Bancos de dados oriundos de processos industriais são caracterizados por elevado número de variáveis correlacionadas, dados ruidosos e maior número de variáveis do que observações, tornando a seleção de variáveis um importante problema a ser analisado no monitoramento de tais processos. A presente dissertação propõe sistemáticas para seleção de variáveis com vistas à classificação de bateladas produtivas. Para tanto, sugerem-se novos métodos que utilizam Índices de Importância de Variáveis para eliminação sistemática de variáveis combinadas a ferramentas de classificação; objetiva-se selecionar as variáveis de processo com maior habilidade discriminante para categorizar as bateladas em classes. Os métodos possuem uma sistematização básica que consiste em: i) separar os dados históricos em porções de treino e teste; ii) na porção de treino, gerar um Índice de Importância de Variáveis (IIV) que ordenará as variáveis de acordo com sua capacidade discriminante; iii) a cada iteração, classificam-se as amostras da porção de treino e removem-se sistematicamente as variáveis; iv) avaliam-se então os subconjuntos através da distância Euclidiana dos resultados dos subconjuntos a um ponto hipotético ótimo, definindo assim o subconjunto de variáveis a serem selecionadas. Para o cumprimento das etapas acima, são testadas diferentes ferramentas de classificação e IIV. A aplicação dos métodos em bancos reais e simulados verifica a robustez das proposições em dados com distintos níveis de correlação e ruído. / Databases derived from industrial processes are characterized by a large number of correlated, noisy variables and more variables than observations, making of variable selection an important issue regarding process monitoring. This thesis proposes methods for variable selection aimed at classifying production batches. For that matter, we propose new methods that use Variable Importance Indices for variable elimination combined with classification tools; the objective is to select the process variables with the highest discriminating ability to categorize batch classes. The methods rely on a basic framework: i) split historical data into training and testing sets; ii) in the training set, generate a Variable Importance Index (VII) that will rank the variables according to their discriminating ability; iii) at each iteration, classify samples from the training set and remove the variable with the smallest VII; iv) candidate subsets are then evaluated through the Euclidean distance to a hypothetical optimum, selecting the recommended subset of variables. The aforementioned steps are tested using different classification tools and VII’s. The application of the proposed methods to real and simulated data corroborates the robustness of the propositions on data with different levels of correlation and noise.

The development of fuel efficient internal combustion engines (ICE)have resulted in a variety of different solutions. One of those are the variable valve timing and an implemenation of such is the Continuous Variable Cam Phasing (CVCP). This thesis have used a simulation package, psPack, for the simulation of the gas exchange process for an ICE with CVCP. The purpose of the simulations was to investigate what kind of design parameters, e.g. the length of an intake pipe or the duration of combustion, that were significant for the gas exchange process with the alternation of intake pressure, engine speed and valve setting. The parameters that showed a vast impact were those who affected the amount of residual gas and the temperature of the air charge. Furthermore a validation was made between simulation data acquired from psPack and measured data provided in Heywood (1988). The validation showed that for the general behaviour the simulation results from psPack corresponded well to the measured data.

Ferrofluids are stable colloidal homogeneous mixtures of nano-size single-domain ferromag¬netic particles covered by surfactant layer, and suspended in a carrier fluid compatible with the surfactant. Physical properties of ferrofluid allows one to control it externally using magnetic field without being in direct contact with it. The thesis presents a novel mechanism to pump ferrofluid based on the principle of variable reluctance, in an external magnetic field. The static and dynamic pressure behavior of ferrofluid in a switched DC magnetic field gives an insight into the variable reluctance behavior of ferrofluid. A ferrofluid pump based on the prin¬ciple of variable reluctance of ferrofluid subjected to an external magnetic field, is developed and hence the pump is named as Variable Reluctance Ferrofluid Pump (VRFP). Three configurations of VRFP are developed: • Single stage one-phase VRFP • Single stage two phase VRFP • Multi-stage VRFP A one-phase VRFP consisting of a check valve along with an electromagnet is designed. The valve is modeled and its transfer function is estimated using System Identification method. This model is then used in the simulation model of the pump. The pump is modeled based on the hydraulic-electric analogies. An electric circuit which represents the gross level equivalent of the pump is simulated and the results are compared with that of the experimental measurements. A two phase VRFP is implemented with two electromagnets on either side of the valve around the tube. Two types of magnetic actuation methods are introduced based on the switching sequences of the two electromagnets, namely Full Step Sequencing and Half Step Sequencing. Simulations and experiments were conducted for different pumping conditions. The one phase and two phase VRFPs are single stage structures. A multi-stage VRFP concept, in which the ferrofluid flow channel (tube) is looped through the electromagnets multiple times, is introduced. For the implementation purpose, a two-stage VRFP is discussed in this thesis. Simulations and experiments resulted in significant improvement in case of two-stage VRFP in the pump performance compared to that of single-stage VRFPs. The work presents a simple and novel design of a ferrofluid pump, which is capable of higher flow rates and pumping against higher back pressure compared to the ferrofluid pumps reported in literature. Keywords: Ferrofluid Pump, VRFP, Variable Reluctance, Multi Stage, Magnetic Fluid,

Les systèmes méandriformes construisent, à long terme, des architectures sédimentaires composées de corps poreux disséminés dans un volume de sédiments peu perméables. Ces corps sont susceptibles de contenir des ressources naturelles. Afin d'optimiser leur exploitation, il importe d'estimer la répartition et la connectivité de ces corps. À cet effet, Mines-Paristech développe Flumy, un logiciel de modélisation par processus de ces architectures . Cette thèse vise à améliorer la simulation de la migration dans Flumy en introduisant une influence.Pour cela, trois différents modèles ont été reliés entre eux et comparés. Le premier (modèle à pente constante) est celui actuellement en place dans Flumy, héritier des travaux de [Ikeda 1981] et dans lequel la pente de la surface du cours d'eau est constante. Le deuxième (modèle à pente variable) fut proposé par [Lopez 2003]. Il attribue à la surface libre la pente locale de la topographie. Enfin, un dernier modèle (modèle Saint Venant) a été élaboré en modifiant le précédent. L'écoulement moyen, initialement calculé sous une surface libre connue, est désormais obtenu par résolution des équations de Saint Venant sur un fond connu. Ces trois modèles ont été appliqués à la migration à long terme de méandres libres. Les modèles à pente constante et à pente variable ont également été utilisés pour reproduire les méandres confinés de deux rivières canadiennes, la Beaver et la Red Deer.Les résultats montrent un développement plus réaliste des méandres simulés avec le modèle à pente variable que de ceux simulés avec le modèle à pente constante. Cette amélioration est observable dans les méandres isolés (ralentissement de la vitesse d'extension) et dans les trains de méandres (confinement en ceintures de méandres). Le modèle à pente variable permet également une meilleure reproduction de la morphologie des méandres confinés. L'utilisation du modèle Saint Venant apporte les mêmes améliorations dans la simulation des méandres libres, quoique moins prononcées. Il permet surtout de construire une surface libre physique sur une grande variété de fonds, résolvant ainsi une limite identifiée dans le modèle à pente variable. / Over a long time, Meandering systems build sedimentary architectures composed of porous bodies scattered inside a volume of low-permeability sediments. These bodies may contain natural resources. In order to optimize their mining, it is essential to estimate the distribution and connectivity of such bodies. To this end, Mines ParisTech develops Flumy, a process-based model simulating the formation of these architectures. This thesis aims to improve the simulation of the migration in Flumy by taking into consideration the influence of the local slope.For this purpose, three distinct models were considered in conjunction, and compared. The first one (constant slope model), which constitutes the basis of the current Flumy version, was originally developed by [Ikeda 1981]. The second model (variable slope model), developed by [Lopez 2003], assigns to the free surface the slope of the surrounding topography. Finally, the last model (Saint-Venant model) has been derived from the variable slope model. Initially calculated under a known free surface, the mean flow in each cross-section is now obtained by solving the Saint-Venant equations over a known river bed. Each of those three models has been applied to the simulation of free meanders. Moreover, the constant-slope and variable slope models have been used to reproduce the confined meanders of two Canadian streams.The results point to a more realistic meanders development using the variable slope model than with the constant slope model. This improvement can particularly be observed in individual meanders, whose rate of extension decreases with the age. It is also noticeable in the overall river behavior, which self-confines in a meander belt. The specific morphology of the confined meanders is also better reproduced using the variable slope model than with the constant slope model. Lastly, though a lesser extent, the Saint-Venant model shows the same advantages than the variable slope model. In addition, it allows the construction of a physically meaningful free surface over a wide range of beds and, in doing so, resolves a limit of the variable slope model.

Bancos de dados oriundos de processos industriais são caracterizados por elevado número de variáveis correlacionadas, dados ruidosos e maior número de variáveis do que observações, tornando a seleção de variáveis um importante problema a ser analisado no monitoramento de tais processos. A presente dissertação propõe sistemáticas para seleção de variáveis com vistas à classificação de bateladas produtivas. Para tanto, sugerem-se novos métodos que utilizam Índices de Importância de Variáveis para eliminação sistemática de variáveis combinadas a ferramentas de classificação; objetiva-se selecionar as variáveis de processo com maior habilidade discriminante para categorizar as bateladas em classes. Os métodos possuem uma sistematização básica que consiste em: i) separar os dados históricos em porções de treino e teste; ii) na porção de treino, gerar um Índice de Importância de Variáveis (IIV) que ordenará as variáveis de acordo com sua capacidade discriminante; iii) a cada iteração, classificam-se as amostras da porção de treino e removem-se sistematicamente as variáveis; iv) avaliam-se então os subconjuntos através da distância Euclidiana dos resultados dos subconjuntos a um ponto hipotético ótimo, definindo assim o subconjunto de variáveis a serem selecionadas. Para o cumprimento das etapas acima, são testadas diferentes ferramentas de classificação e IIV. A aplicação dos métodos em bancos reais e simulados verifica a robustez das proposições em dados com distintos níveis de correlação e ruído. / Databases derived from industrial processes are characterized by a large number of correlated, noisy variables and more variables than observations, making of variable selection an important issue regarding process monitoring. This thesis proposes methods for variable selection aimed at classifying production batches. For that matter, we propose new methods that use Variable Importance Indices for variable elimination combined with classification tools; the objective is to select the process variables with the highest discriminating ability to categorize batch classes. The methods rely on a basic framework: i) split historical data into training and testing sets; ii) in the training set, generate a Variable Importance Index (VII) that will rank the variables according to their discriminating ability; iii) at each iteration, classify samples from the training set and remove the variable with the smallest VII; iv) candidate subsets are then evaluated through the Euclidean distance to a hypothetical optimum, selecting the recommended subset of variables. The aforementioned steps are tested using different classification tools and VII’s. The application of the proposed methods to real and simulated data corroborates the robustness of the propositions on data with different levels of correlation and noise.

Une étude expérimentale du développement de jets turbulents à masse volumique variable est menée. Les comportements axiaux et radiaux des propriétés statistiques moyennes et fluctuantes obtenues sur les champs de vitesse et de concentration sont analysés. Ces résultats apportent une meilleure compréhension des phénomènes de mélange et de l’effet de la variation de densité dans le but d’apporter des données complémentaires pour le développement des calculs numériques et accroître leurs précisions. Les cas étudiés dans cette thèse sont deux jets d’hélium marqués par de la vapeur d’acétone pour des nombres de Reynolds Re = 7000 et 11000 permettant d’explorer l’effet de fortes variations de masse volumique (S = 0,39 et 0,41, respectivement) par rapport à l’air ambiant. Le cas d’un jet d’air (Re = 16000) également porteur de vapeur d’acétone est utilisé comme cas de référence comparable au cas d’un contaminant passif avec S = 1,17. L’étude porte sur la région de proche sortie de jet, jusqu’à une distance de 40 fois le diamètre Dj de sortie du jet. La configuration adoptée est celle d’un jet de tube rond, libre, axisymétrique, vertical et ascendant dont le diamètre intérieur du tube est Dj = 3,5mm se développant dans l’air ambiant. La particularité de cette étude est la mise en œuvre d’un système de mesure par diagnostic optique qui permet un couplage spatial et temporel des mesures de vitesse et de concentration sur une région de l’écoulement de quelques cm2 et non sur un seul point. Pour ce faire, le couplage des mesures PIV, pour la mesure du champ de vitesse, et PLIF basée sur la fluorescence de la vapeur d’acétone, pour la mesure du champ de concentration, a été étudié, conçu et testé. Il nous a permis d’obtenir une base de données, encore trop rares aujourd’hui, sur l’évolution des flux turbulents croisés de vitesse et concentration. Ces données ont mis en évidence une évolution plus rapide du jet le plus léger. Cependant, les coefficients de corrélation semblent identiques lorsque l’on atteint la zone autosimilaire du jet. Une approche basée sur les probabilités de densité conjointes vitesse-concentration a permis de mettre en évidence des différences dans la région extérieure des jets, où l’intermittence de frontière donne son empreinte sur les propriétés du mélange. / An experimental study of the development of turbulent jets with variable density is presented. The axial and radial behaviours of the mean and fluctuating statistical properties obtained on the velocity and concentration fields are analysed. These results provide a better understanding of mixing phenomena and the effect of density variation in order to provide complementary data for the development of numerical calculations and to increase their precision. The cases studied in this thesis are two helium jets carrying acetone vapor for Reynolds numbers Re = 7000 and 11000 to explore the effect of large density variations (S = 0.39 and 0.41, respectively) relative to the ambient air. The case of an air jet (Re = 16000) also carrying acetone vapor is used as a reference case comparable to the case of a passive contaminant with S = 1.17. The study focuses on the region of near jet exit, up to a distance of 40 times the jet outlet diameter Dj. The adopted configuration is a round jet tube, free, axisymmetric, vertical and ascending whose internal diameter of the tube is Dj = 3.5 mm developing in the ambient air. The particularity of this study is the implementation of an optical diagnostic measurement system that allows a spatial and temporal coupling of speed and concentration measurements over a region of the flow of a few cm2. To do this, the coupling of the PIV measurements for the measurement of the velocity field and the PLIF based on the fluorescence of the acetone vapor for the measurement of the concentration field has been studied, designed and tested. It allowed us to obtain a database, still too rare today, on the evolution of the turbulent flow of speed and concentration. These data have shown a faster evolution of the lightest jet. However, the correlation coefficients appear to be identical when the self-similar zone of the jet is reached. An approach based on joint speed-concentration density probabilities has revealed differences in the outer region of the jets, where the intermittent boundary gives its imprint on the properties of the mixture.

This paper describes a novel concept for a low cost variable delivery external gear pump (VD-EGP). The proposed VD-EGP is based on the realization of a variable timing for the connections of the internal displacement chambers with the inlet and outlet ports. With respect to a standard EGP, an additional element (slider) is used along with asymmetric gears to realize the variable timing principle. Previously performed tests confirmed the validity of the concept, for a design capable of varing the flow in the 65%-100% range. Although the VD-EGP concept is suitable for various flow control system typologies (manual, electro-actuated, hydraulically flow- or pressure- compensated), this paper particularly details the design and the test results for a prototype that includes both a manual flow control system and a pressure compensator. Flow vs pressure and volumetric efficiency curves are discussed along with transient (outlet flow fluctuation) features of the VD-EGP.

Using an inverter-driven compressor in variable capacity heat pump systems has a main drawback, which is the extra loss in the inverter. The present experimental study aims to recover the inverter losses by using brine-cooled and water-cooled inverters, thereby improving the total efficiency of the heat pump system. In order to achieve this goal, a test rig with the air-cooled, water-cooled and brine-cooled inverters is designed and built, and a comparative analysis of the recovered heat, inverter losses and system performance is conducted when the compressor is driven by the water-cooled, brine-cooled and air-cooled inverters at three different switching frequencies for each inverter. The experimental results show that the inverter losses as a magnitude and as a ratio of the total consumed power are lowest in the brine-cooled inverter and highest in the air-cooled one at all the compressor speeds and all the inverter switching frequencies. Moreover, the recovered energy varies between 45 and 125 (W) in the water-cooled inverter, which corresponds to 63 and 69 (%) of the inverter losses; while it varies between 61 and 139 (W) in the brine-cooled inverter, which corresponds to 79 and 90 (%) of the inverter losses. It is also proved that the improvement of the system coefficient of performance (COPsys) is almost the same when the water-cooled or the brine-cooled inverter is used and varies between 0.54 and 3 (%) in comparison with using the air-cooled one. Indeed, the total isentropic efficiency of the compressor is improved slightly when using the water-cooled inverter and little more when using the brine-cooled one at the same running conditions. In addition, the total isentropic efficiency of the compressor is improved by increasing the inverter switching frequency when any of the inverters is used. The experimental results also show that cooling the inverter by the water, which comes out from the condenser, increases the maximum temperature of the base plate of the inverter about 10 °C which could cause a two-fold deterioration in the inverter median life in comparison with cooling the inverter by air. On the contrary, using the brine for cooling the inverter decreases the maximum temperature of the base plate of the inverter about 30 °C which could cause about a six-fold improvement in the inverter median life. / Capacity-controlled Ground Source Heat Pump single-family dwellings

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第18311号 / 工博第3903号 / 新制||工||1599(附属図書館) / 31169 / 京都大学大学院工学研究科化学工学専攻 / (主査)教授 長谷部 伸治, 教授 大嶋 正裕, 教授 宮原 稔 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

No description available.