncertain and the analysis is performed using Monte-Carlo simulation (MCS) under the environment of SFEM. The randomness in the material properties are characterized by three different distributions namely normal, Weibul and extreme value distribution and their effect on wave propagation, in beam is investigated. Even a study is performed on the usage of different beam theories and their uncertain responses due to dynamic impulse load. A study is also conducted to analyze the wave propagation response In a composite structure in an uncertain environment using Neumann expansion blended with Monte-Carlo simulation (NE-MCS) under the environment of SFEM. Neumann expansion method accelerates the MCS, which is required for composites as there are many number of uncertain variables. The effect of the parameters like, fiber orientation, lay-up sequence, number of layers and the layer thickness on the uncertain responses due to dynamic impulse load, is thoroughly analyzed. Finally, a probabilistic sensitivity analysis is performed to estimate the sensitivity of uncertain material and fabrication parameters, on the SFEM based damage models for a porous laminated composite beam. MCS is coupled with SFEM, for the uncertain wave propagation analysis and the Kullback-Leibler relative entropy is used as the measure of sensitivity. The sensitivity of different input variables on the wave number, group speed and the values of DFI, are mainly considered in this study. The thesis, written in nine chapters, presents a unified document on wave propagation in healthy and defective composite structure subjected to both deterministic and highly uncertain environment.