Th. doct.-ing.--Lille 1, 1977. N°: 199.

Expansion Deflection (ED) nozzles have long been considered as an alternative to conventional bell and conical designs. The ED nozzle has two primary benefits over such nozzle configurations, being shorter, and with the potential for altitude compensation. However, the difficulties involved with modelling the complex flow interactions within the nozzle type have thus far prevented the creation of a reliable method for its design and analysis. The work presented within this dissertation makes use of a combination of several different approaches to flow solution to provide a more complete analysis than previously achieved. The primary advance is the use of a CFD scheme to analyse the transonic throat region, the results of which are used as input to a Method of Characteristics based algorithm for the solution of the inviscid supersonic flow-field. This method is both efficient, and allows contour optimisation through the calculus of variations. The viscous flow region is treated in a partially or fully empirical manner, depending upon ambient pressure. The results of this analysis reveal several previously unidentified flow-field behaviours and design parameter interactions. Careful selection of the variables used to define the throat region is shown to be of considerable importance, as they effect not just the thrust produced, but also whether a real flow through such a nozzle may' exist. Secondly, the high probability of shock wave formation and interaction with the nozzle wall within length optimised ED nozzles is demonstrated. This has consequences for both the design and analysis of the ED nozzle. Overall, a framework is established which allows the flow behaviour and performance characteristics of the nozzle type to be estimated in greater detail than previously possible. The performance increments demonstrated over conventional bell nozzles include an approximately 25 to 50 percent length reduction for nozzles designed for solely high altitude operation, and a similar reduction in length plus a noticeable increase in low altitude thrust for trans-atmospheric nozzles. However, before application of the ED nozzle concept to a real system becomes possible, several areas of uncertainty must be addressed, and these are highlighted at the end of this dissertation.

L'objectif de cette thèse est l’implémentation d'algorithmes d'optimisation pour l'identification du comportement mécanique des matériaux. Ces outils sont considérés ici pour automatiser la recherche des paramètres des lois mécaniques implémentés dans un calcul en éléments finis (EF) et diminuer le nombre de calculs nécessaires pour résoudre le problème d’identification. La base de données expérimentale comprend des réponses mécaniques en flexion et en indentation instrumentée qui sont utilisés pour l’identification du comportement d’un biocomposite et d'une variété de métaux. Plusieurs approches d’optimisation ont été développées comprenant des méthodes déterministes, hybrides et stochastiques. Ces méthodes sont évaluées en termes d'efficacité, coût de calcul et précision de l’approche. Le couplage EF-outil d'optimisation s'est avéré efficace et incontournable pour prédire le comportement des matériaux étudiés notamment quand il s'agit d'un essai local. / The objective of this work is to implement optimization algorithms for the identification of the mechanical behavior of materials. The tools developed here are intended to automatic search of mechanical parameters that represent variables in a constitutive law implemented in Finite Element (FE) models. The advantage of the implemented tools is the reduction of computational time needed to solve the identification problem. The experimental database includes a variety of mechanical responses in bending and instrumented indentation that are used for the identification the behavior of a biocomposite and a variety of metals. Several optimization approaches are developed in the work including deterministic, stochastic and hybrid methods. These methods are evaluated in terms of efficiency; calculation costs and accuracy. The coupling FE-optimization tool proved to be effective and essential to predict the behavior of the studied materials, especially for instrument indentation.

In the last decade, efficient use of energy has become a topic of global significance, touching almost every area of modern life, including computing. From mobile to desktop to server, energy efficiency concerns are now ubiquitous. However, approaches to the energy problem are often piecemeal and focus on only one area for improvement. I argue that the strands of the energy problem are inextricably entangled and cannot be solved in isolation. I offer a high-level view of the problem and, building from it, explore a selection of subproblems within the field. I approach these with various levels of formality, and demonstrate techniques to make improvements on all levels.

No description available.

No description available.

Volumetric modulated arc therapy (VMAT) is a technique for the delivery of intensity modulated radiotherapy (IMRT) whereby the linear accelerator (linac) delivers dose continuously while rotating around the patient. VMAT has gained attention due to its ability to produce complex dose distributions, deliverable in a much shorter treatment time than IMRT. The purpose of this thesis was to investigate the clinical application of VMAT, and to identify any benefits over IMRT in the areas of treatment planning, delivery, and imaging. A VMAT planning strategy was developed which demonstrates that complex dynamic arc deliveries can be sequenced from static, IMRT-based control points. For prostate patients, the VMAT solution demonstrated superior sparing of critical structures compared to IMRT plans. A further comparison of VMAT and IMRT was performed with the development of an automated planning methodology, which aimed to reduce the impact of planner bias. Applied over a series of nasopharynx patients, the technique showed that VMAT achieved an improvement in parotid sparing compared to IMRT.To investigate the limitations on the delivery of VMAT plans, a software emulator was produced to accurately simulate linac motion. The emulator was used to determine 'ideal' linac parameters for a range of VMAT plans. Leaf speed was found to be a limiting factor for the achievable plan complexity, along with the availability of continuous variable dose rate (CVDR). For a commercial CVDR system, experiments confirmed the improved delivery efficiency, and an improvement in dosimetric accuracy compared to the binned dose rate (BDR) system.An independent dose calculation methodology was developed for VMAT, such that accurate pre-treatment plan QC can be performed. It was found that the accuracy of a Monte Carlo simulation was improved when accounting for the effects of realistic linac motion. Finally, the impact of MV scatter on simultaneously-acquired cone beam CT images was investigated, and a scatter correction methodology was developed and validated.This thesis shows that VMAT can offer an alternative to static-field IMRT, provided that knowledge of the limitations of dynamic linac motion are accounted for within planning. Results suggest that modern linac designs (i.e. faster MLC speed, and a higher, continuously-variable dose rate) are required to achieve robust delivery of complex plans. The workflow benefits of VMAT can also be optimised through the use of independent dose calculations incorporating delivery characteristics, and through the use of image guidance from CBCT scans acquired during treatment.

De nombreux problèmes d'optimisation issus du monde réel. notamment dans le domaine de la logistique, doivent faire face à beaucoup de difficultés, En effet, ils sont souvent caractérisés par des espaces de recherche vastes et complexes, de multiples fonctions objectif contradictoires, et une foule d'incertitudes qui doivent être prises en compte, Les métaheuristiques sont des candidats naturels pour résoudre ces problèmes, ce qui les rend préférables aux méthodes d'optimisation classiques, Toutefois, le développent de métaheuristiques efficaces résulte en un processus d'ingénierie complexe, Le coeur de ce travail réside en la conception, l'implémentation et l'analyse expérimentale de métaheuristiques pour l'optimisation multiobjectif ainsi que leurs applications à des problèmes logistiques de tournées et d'ordonnancement. Tout d'abord, une vue unifiée de ces approches est présentée, puis intégrée dans une plateforme logicielle dédiée à leur implémentation, ParadisEO-MOEO, Ensuite, plusieurs approches de coopération, combinant des métaheuristiques pour l'optimisation multiobjectif, sont proposées, Enfin, la question de la prise en compte de l'incertitude est abordée dans le context de l'optimisation multiobjectif. / Many real-world optimization problems, especially in the field of logistics, have to face a lot of difficulties, Indeed, they are often characterized by large and complex search'spaces, multiple conflicting objective functions, and a host of uncertainties that have ta be taken into account. Metaheuristics are natural candidates ta solve those problems and make them preferable to classical optimization methods, However, the development of efficient metaheuristics results in a complex engineering process, The core subject of this work lies in the design, implementation and experimental analysis of metaheuristics for multiobjective optimization, together with theu applications to logis tic problems from routing and scheduling, Firstly, a unitïed view of such approaches is presented and then integrated into a software framework for their implementation, ParadisEO-MOEO, Next, sorne cooperative approaches combining metaheuristics for multiobjective optimization are proposed, At last, the issue of uncertainty handling is discussed in the context of multiobjective optimization.

Le feu est à la cause des pertes en vies humaines et des dégâts matériels considérables. En cas d’incendie dans un système fermé tel que les bâtiments, navires, ou avions, un feu doit être compartimenté et sa propagation restreinte afin de sauver des vies et des biens en laissant le temps aux personnes d’évacuer. Afin d’y arriver, une barrière de feu ayant une très faible inflammabilité qui limiterait la propagation du feu doit être conçu et assemblé. L’objective de ce travail de thèse est de fournir les bases fondamentales pour la conception de barrières feu efficace quand ils subissent une contrainte thermique. En se basant sur la compréhension du comportement au feu de matériaux de références c.-à-d. le silicate de calcium, la mousse de mullite à cellules fermées, la mousse phénolique, et un revêtement intumescent à base de silicone, des tests feux avec un flux radiant externe basé sur la norme ISO 5660 du cône calorimètre sont effectués. En parallèle, des modèles numériques sont développés afin de prédire le comportement au feu de ces barrières feu. Leurs domaines d’application ainsi que leurs limites sont expliqués. Les propriétés physiques d’entrées requises pour alimenter les modèles sont obtenus soit par mesure directe par analyse thermique, soit de la littérature. Par ailleurs, des études de sensibilité sont effectués afin d’identifier les paramètres essentiels qui contrôlent le comportement au feu des matériaux de référence. Les modèles numériques sont ensuite appliqués à la conceptualisation de nouvelles barrières feu grâce à la méthodologie basée sur le "performance-based design" ainsi que l’optimisation. Enfin, après les étapes de conception et les études de sensibilité, les règles fondamentales sur la conceptualisation de barrières de feu pour un scenario feu précis en accord avec des normes définies sont énoncées. / Fire causes injuries, the loss of lives and property. In the case of a fire in an enclosed system such as buildings, naval ships or aircraft, the fire should be compartmentalized and restricted from spreading from one point to another in order to save lives and property as well as give people enough time to evacuate. To accomplish this, fire barriers exhibiting low flammability and limiting fire spread need to be designed and assembled. The aim of this PhD is to provide with the fundamentals on how to design efficient fire barriers when subjected to a thermal stress. Based on understanding the fire behavior of selected reference fire retardant materials i.e. calcium silicate, closed-cell mullite foam, phenolic foam and a silicone-based intumescent coating, fire tests using the external radiant heat flux from the ISO 5660 cone calorimeter are conducted. At the same time, numerical models are developed to predict the fire behavior of these fire barriers and their applicability as well as limitations are explained. The input materials properties to run the numerical models are obtained from both direct measurements and from the literature. In addition, sensitivity studies are conducted in order to identify the governing parameters that control the fire behavior of the reference fire resistant materials. The numerical models are then used for the conceptualization of innovative fire safe materials based on the performance-based design process and optimization. Finally, based on the sensitivity studies and the conceptualization process, fundamental rules on how to make an efficient fire barrier in order to meet certain requirements in a given fire scenario are clearly stated.

L’électronique de puissance nécessite des interconnexions à faible inductance et des températures de fonctionnement en dessous de certaines valeurs critiques. Les busbars laminés présentent une inductance très faible en comparaison avec les câbles. De plus, leur forme aplatie permet d’augmenter la surface d’échange avec l’environnent garantissant des bonnes performances thermiques. Dans cette thèse nous proposons une comparaison entre différentes méthodes de calcul d’inductance. Le but est d’en privilégier une en fonction de la complexité du problème à résoudre. La valeur d’inductance à calculer correspond à la boucle de commutation ayant lieu dans les convertisseurs de puissance. Des règles de conception permettant de minimiser la valeur de cette boucle sont aussi présentées. Concernant le calcul thermique, on propose une approche basée sur les éléments finis 2D. La superposition de plusieurs plaques bidimensionnelles permet de modéliser des busbars laminés de façon précise et plus rapide qu’une étude tridimensionnelle. Les conditions aux limites sont appliquées en fonction de la superposition des plaques. Grâce à la rapidité de cette approche, des optimisations topologiques peuvent être appliquées. Des algorithmes d’optimisation adaptés aux besoins industriels ont été développés pour la minimisation de l’inductance et de la température maximale des busbars. Ces algorithmes présentent de bonnes performances et sont utilisés lors des optimisations mono et multi-objectif. / Power electronics needs low inductance interconnections and operating temperatures below certain limits. Laminated busbars present low inductance compared with cables. Moreover, their flat geometry increases the heat exchange surface with the environment ensuring good thermal performances. In this PhD thesis we propose a comparison between different inductance calculation methods. The objective is to favor one according to the complexity of the problem to be solved. The inductance value is calculated during the switching loop taking place in the power converters. Several design rules are presented in order to minimize the inductance of the loop. Regarding the thermal management, we present an approach based on the 2D finite element method. The overlapping of the two-dimensional plates allows us to model the whole laminated busbar and faster way than a tridimensional study without losing precision. The boundary conditions are applied according to the overlapping between layers. Thanks to this approach, topology optimization techniques can be applied. Algorithms adapted to the industrial needs have been developed for the minimization of the inductance and the maximum temperature in busbars. These algorithms show good performances and are used in mono and multi-objective optimizations.