The doctoral thesis deals with the analysis of the flow in the circular and not circular cross-section pipes by methods using the distribution of the vorticity density. This analysis is particularly focused on the derivation of the new velocity profiles formulas using the above mentioned method. In this work is presented a historical overview of the derived velocity profiles. This overview of already derived velocity profiles will be a fundamental benchmark for newly derived velocity profiles. These new velocity profiles are derived for the circular and not circular cross-section pipes and the derivation is based on the analogy of electromagnetic induction by using Biot-Savart law. It is necessary to apply this analogy at first on solitary vortex filament. By taking this step is possible to get the value of the induced velocity from one solitary vortex filament. Subsequently it is possible to obtain the value of the induced velocity from the vorticity wall and afterwards from the vorticity density distribution over the cross section. This work contains also the results of the experimental measurements of the velocity profiles, and of the CFD simulations. Experimentally measured results are used besides other for the selecting of the most suitable CFD computational model. Selected CFD model will be subsequently declared as a reference model and the valid velocity profiles for this model will serve with the experimentally measured data as a benchmark for the newly derived velocity profiles.

La perméabilité aux gaz et aux liquides des matériaux polymères est une propriété qui est mise à profit dans de nombreux domaines industriels. Cette thèse est effectuée dans l'optique de mieux appréhender la problématique de l'étanchéité des conduites flexibles par les polymères. Ainsi, les perméabilités de H2S, CO2 et CH4 dans le polyéthylène (PE) ont fait l'objet d'une étude effectuée dans le contexte de cette thèse. La perméabilité est une propriété qui résulte de la solubilisation des gaz dans le polymère puis de la diffusion de ces produits à travers la matière. La solubilité, qui caractérise l’aptitude d’un gaz à pouvoir s’absorber dans le polymère, est une propriété d’équilibre, qui pourra être étudiée par les techniques de Monte Carlo. La diffusion, qui caractérise l’aptitude d’un gaz à se mouvoir plus ou moins rapidement dans le réseau polymère, sera quant à elle, étudiée par dynamique moléculaire. Au dessous de la température de fusion, le polyéthylène est à l'état semi-cristallin. Cet état est composé de régions contenant des chaînes orientées aléatoirement (régions amorphes) et des régions contenant des chaînes orientées sur un réseau (régions cristallines). La morphologie complexe des polymères semi-cristallins présente des hétérogénéités de dimensions nanométriques, ce qui est difficilement accessible par la simulation moléculaire. A fin d'étudier la solubilité et la diffusion de gaz dans le polyéthylène semi-cristallin, nous modéliserons uniquement la phase amorphe au cours de ce travail. Par contre, l’effet des régions cristallines sur la phase amorphe sera pris en compte dans la simulation par une contrainte ad-hoc. / The gas permeability through the polymers is a property that is exploited in many industrial fields. The objective of this thesis is to better understand the problem of sealing of flexible pipes with polymers. Thus, the permeability of H2S, CO2 and CH4 in polyethylene (PE) was studied during this work. Permeability is a property resulting from the dissolution of gases in the polymer and then diffusion of these products through the material. Solubility, which characterizes the ability of a gas to be absorbed in the polymer, is a property of equilibrium, which can be studied by Monte Carlo techniques. Diffusion coefficient, which characterizes the ability of a gas to move more or less rapidly into the polymer network, will in turn studied by molecular dynamics.Below the melting temperature, polyethylene is in semi-crystalline state. This state is composed of regions containing randomly oriented chains (amorphous regions) and regions containing chains oriented regularly on a network (crystalline regions). The complex morphology of semi-crystalline polymers has nanometric heterogeneities, which is not easily accessible by molecular simulation. In order to study the solubility and diffusion coefficient of gases in semi-crystalline polyethylene, we model only the amorphous phase in this work. However, the effect of crystalline regions on the amorphous phase will be taken into account in the simulation by an ad-hoc constraint.

[pt] Apresenta-se uma metodologia de análise numérica para dutos enterrados usados no transporte de petróleo e gás, considerando não-linearidades geométricas e não linearidades de material baseada na formulação Lagrangeana Total. Emprega-se uma modelagem com base em uma discretização com elementos especiais de viga. As equações de equilíbrio são formuladas a partir do principio dos trabalhos virtuais, segundo as componentes de tensão e deformação no elemento viga-duto, com emprego da técnica do Módulo Reduzido de Integração Direta (RMDI), na qual incorpora-se o comportamento plástico do material. Esta técnica exclui da análise os efeitos da flambagem local nas paredes do duto. As matrizes para resolução por elementos finitos destas equações são derivadas. Incorporam- se, nesta metodologia os efeitos de pressão interna constante no duto assim como a interação solo-duto através da modelagem do solo por meio de molas elasto-plásticas verticais e horizontais. Na modelagem desenvolvida as cargas distribuídas são consideradas como constantes no sistema global de eixos. Desenvolve-se um programa para computador segundo a metodologia apresentada. São estudados alguns exemplos com o objetivo de avaliar numericamente os resultados de análise obtidos e formular algumas conclusões sobre o comportamento de dutos enterrados. / [en] This work presents a numerical methodology for the analysis of buried pipes employed by the transport of oil and gas. Both geometric and material nonlinearities are considered in a total Lagrangean formulation. A special type of beam element is employed in the finite element discretization. The equilibrium equations are formulated based on the virtual work principle considering the stress and deformation components of the beam-pipe element. The Reduced Modulus Direct Integration (RMDI) technique is employed through which the plastic material behavior is incorporated. This technique excludes from the analysis the local buckling effects of the pipe walls. The corresponding finite element matrices for this element are obtained. In this methodology the effects of the constant internal pressure as well as the soil-pipe interaction by means of elastic-plastic uplift, bearing and longitudinal springs are included. Distributed loads are considered constant with respect to the global axis. A computer code is developed according to the methodology presented. Some examples are studied with the objective to evaluate numerically the analysis results and to formulate some conclusions to the behavior of buried pipes.

An important challenge in natural language surface realization is the generation of grammatical sentences from incomplete sentence plans. Realization can be broken into a two-stage process consisting of an over-generating rule-based module followed by a ranker that outputs the most probable candidate sentence based on a statistical language model. Thus far, an n-gram language model has been evaluated in this context. More sophisticated syntactic knowledge is expected to improve such a ranker. In this thesis, a new language model based on featurized functional dependency syntax was developed and evaluated. Generation accuracies and cross-entropy for the new language model did not beat the comparison bigram language model.

No description available.

Ce manuscrit synthétise plus d'une décade de notre recherche académique sur le sujet d'optimisation de codes bas niveau, dont le but est une intégration dans un compilateur optimisant ou dans un outil d'optimisation semi-automatique. Dans les programmes bas niveau, les caractéristiques du processeur sont connues et peuvent être utilisées pour générer des codes plus en harmonie avec le matériel. Nous commençons notre document par une vue générale sur le problème d'ordonnancement des phases de compilation. Actuellement, des centaines d'étapes de compilation et d'optimisation de codes existent; un problème fondamental et ouvert reste de savoir comment les combiner et les ordonner efficacement. Pour pallier rapidement cette difficulté, une stratégie du moindre effort consiste à appliquer une compilation itérative en exécutant successivement le programme avant de décider de la technique d'optimisation de code à employer et avec quels paramètres. Nous prouvons que l'approche de compilation itérative ne simpli fie pas fondamentalement le problème, et l'utilisation de modèles statiques de performances reste un choix raisonnable. Un problème classique de con it entre deux étapes de compilation est celui qui lie l'allocation de registres et l'ordonnancement d'instructions. Nous montrons comment gérer efficacement cet antagonisme en séparant les contraintes de registres des contraintes d'ordonnancement d'instructions. Cela est possible grâce à la notion de saturation en registres (RS), qui est le besoin maximal en registres pour tous les ordonnancements possibles d'un graphe. Nous apportons une contribution formelle et une heuristique efficace, qui permettent la détection de contraintes de registres toujours véri fiées; ils peuvent par conséquent être négligées. Nous introduisons la plate-forme SIRA, qui permet de garantir l'absence de code de vidage avant l'ordonnancement d'instructions. SIRA est un modèle basé sur la théorie des graphes permettant de borner le besoin maximal en registres pour tout pipeline logiciel, sans altérer, si possible, le parallélisme d'instructions. SIRA modélise les contraintes cycliques des registres dans différentes architectures possibles : avec plusieurs types de registres, avec tampons ou les d'attente, et avec des bancs de registres rotatifs. Nous apportons une heuristique efficace qui montre des résultats satisfaisants, que ce soit comme outil indépendant, ou comme passe intégrée dans un vrai compilateur. Dans le contexte des processeurs exhibant des retards d'accès aux registres (VLIW, EPIC, DSP), nous attirons l'attention sur le problème qui peut survenir lorsque les contraintes de registres sont traitées avant l'ordonnancement d'instructions. Ce problème est la création de circuits négatifs ou nuls dans le graphe de dépendances de données. Nous montrons comment éliminer ces circuits indésirables dans le contexte de SIRA. SIRA définit une relation formelle entre le nombre de registres alloués, le parallélisme d'instructions et le facteur de déroulage d'une boucle. Nous nous basons sur cette relation pour écrire un algorithme optimal qui minimise le facteur de déroulage tout en sauvegardant le parallélisme d'instructions et en garantissant l'absence de code de vidage. D'après nos connaissances, ceci est le premier résultat qui démontre que le compactage de la taille de code n'est pas un objectif antagoniste à l'optimisation des performances de code. L'interaction entre la hiérarchie mémoire et le parallélisme d'instructions est un point central si l'on souhaite réduire le coût des latences d'opérations de chargement. Premièrement, notre étude pratique avec des micro-benchmarks montre que les processeurs superscalaires ayant une exécution dans le désordre ont un bug de performances dans leur mécanisme de désambiguation mémoire. Nous montrons ensuite qu'une vectorisation des opérations mémoire résoud ce problème pour des codes réguliers. Deuxièmement, nous étudions l'optimisation de préchargement de données pour des codes VLIW embarqués irréguliers. Finalement, avec l'arrivée des processeurs multicoeurs, nous observons que les temps d'exécution des programmes deviennent très variables. A fin d'améliorer la reproductibilité des résultats expérimentaux, nous avons conçu le Speedup-Test, un protocole statistique rigoureux. Nous nous basons sur des tests statistiques connus (tests de Shapiro-Wilk, F de Fisher, de Student, de Kolmogorov-Smirnov, de Wilcoxon- Mann-Whitney) a n d'évaluer si une accélération observée du temps d'exécution médian ou moyen est signi cative.

Thesis (LL.M. (Import and Export Law))--North-West University, Potchefstroom Campus, 2008.

Estimation of depth within an imaged scene can be formulated as a stereo correspondence problem. Software solutions tend to be too slow for high frame rate (i.e. > 30 fps) performance. Hardware solutions can result in marked improvements. This thesis explores one such hardware implementation that generates dense binocular disparity estimates at frame rates of over 200 fps using a dynamic programming formulation (DPML) developed by Cox et. al. A highly parameterizable field programmable gate array implementation of this architecture demonstrates equivalent accuracy while executing at significantly higher frame rates to those of current approaches. Existing hardware implementations for dense disparity estimation often use sum of squared difference, sum of absolute difference or other similar algorithms that typically perform poorly in comparison to DPML. The presented system runs at 248 fps for a resolution of 320 x 240 pixels and disparity range of 128 pixels, a performance of 2.477 billion DPS.

lthough extensive research work has been carried out on the drag reduction behavior of polymers and surfactants alone, little progress has been made on the synergistic effects of combined polymers and surfactants. A number of studies have demonstrated that certain types of polymers and surfactants interact with each other to form surfactant-polymer complexes. The formation of such complexes can cause changes in the solution properties and may result in better drag reduction characteristics as compared with pure additives. A series of drag-reducing surfactants and polymers were screened for the synergistic studies. The following two widely used polymeric drag reducing agents (DRA) were chosen: a copolymer of acrylamide and sodium acrylate (referred to as PAM) and polyethylene oxide (PEO). Among the different types of surfactants screened, a cationic surfactant octadecyltrimethylammonium chloride (OTAC) and an anionic surfactant Sodium dodecyl sulfate (SDS) were selected for the synergistic study. In the case of the cationic surfactant OTAC, sodium salicylate (NaSal) was used as a counterion. No counterion was used with anionic surfactant SDS. The physical properties such as viscosity, surface tension and electrical conductivity were measured in order to detect any interaction between the polymer and the surfactant. The drag reduction (DR) ability of both pure and mixed additives was investigated in a pipeline flow loop. The effects of different parameters such as additive concentration, type of water (deionized (DI) or tap), temperature, tube diameter, and mechanical degradation were investigated. The addition of OTAC to PAM solution has a significant effect on the properties of the system. The critical micelle concentration (CMC) of the mixed surfactant-polymer system is found to be different from that of the surfactant alone. The anionic PAM chains collapse upon the addition of cationic OTAC and a substantial decrease in the viscosity occurs. The pipeline flow behaviour of PAM/OTAC mixtures is found to be consistent with the bench scale results. The drag reduction ability of PAM is reduced upon the addition of OTAC. At low concentrations of PAM, the effect of OTAC on the drag reduction behavior is more pronounced. The drag reduction behavior of polymer solutions is strongly influenced by the nature of water (de-ionized or tap). The addition of OTAC to PEO solution exhibited a week interaction based on the viscosity and surface tension measurements. However, the pipeline results showed a considerable synergistic effect, that is, the mixed system gave a significantly higher drag reduction (lower friction factors) as compared with the pure additives (pure polymer or pure surfactant). The synergistic effect in the mixed system was stronger at low polymer concentrations and high surfactant concentrations. Also the resistance against mechanical degradation of the additive was improved upon the addition of OTAC to PEO. The mixed PEO/SDS system exhibited a strong interaction between the polymers (PEO) and the surfactant (SDS), Using electrical conductivity and surface tension measurements, the critical aggregation concentration (CAC) and the polymer saturation point (PSP) were determined. As the PEO concentration is increased, the CAC decreases and the PSP increase. The addition of SDS to the PEO solution exhibits a remarkable increase in the relative viscosity compared to the pure PEO solution. This increase is attributed to the changes in the hydrodynamic radius of the polymer coil. The pipeline flow exhibited a considerable increase in DR for the mixed system as compared to the pure PEO solution. The addition of surfactant always improves the extent of DR up to the PSP. Also the mixed PEO/ SDS system shows better resistance against shear degradation of the additive.

Estimation of depth within an imaged scene can be formulated as a stereo correspondence problem. Software solutions tend to be too slow for high frame rate (i.e. > 30 fps) performance. Hardware solutions can result in marked improvements. This thesis explores one such hardware implementation that generates dense binocular disparity estimates at frame rates of over 200 fps using a dynamic programming formulation (DPML) developed by Cox et. al. A highly parameterizable field programmable gate array implementation of this architecture demonstrates equivalent accuracy while executing at significantly higher frame rates to those of current approaches. Existing hardware implementations for dense disparity estimation often use sum of squared difference, sum of absolute difference or other similar algorithms that typically perform poorly in comparison to DPML. The presented system runs at 248 fps for a resolution of 320 x 240 pixels and disparity range of 128 pixels, a performance of 2.477 billion DPS.