This dissertation focuses on the application of Monte Carlo (MC) molecular simulation and Molecular Dynamics (MD) in modeling thermodynamics and flow of subsurface reservoir fluids. At first, MC molecular simulation is proposed as a promising method to replace correlations and equations of state in subsurface flow simulators. In order to accelerate MC simulations, a set of early rejection schemes (conservative, hybrid, and non-conservative) in addition to extrapolation methods through reweighting and reconstruction of pre-generated MC Markov chains were developed. Furthermore, an extensive study was conducted to investigate sorption and transport processes of methane, carbon dioxide, water, and their mixtures in the inorganic part of shale using both MC and MD simulations. These simulations covered a wide range of thermodynamic conditions, pore sizes, and fluid compositions shedding light on several interesting findings. For example, the possibility to have more carbon dioxide adsorbed with more preadsorbed water concentrations at relatively large basal spaces. The dissertation is divided into four chapters. The first chapter corresponds to the introductory part where a brief background about molecular simulation and motivations are given. The second chapter is devoted to discuss the theoretical aspects and methodology of the proposed MC speeding up techniques in addition to the corresponding results leading to the successful multi-scale simulation of the compressible single-phase flow scenario. In chapter 3, the results regarding our extensive study on shale gas at laboratory conditions are reported. At the fourth and last chapter, we end the dissertation with few concluding remarks highlighting the key findings and summarizing the future directions.

Un système logistique est une composante essentielle d'un système spatial dans lequel les acteurs s'organisent autour d'infrastructures pour faire circuler des flux (de marchandises, d'information et financier) sur un territoire. L'organisation logistique globale résulte d'un processus auto-organisé et distribué de la part des acteurs. Ce travail vise à comprendre, à de multiples échelles, comment des acteurs autonomes et très hétérogènes (dans leurs modes de fonctionnements et dans leurs objectifs), s'organisent collectivement autour des infrastructures à leurs dispositions pour gérer des flux soumis à un ensemble de contraintes (temporelles, spatiales,...). On propose ici un modèle orienté agent permettant de simuler les processus de création et d'organisation des flux liés à la logistique sur un territoire. Le modèle prévoit de décrire l'interface entre les flux internationaux et les flux urbains afin de comprendre comment les dynamiques portuaires et urbaines cohabitent au sein du système. Le modèle intègre une dynamique structurelle et organisationnelle grâce aux graphes dynamiques afin de représenter l'évolution du système. Le modèle permet ainsi aux agents de s'adapter, comme dans la réalité, à des perturbations du système. / A logistic system is an essential component of a spatial system. Actors are organised around infrastructures in order to move different kinds of flow (of goods, of information, or financial) over a territory. The logistic organisation comes from an auto-organised and distributed process from the actors. This works aims to understand, at different scales, how autonomous and heterogeneous actors (according to their goals and methods to take decisions) are collectively organised around infrastructures to manage different kinds of flow, and despite numerous constraints (temporal, spatial,...). We propose an agent-based model which allows to simulate the processes to create and organise logistic flow over a territory. The model describes an interface between international and urban flow in order to understand how the port and urban dynamics work together. The model integrates a structural and organisational dynamics thanks to dynamic graphs in order to represent the evolution of this kind of system. Thus, the agents can adapt themselves to system's perturbations as in the reality.

Les matériaux à changement de phase (MCP) constituent une alternative prometteuse pour l'amélioration de l'inertie thermique des matériaux de construction. Grâce à leur chaleur latente, ces matériaux permettent de stocker des quantités importantes d'énergie thermique, ce qui permet de réduire la consommation d'énergie liée au chauffage et à la climatisation. Cependant, leur incorporation dans les matériaux cimentaires entraine une baisse de la résistance mécanique du nouveau matériau composite ainsi obtenu. Durant ces dernières décennies, les composites MCP/bétons ont suscité un grand intérêt conduisant à un grand nombre de travaux expérimentaux. Cependant, les modèles théoriques et numériques pour prédire les comportements de ces matériaux complexes sont aujourd'hui très peu développés en raison de la complexité du comportement thermique avec changement de phase, de la séparation d'échelle et de la difficulté que représente la prédiction de l'endommagement par fissuration à l'échelle des hétérogénéités microscopiques. L'objectif de cette thèse est précisément de développer des outils de modélisation numériques pour prédire le comportement thermomécanique effectif du matériau en vue de calculs de structures. Pour cela, des modèles numériques sont développés pour simuler le transfert de chaleur, le comportement mécanique, la fissuration ainsi que la fuite du MCP liquide à travers les fissures, à l'échelle d'un Volume Elémentaire Représentatif du matériau. Après avoir étudié les effets des changements de phase dans le MCP sur le comportement mécanique effectif, une approche multi-échelle (méthode EF²$) est proposée afin de réaliser des calculs de structures en tenant compte des phénomènes à l'échelle des micro capsules. Des caractérisations expérimentales thermo-physiques sont menées afin de prouver l'utilité des MCP dans les matériaux de construction et de faire des comparaisons avec les outils d'homogénéisation développés. Enfin, nous proposons une étude dans le but de comprendre et d'évaluer les effets du MCP dans la dégradation des propriétés mécaniques de ces nouveaux matériaux / A promising way to enhance thermal inertia of buildings is the use of phase change materials (PCMs). Thanks to their high latent heat, PCMs can be used to store a significant amount of thermal energy in order to reduce energy consumption related to air conditioning. However, their use leads to a decrease in the mechanical strength of the obtained composites. During the last decades, the incorporation of PCMs in concrete has been of great interest leading to many experimental works. However, theoretical and numerical models to predict the behavior of such complex materials are not developed so far, due to the complexity of the phase change behavior, the scale separation and issues associated to the damage which is mainly induced by microcracking at the scale of microstructural heterogeneities. The objective of this thesis is precisely to develop numerical modeling tools to predict the effective thermomechanical behavior of the material with aim of structural calculations. For this purpose, numerical tools based on microstructures at the scale of microencapsulated PCM are developed to simulate heat transfer, mechanical response, cracks propagation as well as leakage of liquid PCM through cracks. After studying the effects of phase changes in the PCM on the effective mechanical response of the composites, a multi-scale approach (FE² method) is proposed to carry out structural calculations taking into account phenomena at micro scale. Thermo-physical experimental characterizations are carried out to show the usefulness of PCMs in building materials and to make comparisons with the developed homogenization tools. Finally, we propose a study to understand and evaluate the effects of PCMs in the degradation of the mechanical properties of these new materials

Avec le développement des matériaux fonctionnels (multi-matériaux, multicouches,…), la caractérisation du comportement mécanique par des moyens macroscopiques conventionnels est devenue de plus en plus difficile. Ces méthodes conventionnelles sont donc substituées progressivement par des moyens de caractérisation multi-échelles. Parmi ces moyens, la nanoindentation, qui peut résoudre certains défis de la micro-caractérisation tels que la présence de phases indissociables, les systèmes multicouches, les revêtements ultra-minces, etc. Cet outil est devenu une technique de haute précision capable de solliciter des volumes de matière très faibles et fournir des informations riches pour la caractérisation des matériaux. Cependant, cet outil est utilisé majoritairement pour identifier les propriétés élastiques et qualitativement certains paramètres tels que la dureté, la ductilité et les contraintes internes.Ce travail de thèse s’intéresse à la caractérisation du comportement élastoplastique par nanoindentation à deux échelles : l’échelle macroscopique et l’échelle du cristal.Le premier défi de ce travail est expérimental. Il s’agit de générer des surfaces avec des propriétés représentatives de la microstructure étudiée. Ce défi est d’autant plus relevé que le matériau utilisé comme modèle est l’acier 316L très ductile et dont la surface est sensible au moindre changement. Un protocole expérimentale a été mis en place, à l’issu de ce travail, et les erreurs et dispersions de la réponse en nanoindentation introduites par les différentes étapes de génération de surface ont été quantifiés.Une base de données étendue a été mise en place, par la suite. Différentes géométries d’indent ont été appliquées à plusieurs profondeurs. Cette base de données va alimenter des stratégies d’identification inverse basée sur un couplage entre des algorithmes d’optimisation et une modélisation éléments finis de l’essai. Deux types d’algorithme ont été appliqués : Levenberg-Marquardt et l’algorithme génétique. Ce dernier est très consommateur en temps de calcul. Différents modèles EF axisymétrique et 3D ont été utilisés. Ces modèles ont été soigneusement optimisés par rapport au temps de calcul.Plusieurs stratégies d’identification ont été employées en se basant sur différentes données expérimentales issues de l’essai de nanoindentation telles que la courbe de charge-décharge, la forme de l’empreinte résiduelle et l’association de plusieurs géométries d’indent. Plusieurs modèles d’écrouissage isotrope ont été identifiés. À l’échelle macroscopique, les modèles d’écrouissage isotrope classiques ont été déterminés. À l’échelle du grain, la loi cristalline de Méric et Cailletaud a été identifiée. Les résultats obtenus ont été confrontés, à l’échelle macroscopique, à des identifications réalisées sur le même matériau à partir des essais de traction et de compression et ont montré que l’association de multiples géométries d’indentation permet de reproduire le comportement volumique du 316L avec une précision acceptable. Pour le comportement du cristal, des essais de compression de micropilliers ont été utilisé pour se procurer des données de référence à cette échelle. La comparaison montre beaucoup de dispersion dans les deux cas. En effet, certains phénomènes liés à la densité de dislocation très variables d’un grain à l’autre sont responsables de cette dispersion. Cette densité de dislocation n’est pas prise en compte, en tant que variable, dans le modèle cristallin utilisé. L’utilisation d’un modèle plus physique intégrant la densité de dislocation et son évolution permet d’améliorer ces résultats. Enfin, une nouvelle méthode d’identification a été proposée. Cette méthode est basée sur l’estimation et l’introduction de la géométrie réelle de l’indent dans le modèle EF utilisé pour l’identification. La méthode a été validée dans le cas de la pointe Berkovich et elle montre des résultats très prometteurs. / With the development of functional materials (multi-materials, multilayers, ...), the mechanical behavior characterization by conventional macroscopic methods has become progressively difficult. These conventional methods are therefore gradually substituted by multiscale characterization processes. Among these methods, the nanoindentation, this can solve certain challenges of micro-characterization such as the presence of indissociable phases, multilayer systems, ultra-thin coatings, etc. This tool has become a high-precision technique capable of testing very small volumes of matter and providing rich information for material characterization. However, this tool is used mainly to identify the elastic properties and, qualitatively, some parameters such as hardness, ductility and internal stresses.This thesis work focuses on the characterization of elastoplastic behavior by nanoindentation at two scales: the macroscopic scale and the crystal scale.The first challenge of this work is experimental. It involves generating surfaces with properties representative of the studied microstructure. This challenge is important because the material used as a model is 316L steel which is very ductile and whose surface is sensitive to small perturbations. An experimental protocol was implemented at the end of this work, and the errors and dispersions of the nanoindentation response introduced by the different surface generation steps were quantified. Then, a wide database was implemented with different indenter geometries and several depths. This database will feed inverse identification strategies based on a coupling between optimization algorithms and finite element modeling of this test. Two types of algorithm have been applied: Levenberg-Marquardt and genetic algorithms. The latter is very consumer in computing time. Different axisymmetric and 3D FE models have been used. These models have been carefully optimized with respect to computation time.Several identification strategies were employed based on various experimental databases from the nanoindentation test such as the loading-unloading curve, the residual imprint shape and the association of several indent geometries. Some models of isotropic hardening have been identified. On the macroscopic scale, classical isotropic hardening models have been determined. At the grain scale, the crystal plasticity constitutive model of Méric and Cailletaud has been identified. The results obtained were compared on the macroscopic scale with identifications carried out on the same material from the tensile and compression tests. The comparison showed that the combination of multiple indentation geometries makes it possible to reproduce the volume behavior of the 316L with acceptable accuracy. For crystal behavior, micropillar compression tests were used to obtain reference data at this scale. The comparison shows a lot of dispersion in both cases. Indeed, some phenomena related to the density of dislocation very variable from one grain to another are responsible of this dispersion. This dislocation density is not taken into account, as a variable, in the used crystal constitutive model. The use of a more physical law integrating the dislocation density and its evolution makes it possible to improve these results. Finally, a new identification method has been proposed. This method is based on estimating and introducing the real indent geometry in the FE model used for identification. The method has been validated in the case of Berkovich tip and shows very promising results.

京都大学 / 0048 / 新制・課程博士 / 博士(エネルギー科学) / 甲第17913号 / エネ博第285号 / 新制||エネ||59(附属図書館) / 30733 / 京都大学大学院エネルギー科学研究科エネルギー基礎科学専攻 / (主査)教授 岸本 泰明, 教授 中村 祐司, 教授 前川 孝 / 学位規則第4条第1項該当 / Doctor of Energy Science / Kyoto University / DFAM

No description available.

Ecological communities are most commonly structured by a mixture of bottom-up processes such as habitat or prey, competition within the same trophic level, and top-down forces from higher trophic levels. Carnivore guilds play a vital role in the broader ecological community by stabilizing or destabilizing food webs. Consequently, factors influencing the structure of carnivore guilds can be critical to patterns in ecosystems. Coyotes (Canis latrans), bobcats (Lynx rufus), gray foxes (Urocyon cinereoargenteus), raccoons (Procyon lotor), red foxes (Vulpes vulpes), and striped skunks (Mephitis mephitis) occur sympatrically throughout much of their geographic ranges in North America and overlap in resource use, indicating potential for interspecific interactions. Although much is known about space use, habitat relationships, and activity patterns of the individual species separately, little is known about factors that facilitate coexistence and how interactions within this guild influence distribution, activity, and survival of the smaller carnivores. For example, gray fox populations appear to have declined in Illinois since the early 1990s and it is unknown if the increase in bobcat and coyote populations during the same time period is the cause. I conducted a large-scale non-invasive carnivore survey using an occupancy modeling framework to quantify factors affecting the structure of this widely-occurring carnivore guild. I used baited remote cameras during 3-week surveys to detect carnivores at 1,118 camera-points in 357 2.6-km2 sections (clusters of 3-4 cameras/section) in the 16 southernmost counties of Illinois (16,058 km2) during January-April, 2008-2010. I collected microhabitat data at each camera-point and landscape-level habitat data for each camera-cluster. In a multi-stage approach, I used information-theoretic methods to develop and evaluate models for detection, species-specific habitat occupancy, multi-species co-occupancy, and multi-season (colonization and extinction) occupancy dynamics. I developed hypotheses for each species regarding the occupancy of areas based on anthropogenic features, prey availability, landscape complexity, and vegetative landcover. I used photographic data, Poisson regression, and mixed-model logistic regression to quantify temporal activity of carnivores in the study area and how interspecific factors influence temporal patterns of activity. Of the 102,711 photographs of endothermic animals I recorded photographs of bobcats (n = 412 photographs), coyotes (n = 1,397), gray foxes (n = 546), raccoons (n = 40,029), red foxes (n = 149) and striped skunks (n = 2,467). Bobcats were active primarily during crepuscular periods, and their activity was reduced with precipitation and higher temperatures. The probability of detecting bobcats at a camera point decreased after a bobcat photograph was recorded, suggesting avoidance of remote cameras. Across southern Illinois, bobcat occupancy at the camera-point and camera-cluster scale (point = 0.24 ± 0.04, cluster = 0.75 ± 0.06) was negatively influenced by anthropogenic features and infrastructure. Bobcats had high rates of colonization (0.86) and low rates of extinction (0.07) during the study, suggesting an expanding population, but agricultural land was less likely to be colonized. The number of coyote photographs decreased with increased temperature, but increased with previous coyote photographs, suggesting an attraction to bait in cold weather. Nearly all camera clusters were occupied by coyotes (cluster = 0.95 ± 0.03) during the entire study. At the camera-point scale, coyote occupancy (overall point = 0.58 ± 0.03) was higher in hardwood forest stands with open understories than in other habitats. Similar to coyotes, gray foxes were more likely to be photographed in cold weather and after a previous detection had occurred. However, gray fox occupancy was much lower (point = 0.13 ± 0.01, cluster = 0.29 ± 0.03) at all scales. At the camera-cluster scale, with a buffer-area size that represented 20% of the estimated home-range size of gray foxes, the species selected spatially-complex areas with high proportions of forest, and low proportions of grassland and agriculture land cover. Gray fox occupancy of camera clusters was positively related to anthropogenic features within 100% estimated home-range buffers. Collectively, the results suggest gray fox occupancy was greatest near, but not in, anthropogenic developments. Red foxes occupied a similar proportion of the study area as gray foxes (point = 0.12 ± 0.02, cluster = 0.26 ± 0.04), but were more closely associated with anthropogenic features. Indeed, at all three scales of red fox occupancy analysis, anthropogenic feature models received more support than other hypotheses. Camera-cluster extinction probabilities were higher for both gray foxes (0.57) and red foxes (0.35) than their colonization rates (gray fox = 0.16, red fox = 0.06), suggesting both species may be declining in southern Illinois. I recorded more striped skunk photographs in January and February (i.e., during the breeding period) than in March and April. Striped skunks occupied a large portion of the study area (point = 0.47 ± 0.01, cluster = 0.79 ± 0.03) and were associated primarily with anthropogenic features, especially if the features were surrounded by agricultural land and not forest. Raccoons were essentially ubiquitous within the study area, being photographed in 99% of camera clusters. In some instances, the presence of other carnivores appeared to be an important factor in the occupancy of the 4 smaller species, but in general, habitat models were more supported than co-occurrence models. Habitat had a stronger influence on the occupancy of gray foxes and red foxes than did the presence of bobcats. However, the level of red fox activity, represented by the number of photographs recorded in a camera cluster, was negatively correlated with bobcat activity. Gray fox occupancy and level of activity were reduced in camera-clusters occupied by coyotes, but were not related to bobcat occupancy. When not considering the presence of coyotes, gray foxes appeared to use camera points with fewer hardwood and more conifer trees, which was counter to previous findings. However, when adding the effect of coyote presence, gray fox point models indicated a positive relationship with hardwood stands. Therefore, gray foxes were more likely to occupy camera points in hardwood stands than conifer stands if coyotes were also present; suggesting that hardwood stands may enhance gray fox-coyote coexistence. The 2 fox species appeared to co-occur with each other at the camera-point scale more frequently than expected on the basis of their individual selection of habitat. Similarly, camera-point occupancy of red foxes was higher when coyotes were present. These apparent canid associations may be a response to locally-high prey abundance or an unmeasured habitat variable. Activity levels of raccoons, bobcats, and coyotes were all positively correlated. Collectively, my results suggest that although gray foxes and red foxes currently coexist with bobcats and coyotes, the foxes have reduced activity in the areas occupied by larger carnivores, especially when bobcats and coyotes are highly active. Further, hardwood stands may contain trees with structure that enhances tree-climbing by gray foxes, a behavior that probably facilitates coexistence with coyotes. Therefore, efforts to manage gray foxes should focus on maintaining and increasing the amount of mature oak-hickory forest, which presumably provides a suitable prey base and refugia from intraguild predation. Additionally, the varying results from different scales of analyses underscore the importance of considering multiple spatial scales in carnivore community studies.

Phenomenological material models such as Johnson-Cook plasticity are often used in finite element simulations of large deformation processes at different strain rates and temperatures. Since the material constants that appear in such models depend on the material, experimental data, fitting method, as well as the mathematical representation of strain rate and temperature effects, the predicted material behavior is subject to uncertainty. In this dissertation, evidence theory is used for modeling uncertainty in the material constants, which is represented by separate belief structures that are combined into a joint belief structure and propagated using impact loading simulation of structures. Yager’s rule is used for combining evidence obtained from more than one source. Uncertainty is quantified using belief, plausibility, and plausibility-decision functions. An evidence-based design optimization (EBDO) approach is presented where the nondeterministic response functions are expressed using evidential reasoning. The EBDO approach accommodates field material uncertainty in addition to the embedded uncertainty in the material constants. This approach is applied to EBDO of an externally stiffened circular tube under axial impact load with and without consideration of material field uncertainty caused by spatial variation of material uncertainties due to manufacturing effects. Surrogate models are developed for approximation of structural response functions and uncertainty propagation. The EBDO example problem is solved using genetic algorithms. The uncertainty modeling and EBDO results are presented and discussed.

This thesis provides a new hierarchical clustering algorithm for graphs, named Paris, which can be interpreted through the modularity score and its resolution parameter. The algorithm is agglomerative and based on a simple distance between clusters induced by the probability of sampling node pairs. It tries to approximate the optimal partitions with respect to the modularity score at any resolution in one run. In addition to the Paris hierarchical algorithm, this thesis proposes four algorithms that compute rankings of the sharpest clusters, clusterings and resolutions by processing the hierarchy output by Paris. These algorithms are based on a new measure of stability for clusterings, named sharp-score. Key outcomes of these four algorithms are the possibility to rank clusters, detect sharpest clusterings scale, go beyond the resolution limit and detect relevant resolutions. All these algorithms have been tested on both synthetic and real datasets to illustrate the efficiency of their approaches. / Denna avhandling ger en ny hierarkisk klusteralgoritm för grafer, som heter Paris, vilket kan tolkas av modularitetsresultatet och dess upplösningsparameter. Algoritmen är agglomerativ och är baserad på ett enda avstånd mellan kluster som induceras av sannolikheten för sampling av nodpar. Det försöker att approximera de optimala partitionerna vid vilken upplösning som helst i en körning. Förutom en hierarkisk algoritm föreslår denna avhandling fyra algoritmer som beräknar rankningar av de bästa grupperna, kluster och resolutioner genom att bearbeta hierarkiproduktionen i Paris. Dessa algoritmer bygger på ett nytt koncept av klusterstabilitet, kallad sharpscore. Viktiga resultat av dessa fyra algoritmer är förmågan att rangordna kluster, upptäcka bästa klusterskala, gå utöver upplösningsgränsen och upptäcka de mest relevanta resolutionerna. Alla dessa algoritmer har testats på både syntetiska och verkliga datamängder för att illustrera effektiviteten i deras metoder.

Heart failure occurs when the heart is not capable to pump blood at a sufficient rate to meet the demands of the body. Depending on the health of the heart, doctors may recommend a heart transplant, but finding a suitable donor is often a long duration process and the patient might be at an advance condition or the patient is not adequate for a heart transplant. In such cases Ventricular assist devices (VAD) are implemented. The purpose of a VAD is to aid the heart to pump the correct amount of blood, by doing so it relives the load that is put on the heart while giving the patient a chance for recovery. This study focuses on observing the hemodynamic effects of implementing a left ventricular assist device (LVAD) along the aortic arch and main arteries. Thrombi creation and transportation is other subject included in the study, due to the fact that thrombi can obstruct blood flow to critical arteries, manly carotid and vertebral. Occlusion of these can lead to a stroke with devastating effects on the neurocognitive functions and even death. A multi-scale CFD analysis a patient specific geometry model is used as well as a lumped system which provides the correct conditions in order to simulate the whole cardiovascular system. The main goal of the study is to understand the difference in flow behavior created by the unsteady pulsatile boundary conditions. The model described in this work has a total cardiac output of 7.0 Liters/ minute, this for a healthy heart. Two cardiac output splits are used to simulate heart failure conditions. The first split consists of 5 Liters/minute flowing through the LVAD cannula and 2 Liters/minute via the aortic root. The second scenario is when heart iv failure is critical, meaning that zero flow is being output by the left ventricle, thus a split of 7 Liter/minute trough the LVAD cannula and 0 Liters/minute traveling through the aortic root. A statistical analysis for the thrombi motion throughout the patient aortic arch was performed in order to quantify the influence that pulsatile flow has on the particles being track. Spherical particles of 2mm, 4mm and 5mm were released and accounted in the statistical analysis for each of the two split configurations. The study focuses on particles that escaped on the outlet boundaries of the upper arteries (Right Carotid, Left Carotid, and Vertebral). Results exhibit the statistical comparison of means for each particle diameter as well as for the overall probability for the steady and unsteady flow condition.