Cost effective high volume manufacture of solid oxide fuel cells (SOFCs) is a major challenge for commercial success of these devices. More than fifteen processing methods have been reported in the literature, many of which could be used in various combinations to create the desired product characteristics. For some of these processes, high volume manufacturing experience is very limited or non-existent making traditional costing approaches inappropriate. Additionally, currently available cost models are limited by a lack of incorporation of device performance requirements. Therefore, additional modeling tools are needed to aid in the selection of the appropriate processing techniques prior to making expensive investment decisions. This project describes the development of a SOFC device performance model and a manufacturing process tolerance model. These models are then linked to a preliminary cost model; creating a true multi-process, performance based cost model that permits the comparison of manufacturing cost for different combinations of three processing methods. The three processing methods that are investigated are tape casting, screen printing, and sputtering. . This model is capable of considering production volume, process tolerance and process yield, in addition to the materials and process details. Initial comparisons were performed against processes used extensively within the solid oxide fuel cell industry and the cost results show good agreement with this experience base. Sensitivity of manufacturing costs to SOFC performance requirements such as maximum power density and operation temperature are also investigated.

KISS (Kinematic Support System) is a locally developed gait analysis system at Middle East Technical University (METU), and the performance of the system was evaluated before as a whole. However, such evaluations do not differentiate between the efficacy of the data acquisition system and the model-based gait analysis methodology. In this thesis, kinematic results of the KISS system will be compared with those of the Ankara University based commercial VICON (Oxford Metrics Ltd., Oxford, UK) system, in view of evaluating the performance of data acquisition system and the gait analysis methodology separately. This study is expected to provide guidelines for future developments on the KISS system.

KISS (Kinematic Support System) is a locally developed gait analysis system at Middle East Technical University (METU), and the performance of the system was evaluated before as a whole. However, such evaluations do not differentiate between the efficacy of the data acquisition system and the model-based gait analysis methodology. In this thesis, kinematic results of the KISS system will be compared with those of the Ankara University based commercial VICON (Oxford Metrics Ltd., Oxford, UK) system, in view of evaluating the performance of data acquisition system and the gait analysis methodology separately. This study is expected to provide guidelines for future developments on the KISS system.

Pointer et sélectionner une cible avec un curseur est une interaction omniprésente dans l'utilisation des ordinateurs. La sélection peut néanmoins être difficile à réaliser si le contrôle du curseur n'est pas aisé, ou encore si la cible est petite ou distante. Aussi, faciliter la sélection de cibles est un thème essentiel et actif en Interaction Homme-Machine (IHM) et les techniques conçues sont nombreuses.Notre travail de recherche se focalise sur l'étude des techniques d'expansion de cibles. Celles-ci facilitent le pointage et la sélection en allouant à une cible une plus grande zone active, sans toutefois modifier la cible elle-même. L'algorithme d'expansion d'une technique détermine la zone active étendue de chaque cible, tandis que l'aide visuelle de la technique fournit à l'utilisateur les informations nécessaires pour exploiter l'expansion lors du pointage et sélection. Notre travail est dédié à l'aide visuelle des techniques d'expansion de cibles en abordant leur conception et leur impact sur les performances.Nous proposons dans un premier temps un espace de conception des aides visuelles. Cet espace est accompagné d'une notation matricielle de description des aides visuelles et permet une classification des techniques d'expansion existantes. Son pouvoir génératif s'illustre par la création de huit techniques d'expansion, que nous évaluons au travers de deux expériences en laboratoire. Les résultats de ces deux expériences nous permettent de formuler un ensemble de recommandations pour les concepteurs de techniques d'expansion.Nous avançons ensuite un modèle conceptuel de performance des techniques d'expansion. Ce modèle s'appuie sur la notion de pertinence de l'information contenue dans l'aide visuelle des techniques. Il considère une tâche de sélection sous l'angle cognitif en la décomposant en trois phases déterminées par les buts de l'utilisateur : la phase préliminaire au geste, la phase de transfert du curseur vers la cible et la phase de validation de la sélection. Nous confrontons ce modèle de performance avec des résultats expérimentaux de la littérature et d'une troisième expérience en laboratoire.Nos contributions sont appliquées à l'interaction chirurgien-ordinateur en contexte per-opératoire, c'est-à-dire en bloc opératoire durant une intervention chirurgicale. En particulier, l'entreprise partenaire de cette thèse CIFRE, Aesculap, intègre dans la prochaine version du produit OrthoPilot, un système informatique pour la chirurgie orthopédique, une des techniques d'expansion étudiées. Nous présentons également Medical TapTap, une nouvelle technique d'interaction gestuelle au pied conçue pour la validation de la sélection en contexte per-opératoire. / Pointing to/ Selecting targets is an elementary task universally present in graphical user interfaces (GUI). This task can be difficult to perform if the control of the cursor is not easy or if the target is far away or small. Thus facilitating target selection is a fundamental and active research topic in Human-Computer Interaction (HCI) and researchers have proposed numerous targeting assistance techniques.Our research focuses on targeting assistance techniques that allocate larger activation areas to targets. Such target expansion techniques rely on two basic elements: the expansion algorithm and the visual aid. The expansion algorithm distributes partly or wholly the free space among the targets. The visual aid presents the resulting target expansion to the users. Our work is dedicated to the visual aid that enables the users to take full advantage of the target expansion technique during the pointing/selection tasks.We first propose a three-axes design space for visual aid mechanisms. We further define a matrix-based notation for concisely describing a target expansion technique along the three design axes. We provide an analytical exploration of the design space by classifying existing target expansion techniques and by designing eight novel target expansion techniques, thus demonstrating the generative power of the design space. We also provide an experimental exploration of the design space by conducting two in-lab experiments. Based on the experimental results, we build a set of design recommendations.We then put forward a conceptual predictive model of performance. The model relies on a systematic analysis of the relevance of the visual aid provided by a target expansion technique based on the three goal-oriented phases of a selection task: the starting phase to initiate the movement towards the target, the transfer phase to bring the cursor into the goal target and the selection validation phase. To test the model we consider experimental results of the literature and of a third conducted in-lab experiment.Our contributions are applied to the field of Augmented Surgery and in particular interaction with a distant screen during a surgery in the operating theatre. As part of a Aesculap-CIFRE thesis, the next version of the product Aesculap’s OrthoPilot® Navigation System for orthopaedic surgery will include a target expansion technique for facilitating target selection by the surgeon in the operating theatre. We also provide a new foot gesture-based technique, namely Medical TapTap, for the validation of selection in the operating theatre.

Cette thèse se propose de contribuer à la compréhension de la problématique générale de la performance, ceci par l’exploration du contenu de la vision stratégique appliquée au contexte des propriétaires-dirigeants de PME. Notre thèse émerge ainsi des observations du terrain et de son évolution, notamment les difficultés rencontrées par certaines PME pour survivre et à contrario les facteurs favorisants la réussite et la performance de certaines d’entre elles. Cette recherche s’articule ainsi autour d’un processus continu d’aller-retour entre la littérature et le terrain. Les PME représentent un des terreaux les plus propices à la recherche sur le concept de vision stratégique et sa relation à la performance. La revue de littérature fait en effet apparaitre le propriétaire-dirigeant, qualifié d’entrepreneur, comme un facteur central dans l’explication de la performance de son entreprise. Nous proposons ici une approche cognitive de la performance, approche que nous illustrons au travers des représentations de propriétaires-dirigeants de PME de leur performance. Ainsi, nous nous intéressons aux contenus des représentations véhiculées par les propriétaires-dirigeants de PME, la conception de la performance dans leur vision stratégique. Cette représentation se compose de structures émergentes des concepts reliés entre eux à l’intérieur de ces représentations. En ce sens, nous partons du principe qu’un concept ne prend vraiment sa signification que dans ses relations avec d’autres. La thèse présente son projet en termes de problématique et de méthodologie, au travers d’un processus de collecte de données, par entretiens en profondeur, sur 9 PME performantes primées à plusieurs reprises. A cet effet, nous avons utilisé la méthodologie de la cartographie cognitive qui permet d’analyser la vision stratégique comme un réseau de concepts dans une dynamique des relations les liants. Le traitement et l’analyse des données récoltées nous conduisent à proposer un modèle de lecture cognitive de la performance adapté au contexte des propriétaires-dirigeants de PME. La performance apparait comme un concept multidimensionnel, hiérarchique et dialogique. En effet dans sa PME le propriétaire-dirigeant doit faire des arbitrages entre les différents objectifs et les intérêts multiples des différents acteurs. Enfin, l’étude démontre une importance avérée aux ressources humaines et à l’ancrage territorial de son entreprise. / This thesis aims to contribute to the understanding of the general issue of performance, this by exploring the contents of the strategic vision applied to the context of SME owner-managers. Our thesis emerges from field observations and its evolution, the difficulties of some of SMEs to survive and conversely the contributing factors of success and the performance of some of them. Thus, this research revolves around an ongoing go-return process between literature and results of our field research. Concretely, we find in the SMEs one of the more conducive soils to research on the concept of strategic vision and its relationship to performance. Indeed, the owner-manager, Often described as an entrepreneur, holds in his SME a central predominant position and is one of the major internal factors to consider. Here we propose a cognitive approach to performance, we illustrate that approach through representations of owner-managers of SMEs performance. Thus, we focus on the content of the representations conveyed by the owner-managers of SMEs, the performance conception in their strategic vision, this representation consists of a emerging structures of concepts interconnected within these representations. In this sense, we assume that a concept really takes its meaning only in its relations with others. The thesis presents its problematic and methodological project, through in-depth interviews of nine performing SMEs (award winning repeatedly). To this end, we used the methodology of cognitive mapping that can analyze the strategic vision as a network of concepts in a dynamic relationship. The processing and analysis of collected data lead us to propose a model of cognitive reading of performanceadapted to the context of SME owner-managers. The performance appears as a multidimensional concept, hierarchical, and dialogical. Indeed, in its SME, the owner-manager must make trade-offs between different objectives and different interests of different stakeholders. Finally, the study demonstrates proven importance to human resources and territorial anchorage of its business.

As the globalisation process accelerates there is a growing need for individual countries to understand the bases for effective performance in international trade. Because it makes up such a large share of world trade, it is especially important to understand what determines effectiveness in exporting. Despite much empirical research, especially over recent decades, the state of knowledge on this topic remains fragmented, unclear and unsatisfactory. The motivation for the present study was therefore twofold: dissatisfaction with the present state of knowledge in this vital area and the importance to the UK economy of improving its export performance in a world of increasing competition. Its aim was to contribute to the resolution of both. In addition to finding what appeared to be quite serious methodological problems in a group of earlier studies, our review of the literature indicated that the best prospects for identifying the determinants of effective exporting were to be found, not at national or sectoral level but at that of the individual firm. Accordingly, an empirical survey research project was developed. To minimise unquantifiable inter-sectoral variability, it was focused on a single sector of industry. For a range of reasons, including the limited amount of information available about its current export activity and prospects, the UK knitted apparel industry was chosen.. Special care having been taken to assemble the fullest possible sampling frame and to develop a suitable instrument (which included an export performance model), a mail survey in the form of a stratified random sample of exporting UK manufacturers of knitted apparel was carried through from late 2000. Persistent follow-up by mail and telephone generated a response rate of 70 per cent, comprising close to half of the sampling frame, that was representative of all company size bands, levels of exporting and products. The overall quality of the responses was good; tests of non-response did not find any indications of non-response bias. Data analysis, designed to test thoroughly our 10 export-determinants hypotheses, relied primarily on Pearsonian correlation at the bivariate level then sequentially on Multiple Regression Analysis, Canonical Correlation Analysis and Partial Least Squares. A perhaps slightly novel aspect of the research was that it was not solely cross-sectional in format; a longitudinal element was provided by drawing on the researcher¿s earlier surveys ; and a panel element by following-up, in 2007, the main 2000 field survey. Where possible, these data were drawn upon in the analysis and interpretation. There did not appear to be any conflict between the three multivariate techniques employed and indeed their findings were not dissimilar. The outcome of the data analysis was to uphold, to varying degrees, most of our hypotheses about the determinants of effective/ or successful exporting. Those that did not find support were three: firm size, product adaptation, and price determination method. Most strongly supported as determinants were promotional intensity, serving many markets and visits to trade fairs/ exhibitions; others which were statistically significant, included management commitment, special staff skills and the use of Commission Agents. While the conclusions must remain a bit tentative they are encouraging.

No description available.

Processor architectures are becoming increasingly complex and hence architects have to evaluate a large design space consisting of several parameters, each with a number of potential settings. In order to assist in guiding design decisions we develop simple and accurate models of the superscalar processor design space using a detailed and validated superscalar processor simulator. Firstly, we obtain precise estimates of all significant micro-architectural parameters and their interactions by building linear regression models using simulation based experiments. We obtain good approximate models at low simulation costs using an iterative process in which Akaike’s Information Criteria is used to extract a good linear model from a small set of simulations, and limited further simulation is guided by the model using D-optimal experimental designs. The iterative process is repeated until desired error bounds are achieved. We use this procedure for model construction and show that it provides a cost effective scheme to experiment with all relevant parameters. We also obtain accurate predictors of the processors performance response across the entire design-space, by constructing radial basis function networks from sampled simulation experiments. We construct these models, by simulating at limited design points selected by latin hypercube sampling, and then deriving the radial neural networks from the results. We show that these predictors provide accurate approximations to the simulator’s performance response, and hence provide a cheap alternative to simulation while searching for optimal processor design points.

Deep learning, a sub-topic of machine learning inspired by biology, have achieved wide attention in the industry and research community recently. State-of-the-art applications in the area of computer vision and speech recognition (among others) are built using deep learning algorithms. In contrast to traditional algorithms, where the developer fully instructs the application what to do, deep learning algorithms instead learn from experience when performing a task. However, for the algorithm to learn require training, which is a high computational challenge. High Performance Computing can help ease the burden through parallelization, thereby reducing the training time; this is essential to fully utilize the algorithms in practice. Numerous work targeting GPUs have investigated ways to speed up the training, less attention have been paid to the Intel Xeon Phi coprocessor. In this thesis we present a parallelized implementation of a Convolutional Neural Network (CNN), a deep learning architecture, and our proposed parallelization scheme, CHAOS. Additionally a theoretical analysis and a performance model discuss the algorithm in detail and allow for predictions if even more threads are available in the future. The algorithm is evaluated on an Intel Xeon Phi 7120p, Xeon E5-2695v2 2.4 GHz and Core i5 661 3.33 GHz using various architectures and thread counts on the MNIST dataset. Findings show a 103.5x, 99.9x, 100.4x speed up for the large, medium, and small architecture respectively for 244 threads compared to 1 thread on the coprocessor. Moreover, a 10.9x - 14.1x (large to small) speed up compared to the sequential version running on Xeon E5. We managed to decrease training time from 7 days on the Core i5 and 31 hours on the Xeon E5, to 3 hours on the Intel Xeon Phi when training our large network for 15 epochs

The master thesis is focused on performance evaluation of company Zemědělské družstvo Nové Město na Moravě. The first part contains theoretical foundations dealing with explanation basic theoretical terms and modern methods for performance measurement. The second part is practical, where the analysis company is introduced and after that the performance evaluation using of START Model is performed. The main output of master thesis is the last part, namely suggestions, where are presented measures and recommendations directed at the improvement performance of company.