La durée de vie des pièces mécaniques en contact est fortement affectée par la présence d'hétérogénéités dans le matériau, comme des renforts (fibres, particules), des précipités, des porosités, ou encore des fissures. Des hétérogénéités dures et de formes complexes peuvent créer des surcontraintes locales, initiatrices de fissures par fatigue à proximité de la surface de contact. Une analyse quantitative des surcontraintes créées par les hétérogénéités est nécessaire à la compréhension des mécanismes d'endommagement. Cette étude s'applique à des roulements de ligne d'arbre qui font partie des éléments critiques de moteurs en aéronautique. Elle vise à déterminer précisément la distribution du champ de pression sur l'aire effective de contact et à prédire le profil et l'évolution des champs de contraintes/déformations à chaque passage de la charge sur un volume élémentaire représentatif prenant en compte le gradient de dureté, la présence de carbures et l'existence des contraintes initiales d'origine thermochimique. Une partie de l’étude est consacrée au développement d’un solveur du problème de contact roulant élasto-plastique avec présence d’hétérogénéité par les méthodes semi analytiques assurant un excellent gain en temps et ressources de calculs. Ensuite, un algorithme homogénéisation a été conçu pour analyser le comportement effectif d’un massif élasto-plastique hétérogène sous indentation. Enfin une partie expérimentale est dédiée à la caractérisation microstructurale des aciers étudiés dans le but de déterminer leurs propriétés. Les analyses des résultats de cette étude concourent à soutenir que bien que les inclusions de particules non métalliques soient responsables de la haute résistance de ces matériaux, certaines d’entre elles (celles de longueur dépassant les dizaines de micromètre ou celles qui forment des chaines dans une direction particulière) deviennent, au cours des cycles de fatigue, les principales sources d’endommagement depuis l’échelle locale jusqu’à la rupture globale de la structure. / The lifetime of contacting mechanical parts is strongly affected by the presence of heterogeneities in their materials, such as reinforcements (fibers, particles), precipitates, porosities, or cracks. Hard heterogeneities having complex forms can create local overstress that initiating fatigue cracks near the contact surface. The presence of heterogeneities influences the physical and mechanical properties of the material at microscopic and macroscopic scales. A quantitative analysis of the over-stresses generated by heterogeneities is necessary to the comprehension of the damage mechanisms. The present study is applied to rolling bearings which are the critical elements of the aero-engine's mainshaft. The performance required for these bearings, led SKF Aerospace to introduce a new technology of hybrid bearing with ceramic rolling elements on high-strength steels having experienced a double surface treatment (carburizing followed by nitriding). The study aims to precisely determine the pressure field distribution on the effective contact area and to predict the profile and the evolution of the stress/strain fields at each loading cycle on a representative elementary volume that takes into account the gradient of hardness, the presence of carbides and the existence of an initial compressive stress from thermochemical origin. A major part of this study is devoted to develop a heterogeneous elastic-plastic rolling contact solver, by semi-analytical methods ensuring an excellent saving of calculation time and resources. Thereafter, a homogenization algorithm was built to analyze the effective behavior of a heterogeneous elastic-plastic half-space subjected to an indentation loading. Finally, an experimental part is dedicated to the microstructure characterization of the studied steels with intent to determine their properties. A description of the carbides behavior inside the matrix during micro-tensile tests was carried out under SEM in-situ observation. In the scheme of all analyses conducted in the present work, it can be argued that, although the heterogeneities (such as carbides or nitrides) are responsible for the high resistance of the studied materials, some of them (those whose length exceeds tens of micrometer or those which form stringers in a particular direction) become, over fatigue cycles, the main sources of damage, from their local scale up to the macroscopic failure of the structure.

Dans cette thèse, nous nous intéressons au problème de la détection d’objets faiblement supervisée. Le but est de reconnaître et de localiser des objets dans les images, n’ayant à notre disposition durant la phase d’apprentissage que des images partiellement annotées au niveau des objets. Pour cela, nous avons proposé deux méthodes basées sur des modèles différents. Pour la première méthode, nous avons proposé une amélioration de l’approche ”Deformable Part-based Models” (DPM) faiblement supervisée, en insistant sur l’importance de la position et de la taille du filtre racine initial spécifique à la classe. Tout d’abord, un ensemble de candidats est calculé, ceux-ci représentant les positions possibles de l’objet pour le filtre racine initial, en se basant sur une mesure générique d’objectness (par region proposals) pour combiner les régions les plus saillantes et potentiellement de bonne qualité. Ensuite, nous avons proposé l’apprentissage du label des classes latentes de chaque candidat comme un problème de classification binaire, en entrainant des classifieurs spécifiques pour chaque catégorie afin de prédire si les candidats sont potentiellement des objets cible ou non. De plus, nous avons amélioré la détection en incorporant l’information contextuelle à partir des scores de classification de l’image. Enfin, nous avons élaboré une procédure de post-traitement permettant d’élargir et de contracter les régions fournies par le DPM afin de les adapter efficacement à la taille de l’objet, augmentant ainsi la précision finale de la détection. Pour la seconde approche, nous avons étudié dans quelle mesure l’information tirée des objets similaires d’un point de vue visuel et sémantique pouvait être utilisée pour transformer un classifieur d’images en détecteur d’objets d’une manière semi-supervisée sur un large ensemble de données, pour lequel seul un sous-ensemble des catégories d’objets est annoté avec des boîtes englobantes nécessaires pour l’apprentissage des détecteurs. Nous avons proposé de transformer des classifieurs d’images basés sur des réseaux convolutionnels profonds (Deep CNN) en détecteurs d’objets en modélisant les différences entre les deux en considérant des catégories disposant à la fois de l’annotation au niveau de l’image globale et l’annotation au niveau des boîtes englobantes. Cette information de différence est ensuite transférée aux catégories sans annotation au niveau des boîtes englobantes, permettant ainsi la conversion de classifieurs d’images en détecteurs d’objets. Nos approches ont été évaluées sur plusieurs jeux de données tels que PASCAL VOC, ImageNet ILSVRC et Microsoft COCO. Ces expérimentations ont démontré que nos approches permettent d’obtenir des résultats comparables à ceux de l’état de l’art et qu’une amélioration significative a pu être obtenue par rapport à des méthodes récentes de détection d’objets faiblement supervisées. / In this dissertation we address the problem of weakly supervised object detection, wherein the goal is to recognize and localize objects in weakly-labeled images where object-level annotations are incomplete during training. To this end, we propose two methods which learn two different models for the objects of interest. In our first method, we propose a model enhancing the weakly supervised Deformable Part-based Models (DPMs) by emphasizing the importance of location and size of the initial class-specific root filter. We first compute a candidate pool that represents the potential locations of the object as this root filter estimate, by exploring the generic objectness measurement (region proposals) to combine the most salient regions and “good” region proposals. We then propose learning of the latent class label of each candidate window as a binary classification problem, by training category-specific classifiers used to coarsely classify a candidate window into either a target object or a non-target class. Furthermore, we improve detection by incorporating the contextual information from image classification scores. Finally, we design a flexible enlarging-and-shrinking post-processing procedure to modify the DPMs outputs, which can effectively match the approximate object aspect ratios and further improve final accuracy. Second, we investigate how knowledge about object similarities from both visual and semantic domains can be transferred to adapt an image classifier to an object detector in a semi-supervised setting on a large-scale database, where a subset of object categories are annotated with bounding boxes. We propose to transform deep Convolutional Neural Networks (CNN)-based image-level classifiers into object detectors by modeling the differences between the two on categories with both image-level and bounding box annotations, and transferring this information to convert classifiers to detectors for categories without bounding box annotations. We have evaluated both our approaches extensively on several challenging detection benchmarks, e.g. , PASCAL VOC, ImageNet ILSVRC and Microsoft COCO. Both our approaches compare favorably to the state-of-the-art and show significant improvement over several other recent weakly supervised detection methods.

O planejamento da fabricação de peças de caldeiraria para hidrelétricas tais como caixa espiral, tubo de sucção e conduto forçado, necessita de bons conhecimentos em caldeiraria e muita experiência na área devido à complexidade dos cálculos, tornando-se uma tarefa trabalhosa e com muitas possibilidades de erros. A seqüência de operações, no entanto, é altamente repetitiva facilitando o desenvolvimento de um sistema computacional de apoio. O trabalho proposto a seguir tem como objetivo apresentar um sistema desenvolvido como um aplicativo em software CAD para auxiliar as tarefas de projetar e fabricar caixa espiral, tubo de sucção e conduto forçado. O sistema colocou o conhecimento técnico da engenharia industrial de hidrogeradores de uma empresa de bens de capital ao alcance de projetistas desta área, de forma a aumentar a produtividade nestas, e em outras atividades similares, minimizando as possibilidades de erro e padronizando os procedimentos de cálculos. Os resultados foram, entre outros, diminuição do tempo de geração das informações para o processo de fabricação, tais como croquis de delineamento e desenhos de montagem, acima de 90% e perda zero de peças na fabricação. / The manufacturing process of hydroelectric weld parts, such as spiral cases, draft tubes and penstocks, demands good knowledge and experience in the area due to the calculations needed, revealing to be a difficult work with a high risk of errors. However, the operation sequence is highly repetitive which enables the development of a computational support system. This thesis has the purpose to present a system developed as an application in CAD software to aid the projecting and manufacturing activities of the spiral boxes, draft tubes and penstocks. Such system was disposed to the technical body of an industrial engineering department maintained by a heavy equipment company which manufactures hydro generators, so professionals were able to increase their productivity, being the system also used in similar areas, minimizing possibilities of mistakes and standardizing calculations procedures. Results were, among others, decrease of time needed to develop process information, such as sketch outlines and assembly drawings, above 90% and zero losses on parts production.

O planejamento da fabricação de peças de caldeiraria para hidrelétricas tais como caixa espiral, tubo de sucção e conduto forçado, necessita de bons conhecimentos em caldeiraria e muita experiência na área devido à complexidade dos cálculos, tornando-se uma tarefa trabalhosa e com muitas possibilidades de erros. A seqüência de operações, no entanto, é altamente repetitiva facilitando o desenvolvimento de um sistema computacional de apoio. O trabalho proposto a seguir tem como objetivo apresentar um sistema desenvolvido como um aplicativo em software CAD para auxiliar as tarefas de projetar e fabricar caixa espiral, tubo de sucção e conduto forçado. O sistema colocou o conhecimento técnico da engenharia industrial de hidrogeradores de uma empresa de bens de capital ao alcance de projetistas desta área, de forma a aumentar a produtividade nestas, e em outras atividades similares, minimizando as possibilidades de erro e padronizando os procedimentos de cálculos. Os resultados foram, entre outros, diminuição do tempo de geração das informações para o processo de fabricação, tais como croquis de delineamento e desenhos de montagem, acima de 90% e perda zero de peças na fabricação. / The manufacturing process of hydroelectric weld parts, such as spiral cases, draft tubes and penstocks, demands good knowledge and experience in the area due to the calculations needed, revealing to be a difficult work with a high risk of errors. However, the operation sequence is highly repetitive which enables the development of a computational support system. This thesis has the purpose to present a system developed as an application in CAD software to aid the projecting and manufacturing activities of the spiral boxes, draft tubes and penstocks. Such system was disposed to the technical body of an industrial engineering department maintained by a heavy equipment company which manufactures hydro generators, so professionals were able to increase their productivity, being the system also used in similar areas, minimizing possibilities of mistakes and standardizing calculations procedures. Results were, among others, decrease of time needed to develop process information, such as sketch outlines and assembly drawings, above 90% and zero losses on parts production.

Cette thèse s’intéresse au fonctionnement cyclique multi-produits des ateliers de traitement de surface, et au problème d’ordonnancement associé (HSP), caractérisé par des contraintes fortes et atypiques, dont certaines sont liées aux ressources de transport. Dans le cas de productions en grandes séries, une commande cyclique de ces systèmes est particulièrement adaptée, permettant notamment de réduire la combinatoire de résolution, et sous réserve que les ratios de produits soient connus à l’avance. Notre objectif est de trouver le meilleur ordonnancement des tâches de traitement et de transport en un temps raisonnable. Pour cela, nous proposons une première approche, basée sur un modèle linéaire et une méthode de résolution arborescente de type séparation et évaluation. Nous présentons des modélisations pour différentes extensions du problème dit de base et nous fournissons des exemples illustratifs et des résultats sur des benchmarks. Par la suite et compte tenu de l’analyse de la littérature relative aux ordonnancements cycliques mono-produit et multi-produits, nous proposons tout d’abord une heuristique dédiée au cas multi-produits étudié, et basée sur un algorithme de liste. Avec ce dernier, nous obtenons un ordonnancement cyclique dont le degré du cycle n’est pas fixé au préalable. Enfin, nous présentons une deuxième modélisation approchée sous la forme d’un algorithme génétique pour résoudre un HSP 2-cyclique. Ces différents modèles sont validés par des tests sur des benchmarks de la littérature pour lesquels nous avons obtenus des résultats prometteurs. Nous terminons par une analyse critique des avantages et inconvénients des modèles élaborés et par quelques propositions de perspectives pour ce travail. / In this thesis, we study the Cyclic Hoist Scheduling Problem (CHSP) in automated electroplating lines, when a mass production must be achieved. The CHSP is characterized by specific constraints related to processing and transport resources. To solve it in a multi-parts context, we first elaborate a 2-degree cyclic model and an associated branch and bound algorithm. Then we extend it to more complex configurations. Then, we develop a dedicated heuristic to find a feasible repetitive sequence of hoist moves that minimizes the cycle time, without a priori fixing the cycle degree. Comparisons with existing algorithms are presented to show the efficiency of the proposed heuristic. To reduce the cycle time, we integrate in the general heuristic an algorithm with a set of Minimum Part Set (MPS) configurations’. This one allows us to find the best order in which jobs should be introduced into the line. Finally, we describe a genetic algorithm approach to find a schedule which can reach the optimal 2-cycle. We finally discuss the interest of those various models, based on the promising results obtained and we provide some perspectives which could be explored.

本研究旨在藉由Brandenburge & Nalebuff (1995)的『價值網絡』(Value Net)模型來探討企業之新創事業如何在所屬的『價值網絡』中，依其策略性定位採取與同一網絡中的其他價值競逐者或合作或競爭的合縱連橫，以遂行其為事業創造顧客價值之目的。 企業在其所屬事業的『價值網絡』中，或藉由靈活地調整自身策略定位，或藉由提昇自己的附加價值、壓抑對手的附加價值，或藉由改變競爭規則、競爭範疇等，以形塑其事業的競爭優勢。 研究發現，新創事業即使在『價值網絡』中具有相對於其它競爭者更不對稱的優勢，若無法為事業創造高於所投入成本的價值，並將所獲得的價值價格化，則所具備的是不具價值的優勢。亦即，新創事業若無獲利能力，則再好的事業構想、再好的商業模式，也無助於新創事業的超越失敗。 企業對於已難以超越失敗的新創事業之進退所採取的策略選擇，固然會受到客觀的退出障礙及理性的策略性考量影響；但，企業對於所屬事業前景的主觀看法及經營者的心理素質更會左右對事業殘局的評估及其因應策略的選擇。 經營具充滿不確定性風險的新創事業並非參加一場僅是一次行動的比賽。因此，企業除了要熟稔其事業在『價值網絡』的定位、靈活運用各項要素以遂行其創造顧客價值的優勢外；對於不如預期之新創事業，更應理性地面對事實及客觀地評估與選擇或進或退之策略，避免因無謂的資源消耗而陷入僵局；也避色因陷入僵局而使得企業失却其它再興的新局！

臺灣自行車產業自1970年代開始，逐漸開始萌芽，一直到1990年代之前，臺灣一直是物廉價美的代工最佳地區。自從1990年起臺灣自行車產業陸續往大陸西移，幸好臺灣自行車業者還是把研發基地留在本國，正因為如此不管生產基地往哪裡擴張，臺灣本廠一直能保有營運總部的根基。隨著產業的壯大，如何能將研發的實力及銷售的利潤留在自家內，發展品牌變成熱門的話題。但由於自行車由許多零配件供應衛星工廠協同生產，這些規模大小不一的廠家，在發展品牌的困難程度遠較自行車整車高出很多。有部分廠商因其經濟規模較大，在其產品橫向發展上非常成功，在自有品牌的行銷上確實有亮麗的成績，如瑪吉斯輪胎、建大輪胎等。本研究針對20家廠商做自有品牌及品牌聯盟之問卷調查，及對四家自行車零配件代表性廠商做深入產業品牌經營訪談研究。根據其對發展自有品牌及品牌聯盟之現況，分析業者未來經營品牌成功之關鍵做法。 臺灣自行車零配件業者雖然在研發及製造實力在國際上位於領導地位，但是品牌實力卻落後於別人，以致開發的報酬的相對不高。現在由於臺灣本國市場、中國大陸及其他新興市場的崛起，此時台灣自行車零配件最有機會建立品牌，培養忠誠客戶，把臺灣業者最擅長的服務行銷成為最佳的市場競爭優勢，掌握產業主導，開創台灣自行車業者品牌價值新頁。 關鍵字: 代工，品牌，自有品牌，品牌聯盟，自行車，自行車零配件

Through continued innovation and growth, traveling wave tube amplifiers (TWTAs) remains the microwave power amplifiers of choice in a wide range of high power microwave and millimeter-wave applications specifically for aerospace applications with the volume, weight, bandwidth and power constraints. These advances can be credited to device innovation, improved modeling and design and development of advanced materials and construction techniques. This thesis aims at advancing the present technology of TWTs with coupled resonator slow-wave structures (SWSs) by a combination of device innovation, development of enhanced analytical and field analysis codes and understanding gained through improved modeling, simulation and experimentation. In a TWT, the SWS that slows the RF wave velocity down to near the electron beam velocity for interaction with the electron beam primarily determines the microwave performances of the tube. As compared to helix SWS, the coupled resonator SWS is capable of handling high peak and average powers with higher efficiency and TWTs based on these SWS are well suited for air-borne or space-borne radar systems and the major focus of this thesis is on the analysis and design of coupled resonator SWSs. As a part of this thesis, improved analytical codes based on quasi-TEM analysis and equivalent circuit analysis have been developed. The technical formulation is explained and the improvements made for enhanced accuracy and for incorporation of different types of coupled resonator SWSs detailed. Using these models new variants of coupled resonator SWSs have been investigated. The SWSs proposed are the ladder-core inverted slot mode SWS and the inductively loaded inter digital SWS (ILID-SWS). The possibility of achieving both coalesced mode design that gives wide bandwidth and multi beam design that improves the peak power and gain using rectangular ILID-SWS is presented. The properties of these proposed SWSs have been compared with the existing SWSs and found to give superior performance. Also an improved modeling and simulation technique using 3-D electromagnetic codes has been proposed and the conventional cold test measurement procedure has been modified for more accurate results. Numerous illustrative examples are presented throughout the thesis highlighting the analytical model and simulation code validation with experimental results. The experimentations have been carried out on the real SWS model that have been fabricated and assembled. Further, the contribution of the thesis is towards the development of a field analysis model for analysis of a corrugated waveguide SWS, based on the coupled integral equation technique (CIET), which is a combination of mode matching technique (MMT) and method of moments. The technical formulation and computational methodology employed in the model are explained and some of the most important aspects of implementation like the handling of singularities and choice of parameters controlling the accuracy is discussed. The accuracy and speed of the code is demonstrated by comparing CIET with MMT and 3-D electro magnetic simulators based on finite difference time domain (FDTD) method and finite element method (FEM). The CIET code developed is quite faster than the existing numerical methods and helps in solving the convergence problem associated with the MMT.

The Pebble Bed Modular Reactor (PBMR) is one of the most technologically advanced developments in South Africa. In order to build a commercially viable demonstration power plant, all the specifically and uniquely designed equipment must first be qualified. All the prototype equipment is tested at the Helium Test Facility (HTF) at Pelindaba. One of the largest components that are tested is the Core Unloading Device (CUD). The main function of the CUD is to unload fuel from the bottom of the reactor core to enable circulation of the fuel core. The CUD housing vessel forms part of the reactor pressure boundary. Pebble-directing valves and other moving machinery are installed inside its machined inner surface. It is essential that the interior surfaces of the CUD are case hardened to provide a corrosion- and wear-resistant layer. Cold welding between the moving metal parts and the machined surface must also be prevented. Nitriding is a case hardening process that adds a hardened wear- and corrosion-resistant layer that will also prevent cold welding of the moving parts in the helium atmosphere. Only a few nitriding furnaces exist that can house a forging as large as the CUD of the PBMR. Commercial nitriding furnaces in South Africa are all too small and have limited flexibility in terms of the nitriding process. The nitriding of a vessel as large as the CUD has not yet been carried out commercially. The aim of this work was to design and develop a custom-made nitriding plant to perform the nitriding of the first PBMR/HTF CUD. Proper process control is essential to ensure that the required nitrided case has been obtained. A new concept for a gas nitriding plant was developed using the nitrided vessel interior as the nitriding process chamber. Before the commencement of detail design, a laboratory test was performed on a scale model vessel to confirm concept feasibility. The design of the plant included the mechanical design of various components essential to the nitriding process. A special stirring fan with an extended length shaft was designed, taking whirling speed into account. Considerable research was performed on the high temperature use of the various components to ensure the safe operation of the plant at temperatures of up to 600°C. Nitriding requires the use of hazardous gases such as ammonia, oxygen and nitrogen. Hydrogen is produced as a by-product and therefore safety was the most important design parameter. Thermohydraulic analyses, i.e. heat transfer and pressure drop calculations in pipes, were also performed to ensure the successful process design of the nitriding plant. The nitriding plant was subsequently constructed and operated to verify the correct design. A large amount of experimental and operating data was captured during the actual operation of the plant. This data was analysed and the thermohydraulic analyses were verified. Nitrided specimens were subjected to hardness and layer thickness tests. The measured temperature of the protruding fan shaft was within the limits predicted by Finite Element Analysis (FEA) models. Graphs of gas flow rates and other operation data confirmed the inverse proportionality between ammonia supply flow rate and measured dissociation rate. The design and operation of the nitriding plant were successful as a nitride layer thickness of 400 μm and hardness of 1 200 Vickers hardness (VHN) was achieved. This research proves that a large pressure vessel can successfully be nitrided using the vessel interior as a process chamber. / Thesis (M.Ing. (Mechanical Engineering))--North-West University, Potchefstroom Campus, 2010.

Access to abstract restricted until 05/2017 / Access to thesis restricted until 05/2017 / School of Music