La conception des structures multi-échelle s’appuie sur des modélisations expérimentales et prédictives. Pour accéder à des niveaux de précision élevés, ces modélisations reposent sur des campagnes expérimentales nombreuses et des développements prédictifs sophistiqués analytiques ou numériques qui intègrent des connaissances sur les paramètres d’intérêt. L’intégration de connaissances diminue l’incertitude sur les grandeurs d’intérêt et impacte de façon significative le coût de modélisation des structures multi-échelle, facteur majeur du coût de conception. Le concepteur doit alors être en mesure de maîtriser la pertinence de l’intégration de connaissances pour la prédiction des grandeurs d’intérêt et son impact sur le coût de modélisation. Les recherches menées sont structurées autour du développement d’une méthodologie d’aide à la conception sous incertitudes permettant au concepteur de choisir des combinaisons de modèles prédictifs et expérimentaux, appelées chemins de modélisation, présentant des compromis différents entre le coût de modélisation et l’incertitude sur les paramètres d’intérêt. Le travail se base sur une représentation pyramidale des modélisations expérimentales et prédictives. Les incertitudes aléatoires et épistémiques liées aux matériaux, aux modèles ainsi qu’aux tolérances géométriques sont agrégées et propagées dans la pyramide jusqu’aux grandeurs d’intérêt de la structure. Une méthode adaptative d’estimation du coût de modélisation, basée sur la logique floue, a été proposée. Le problème multi objectif visant à minimiser les incertitudes sur les paramètres d’intérêt et le coût de modélisation est résolu au moyen d’un algorithme « NSGA-II » permettant l’identification de chemins optimisés robustes. Les travaux sont appliqués au cas d’un réservoir composite épais destiné au stockage d’hydrogène. La méthodologie proposée démontre qu’il est possible de rationaliser les modélisations expérimentales et prédictives menées pour obtenir la pression d’éclatement du réservoir avec une précision maîtrisée. Dans un second temps, la méthodologie est utilisée pour obtenir des solutions de reconception sur des réservoirs présentant des volumes plus importants ou plus faibles et atteignant des pressions cibles différentes. Les chemins de modélisations robustes obtenus délivrent des solutions de dimensionnement adaptées aux exigences de reconception présentant un coût de modélisation et un niveau d’incertitude maitrisés. / The design of multi-scale structures is based on predictive and experimental modelling. To achieve a high level of precision, modelling rest on a high number of experimental tests and sophisticated analytical and numerical developments integrating all possible knowledge about the quantity of interest. Adding knowledge into models diminishes the uncertainty on quantities of interest and significantly impacts the cost of modelling, a high impact factor on the design cost. The designer must be able to control the suitability of the integration of knowledge into the prediction of quantities of interest and its impact on the cost of modelling. The research carried out in this work is structured around the development of a methodology of assistance to the design under uncertainties allowing the designer to choose combinations between several predictive and experimental models, called modelling paths, presenting different compromises between the cost of modelling and the uncertainty on quantities of interest. The work is based on a pyramidal representation of experimental and predictive modelling. Random and epistemic uncertainties related to materials, models and geometrical tolerances are aggregated and propagated in the pyramid up to the quantities of interest of the structure. An adaptive method based on fuzzy logics for estimating the cost of modelling has been proposed. The multi objective problem aiming to minimizing the uncertainties on the quantities of interest and the cost of modelling is solved by means of the « NSGA-II » genetic algorithm, allowing to identify robust optimized modelling paths. This methodology is applied to a thick composite vessel for hydrogen storage. The proposed methodology demonstrates the possibility of rationalization of experimental and predictive models carried out to obtain the burst pressure of the vessel with a controlled precision. In a second step, the methodology is used to redesign the vessel considering larger or smaller volumes and with different burst pressure targets. Robust modelling paths obtained deliver design solutions adapted to the redesign requirements with a controlled modelling cost and a managed level of uncertainty.

Advanced understanding of the global carbon budget requires large-scale and long-term information on forest carbon pools and fluxes. In situ and remote sensing measurements have greatly enhanced monitoring of forest carbon dynamics, but incomplete data coverage in space and time results in significant uncertainties in carbon accounting. Although theoretical and mechanistic models have enabled continental-scale and global mapping, robust predictions of forest carbon dynamics are difficult without initialization, adjustment, and parameterization using observations. Therefore, this dissertation is focused on a synergistic combination of lidar measurements and modeling that incorporates biophysical principles underlying forest growth. First, spaceborne lidar data from the Geoscience Laser Altimeter System (GLAS) were analyzed for monitoring and modeling of forest heights over the U.S. Mainland. Results showed the best GLAS metric representing the within-footprint heights to be dependent on topography. Insufficient data sampling by the GLAS sensor was problematic for spatially-complete carbon quantification. A modeling approach, called Allometric Scaling and Resource Limitations (ASRL), successfully alleviated this problem. The metabolic scaling theory and water-energy balance equations embedded within the model also provided a generalized mechanistic understanding of valid relationships between forest structure and geo-predictors including topographic and climatic variables. Second, the ASRL model was refined and applied to predict large-scale patterns of forest structure. This research successfully expanded model applicability by including eco-regional and forest-type variations, and disturbance history. Baseline maps (circa 2005; 1-km2 grids) of forest heights and aboveground biomass were generated over the U.S. Mainland. The Pacific Northwest/California forests were simulated as the most favorable region for hosting large trees, consistent with observations. Through sensitivity and uncertainty analyses, this research found that the refined ASRL model showed promise for prognostic applications, in contrast to conventional black-box approaches. The model predicted temporal evolution of forest carbon stocks during the 21st century. The results demonstrate the effects of CO2 fertilization and climate feedbacks across water- and energy-limited environments. This dissertation documents the complex mechanisms determining forest structure, given availability of local resources. These mechanisms can be used to monitor and forecast forest carbon pools in combination with satellite observations to advance our understanding of the global carbon cycle.

La photocatalyse est un procédé d’oxydation avancée et son utilisation est répandue dans le traitement de l’eau. Cette thèse traite de la dépollution d’eau au sein d’un réacteur original mettant en oeuvre un textile lumineux photocatalytique. Le textile est composé de fibres optiques parallèles situées sur une face d’un tissu fibreux. L’unité d’un tel système est assurée par des points de liage répartis périodiquement fixant les fibres optiques au tissu. Un traitement de microtexturation des fibres optiques permet la création d’une multitude de trous sur leur surface latérale. Une émission de lumière macroscopiquement homogène est provoquée lors de la connexion des fibres optiques à une lampe UV. Un dépôt de catalyseur, tel que le dioxyde de titane, sur l’intégralité du textile, conjuguée au rayonnement UV induit une activité photocatalytique. Cette thèse consiste à l’étude des phénomènes agissant dans un dispositif intégrant le textile lumineux photocatalytique. Dans ce réacteur plan modèle, le textile est confiné entre deux plaques et un écoulement unidirectionnel parallèle aux fibres optiques est mis en oeuvre. La dépollution d’un fluide par photocatalyse résulte du couplage de plusieurs mécanismes : écoulement, transport et réaction. Des modèles numériques sont ainsi développéssur un volume élémentaire représentatif du textile (appelé RVE) pour simuler la dépollution d’une eau comportant une molécule test, à l’échelle microscopique. Cette géométrie est choisie en tenant compte des caractéristiques structurelles du textile photocatalytique. La première étape est l’analyse de l’hydrodynamique au sein du textile, qui couple des écoulements fluide et en milieu poreux. Une étude expérimentale préliminaire a permis l’acquisition de données nécessaires à une représentation réaliste de l’écoulement en milieu poreux. Dans un second temps, le transport est caractérisé par une étude de la distribution des temps de séjour (DTS) au sein du réacteur. Des simulations successives utilisant des conditions aux limites pseudo-périodiques sont réalisées pour calculer numériquement la DTS. Elles sont validées par des mesures expérimentales de traçage de colorant. Enfin, la dégradation d’une molécule test est analysée expérimentalement et numériquement. L’étude numérique présente des approches macroscopique et microscopique. L’étude à l’échelle macroscopique permet de quantifier globalement les performances du réacteur et de fournir des valeurs de constantes cinétiques nécessaires aux simulations àl’échelle microscopique. Une analyse fine et précise de la dépollution est ainsi réalisée au sein du RVE. Elle montre les atouts et limitations du réacteur modèle en termes d’efficacité de dépollution et d’homogénéité de fonctionnement. Des propositions d’améliorations sont finalement émises, notamment une configuration de réacteur comportant un empilement de textiles photocatalytiques. / The photocatalysis is known as an advanced oxidation process and its use is common for the water treatment. This thesis deals with the water depollution within an original reactor integrating the UV-light photocatalytic textile. The textile is composed of parallel optical fibres located on a side of a fibrous fabric. The unity of the system is ensured by bonding points periodically distributed fixing the optical fibres to the fabric. A microtexturization treatment is applied to the optical fibres and a multitude of punctual light sources are thus created on their lateral surface. A light emission macroscopically homogeneous is provided by the connection of optical fibres to an UV lamp. The coating of catalyst, such as titanium dioxide, associated with UV irradiation generates photocatalytic activity. This thesis consists in studying phenomena which occurs within a setup containing the UV-light photocatalytic textile. In this model plane reactor, the textile is confined between two plates and a unidirectional flow parallel to optical fibres is applied. The fluid depollution results of the coupling between several mechanisms : fluid flow, transport and reaction. Numerical models are thus developed on a representative volume element of the textile (called RVE) to simulate at the microscopic scale the depollution of water containing a test molecule. This geometry is designed by taking account the structural characteristics of the photocatalytic textile. The first stage is the analysis of the hydrodynamic within the textile that combines free flow regions and porous medium flows. A preliminary experimental study allows the acquisition of data necessary to a realistic representation of the porous medium flow. Secondly, the transport is characterized by a study of the residence time distribution (RTD) within the reactor. Successive simulations using pseudo-periodic boundary conditions are performed to numerically calculate the RTD. They are validated by experimental measurements using dye tracing. Finally, the degradation of a test molecule is analysed experimentally and numerically. The numerical study presents both approaches macroscopic and microscopic. The study at the macroscopic scale allows to globally quantify the reactor performances. On the other hand, kinetic constants necessary to simulations at the microscopic scale are determined by fitting of the macroscopic model with experimental measurements. An accurate analysis is thus realized within the RVE. It points the advantages and limitations of the model reactor in terms of depollution efficiency and functioning homogeneity. Suggestions of structural improvement are proposed and especially a reactor integrating a stack of photocatalytic textiles.

Nous avons appliqué notre interface QM/MM pour modéliser les propriétés photophysiques et photochimiques du chromophore rétinal dans plusieurs environnements.Nous avons commencé par montrer que la méthylation du squelette carbonné du rétinal, qui transforme une photochimie lente en un processus ultra-rapide, comme dans une protéine, dans une solution de méthanol, modifie l’interaction entre les états excités du rétinal, et favorise la formation d’une espèce transitoire réactive. Nous avons ensuite étudié l’effet direct de l’environnement dans le cas de mimiques de la rhodopsine, où des mutations ponctuelles de quelques acides aminés donnent des systèmes qui absorbent sur toute la gamme du visible. En combinaison avec la spectroscopie pompe-sonde ultra-rapide, notre méthode a montré que le potentiel électrostatique autour du rétinal peut affecter la forme des surfaces d’énergies potentielles excitées, et peut moduler le temps de vie de l’état excité ainsi que le lieu de photo-isomérisation. Ensuite, nous avons montré que l’état de protonation standard des acides aminés proches du rétinal dans la bacteriorhodopsine mènent à une surestimation de l’énergie d’absorption,alors que la protonation du résidu Asp212 donne des résultats plus précis ; nous souhaitons maintenant valider ce modèle par le calcule des propriétés de fluorescence et de temps de vie de l’état excité. Enfin, nous avons modélisé la photophysique de la base de Schiff non-protonée d’un pigment UV, où une photochimie originale, et non encore documentée, a lieu, impliquant notamment un état doublement excité. Ces études ont montré la robustesse de notre potentiel QM/MM pour modéliser une large gamme d’environnements. / We have used our QM/MM interface to model the photochemical and photophysical properties of the retinal chromophore in several environments.First, we proved that methylation of the retinal backbone, which converts a slow photochemistry to an ultra-fast protein-like behaviour in methanol solution, modifies the interplay between the retinal excited states, favouring the formation of a photo-active transient intermediate. Then, we have studied the direct effect of the environment in the case of rhodopsin mimics, where point mutations of a few amino-acids lead to systems that can absorb in the wide visible range. Combined with ultra-fast pump-probe spectroscopy, our method has shown that the electrostatic potential around the retinal can affect the shape of the excited potential energy surface, and is able to tune the excited state lifetime as well as the location of the photoisomerization. Next, we showed that the currently accepted protonation state of amino-acids in the vicinity of the retinal in bacteriorhodopsin leads to a strongly blue shifted absorption, while the protonation of Asp212 leads to accurate results; we now aim toward a validation of this protonation by computation of fluorescence and excited state lifetime. Finally, we have modeled the photophysics of the unprotonated Schiff base in a UV-pigment, where an original an previously unreported photochemistry takes place, especially with the direct involvement of a doubly excited state. These studies have shown the reliability of our QM/MM potential for modeling a wide range of different environments.

Embryonic development depends on the precise coordination of cell fate specification, patterning and morphogenesis. Although great strides have been made in the molecular understanding of each of these processes, how their interplay governs the formation of complex tissues remains poorly understood. New techniques for experimental manipulation and image quantification enable the study of development in unprecedented detail, resulting in new hypotheses on the interactions between known components. By expressing these hypotheses in terms of rules and equations, computational modeling and simulation allows one to test their consistency against experimental data. However, new computational methods are required to represent and integrate the network of interactions between gene regulation, signaling and biomechanics that extend over the molecular, cellular and tissue scales. In this thesis, I present a framework that facilitates computational modeling of multiscale multicellular systems and apply it to investigate pancreatic development and the formation of vascular networks. This framework is based on the integration of discrete cell-based models with continuous models for intracellular regulation and intercellular signaling. Specifically, gene regulatory networks are represented by differential equations to analyze cell fate regulation; interactions and distributions of signaling molecules are modeled by reaction-diffusion systems to study pattern formation; and cell-cell interactions are represented in cell-based models to investigate morphogenetic processes. A cell-centered approach is adopted that facilitates the integration of processes across the scales and simultaneously constrains model complexity. The computational methods that are required for this modeling framework have been implemented in the software platform Morpheus. This modeling and simulation environment enables the development, execution and analysis of multi-scale models of multicellular systems. These models are represented in a new domain-specific markup language that separates the biological model from the computational methods and facilitates model storage and exchange. Together with a user-friendly graphical interface, Morpheus enables computational modeling of complex developmental processes without programming and thereby widens its accessibility for biologists. To demonstrate the applicability of the framework to problems in developmental biology, two case studies are presented that address different aspects of the interplay between cell fate specification, patterning and morphogenesis. In the first, I focus on the interplay between cell fate stability and intercellular signaling. Specifically, two studies are presented that investigate how mechanisms of cell-cell communication affect cell fate regulation and spatial patterning in the pancreatic epithelium. Using bifurcation analysis and simulations of spatially coupled differential equations, it is shown that intercellular communication results in a multistability of gene expression states that can explain the scattered spatial distribution and low cell type ratio of nascent islet cells. Moreover, model analysis shows that disruption of intercellular communication induces a transition between gene expression states that can explain observations of in vitro transdifferentiation from adult acinar cells into new islet cells. These results emphasize the role of the multicellular context in cell fate regulation during development and may be used to optimize protocols for cellular reprogramming. The second case study focuses on the feedback between patterning and morphogenesis in the context of the formation of vascular networks. Integrating a cell-based model of endothelial chemotaxis with a reaction-diffusion model representing signaling molecules and extracellular matrix, it is shown that vascular network patterns with realistic morphometry can arise when signaling factors are retained by cell-modified matrix molecules. Through the validation of this model using in vitro assays, quantitative estimates are obtained for kinetic parameters that, when used in quantitative model simulations, confirm the formation of vascular networks under measured biophysical conditions. These results demonstrate the key role of the extracellular matrix in providing spatial guidance cues, a fact that may be exploited to enhance vascularization of engineered tissues. Together, the modeling framework, software platform and case studies presented in this thesis demonstrate how cell-centered computational modeling of multi-scale and multicellular systems provide powerful tools to help disentangle the complex interplay between cell fate specification, patterning and morphogenesis during embryonic development.

The presented research investigated the viability of a double-braided synthetic fiber rope for providing improved performance of steel moment frames subjected to earthquake-induced ground motions. A series of experimental tests, including a 1:3-scale dynamic test and 1:6-scale shaking table tests, was conducted using Northridge ground-motion input. A series of nonlinear dynamic analytical studies, using DRAIN-2DX, was conducted to develop the experimental tests. Throughout experimental testing, the ropes exhibited a hyper-elastic loading response and a reduced-stiffness unloading response. A conditioning cycle was defined as a loading cycle induced in the rope above the highest load expected to be experienced by the rope, and was determined to be requisite for ropes intended to be used for the stated objectives of the research program. After experiencing a conditioning cycle, the rope response returned to initial conditions without permanent deformation, demonstrating repeatability of response through several loading cycles below the conditioning load. In the 1:6-scale shaking-table experiments, the ropes drastically improved the performance of the steel moment frames. Maximum and residual drift were reduced significantly, with a corresponding minimal increase to the maximum base shear. Base shear was reduced at several peaks subsequent to the initial pulse of the Northridge ground-motion input. The analytical model developed was excellent for predicting elastic response of the 1:6-scale shaking table experiments and adequate for the purpose of planning shaking table studies. Correlation of peak rope forces between the analytical model and experimental results was poor, and was attributed to limitations of the pre-defined elements used to represent the rope devices in the software program. The inability of the elements to capture the complex unloading response of the rope was specifically noted. / Ph. D.

Geological Carbon Storage (GCS) is considered as one of the key techniques to reduce the rate of atmospheric emissions of CO2 and thereby to contribute to controlling the global warming. A successful application of a GCS project requires the capability of the formation to trap CO2 for a long term. In this context, processes related to CO2 trapping and also possible leakage of CO2 to the near surface environment need to be understood. The overall aim of this thesis is to understand the flow and transport of CO2 through porous media in the context of geological storage of CO2. The entire range of scales, including the pore scale, the laboratory scale, the field experiment scale and the industrial scale of CO2 injection operation are addressed, and some of the key processes investigated by means of experiments and modeling.  First, a numerical model and laboratory experimental setup were developed to investigate the CO2 gas flow, mimicking the system in the near-surface conditions in case a leak from the storage formation should occur. The system specifically addressed the coupled flow and mass transport of gaseous CO2 both in the porous domain as well as the free flow domain above it. The comparison of experiments and modelling results showed a very good agreement indicating that the model developed can be applied to evaluate monitoring and surface detection of potential CO2 leakage. Second, the field scale CO2 injection test carried out in a shallow aquifer in Maguelone, France was analyzed and modeled. The results showed that Monte Carlo simulations accounting for the heterogeneity effects of the permeability field did capture the key observations of the monitoring data, while a homogeneous model could not represent them. Third, a numerical model based on phase-field method was developed and model simulations carried out addressing the effect of wettability on CO2-brine displacement at the pore-scale. The results show that strongly water-wet reservoirs provide a better potential for the dissolution trapping, due to the increase of interface between CO2 and brine with very low contact angles. The results further showed that strong water-wet conditions also imply a strong capillary effect, which is important for residual trapping of CO2. Finally, numerical model development and model simulations were carried out to address the large scale geological storage of CO2 in the presence of impurity gases in the CO2 rich phase. The results showed that impurity gases N2 and CH4 affected the spatial distribution of the gas (the supercritical CO2 rich phase), and a larger volume of reservoir is needed in comparison to the pure CO2 injection scenario. In addition, the solubility trapping significantly increased in the presence of N2 and CH4.

Le plâtre est un matériau de grande disponibilité et très connu pour ses qualités : il est favorable à la protection de l’environnement, assez malléable, de faible densité, aux propriétés fonctionnelles remarquables (coupe-feu, isolant thermique, régulateur de l’hygrométrie des enceintes), décoratif, ... Ce qui justifie l’intérêt accordé à ce matériau pour les constructions. Sa grande fragilité préoccupante est à l’origine des travaux de recherches dans le monde entier en vue de son renforcement. Les fibres de verre et de sisal sont les renforts les plus utilisés à ce jour. Le renforcement par des fibres végétales est de plus en plus recherché. La texture micro structurale poreuse du plâtre favorise son caractère d’isolant thermique. Les textures mises en œuvre jusqu’à présent sont limitées à des porosités comprises entre 30 et 55%. La réduction du coût de ce matériau pour une large utilisation est encore possible et souhaitée. Deux leviers sont exploités dans ce travail, notamment un allègement de la masse de plâtre pour augmenter le taux de porosité et un renforcement de la tenue mécanique par incorporation de fibres végétales produites localement. L’objectif de ce travail est d’évaluer les caractéristiques mécaniques, thermiques et hygrométriques d’un matériau constitué de plâtre pris, à grande porosité, renforcé d’une nouvelle fibre végétale : le Rhecktophyllum Camerunense (RC), une fibre des forêts humides équatoriales. La fibre de sisal, d’utilisation connue pour le renforcement du plâtre, sert de référence à des fins de comparaison. Une série d’expérimentations est menée à cet effet. Une caractérisation physico-chimique des constituants est effectuée, des essais mécaniques de traction et de flexion sont effectués sur les constituants et les matériaux composites plâtre/fibres résultants, la cinétique d’adsorption d’humidité par les constituants et le matériau fibreux est suivie. Le comportement thermique des matériaux plâtre et plâtre/fibres est aussi mesuré. Les fibres utilisées, le sisal et le RC, sont à fort taux de cellulose (entre 49 et 78,8%), la fibre de RC est tubulaire avec 35,5% de porosité. Le plâtre est gâché à l’eau déminéralisée à un rapport massique E/P égal à 1 à partir de la poudre de semihydrate β. Sa microstructure cristalline est constituée de cristaux de gypse sous forme d’aiguilles enchevêtrées avec des vides intercristallins. Sur le plan du comportement mécanique, les résultats obtenus révèlent que le plâtre se montre fragile et présente un module d’élasticité en traction de 1,72 GPa, une résistance à la traction de 0,86 MPa et un allongement à la rupture de 1,16%. En flexion trois points, son module d’élasticité est de 0,64 GPa et sa contrainte à la rupture, de 0,13 MPa. La fibre de sisal est raide et fragile. Son module d’élasticité est compris entre 9 et 21 GPa, elle admet un allongement à rupture de 3 à 7%. Par contre, la fibre de RC est assez ductile avec un module d’Young moyen de 0,7 GPa et un allongement à rupture de 24,2%. L’adhésion du plâtre sur les fibres est faible : il adhère plus sur le sisal que sur le RC. Le sisal renforce mieux le plâtre par une augmentation plus sensible du module d’élasticité de 42,5%, contre 16,3% pour le RC, ce dernier lui apportant plutôt une grande ductilité élastique. Les fibres de RC apportent le maximum de renforcement en traction au plâtre lorsqu’elles sont tissées en unidirectionnel et en flexion lorsqu’elles sont uniformément réparties dans le volume suivant la direction longitudinale de la structure. (...) / The plaster is a material of high availability and very known for its qualities : it is favourable to the protection of the environment, quite malleable, of low density, its functional properties are remarkable (firewall, thermal insulation, regulator of the hygroscopy of enclosures), decorative, ... What justifies the interest attached to this material for constructions. Its great alarming brittleness is at the origin of the research tasks in the whole world for its strengthening. The glass fibers and sisal are the more used reinforcements to this day. The strengthening by plant fibers is more and more researched. The microstructure of the plaster is porous ; that promotes its heat insulation character. The textures implemented so far are limited to porosities ranging between 30 and 55%. The reduction of cost of this material for a wide use is still possible and desired. Two levers are exploited in this work, in particular a lightening of the plaster weight to increase the proportion of air voids and a reinforcement of the mechanical resistance with locally produced fibers. The objective of this work is to evaluate the mechanical, thermal and hygrometrical characteristics of a material made up of harden plaster, with high porosity, strengthened by a new plant fiber : the Rhecktophyllum Camerunense (RC), a fiber of humid equatorial forests. The sisal fiber, of known use for the strengthening of the plaster, serves as a reference for comparison purposes. A serie of experiments is conducted to this effect. A physicochemical characterization of constituents is performed. Mechanical tests of tensile and of bending are performed on the constituents and the resulting plaster/fiber composite materials. The kinetic adsorption of moisture by the constituents is followed. The thermal behaviour of plaster and plaster/fiber is also measured. The fibers used, sisal and RC are with high rates of cellulose (between 49 and 78.8% ), the fiber of RC is tubular with 35.5 % of porosity. The plaster is dissolved in demineralized water to a mass ratio W/P equals to 1 from the powder of semihydrate β. Its crystalline microstructure is composed of gypsum crystals in the form of needles tangled with the empty intercristallins. As far as the mechanical behavior is concerne, the result reveals that the plaster is weak, its Young’s modulus in tensile is 1.72 GPa, its tensile strength is 0.86 MPa and its elongation at break is 1.16 %. In three points bending test, its modulus of elasticity is 0.64 GPa and its constraint at break is 0.13 MPa. The sisal fiber is stiff and fragile. Its Young’s modulus is between 9 and 21 GPa, it admits an elongation at break of 3 to 7 %. On the other side, the fiber of RC is quite ductile : the means of Young’s module is 7 GPa and the elongation at break is 24.2 %. The adhesion of the plaster on the fiber surface is low : it adheres more on the sisal than on the RC. The sisal strengthened better the plaster with a sensitive increase of the Young’s modulus of 42.5 %, against 16.3 % for the RC. But the RC fiber gives rather high elastic ductility. The fibers of RC deliver maximum capacity in tensile to the plaster when they are woven into unidirectional. They offer high capacity in bending when they are uniformly distributed inside the volume according to the longitudinal direction of the structure. (...)

Les composites C/C sont utilisés dans les domaines du spatial et de l'aéronautique pour leurs excellentes propriétés thermomécaniques depuis la température ambiante jusqu'aux très hautes températures (> 3000°C). Ces matériaux ont une architecture complexe constituée de nappes de fibres stratifiées et aiguilletées. Leur utilisation en zones fortement sollicitées et à haute température nécessite une maîtrise des propriétés thermiques et mécaniques. Actuellement, la conception du matériau se fait de manière empirique et itérative. Pour l'accélérer, le développement d'un modèle numérique multi-échelle prédictif du comportement du composite C/C est proposé. Ce matériau a tout d'abord été caractérisé morphologiquement à ses différentes échelles caractéristiques, les propriétés thermomécaniques de ses constituants élémentaires ont également été identifiées. A l'échelle microscopique, les fils sont représentés de façon homogène et thermoélastique à partir des taux de constituants qui leurs sont associés. A l'échelle mésoscopique, deux aspects morphologiques semblent prédominants : son architecture ainsi que ses porosités et endommagements. Leur effet sur le comportement effectif du composite C/C est étudié dans le but d'évaluer leur influence relative et d'aboutir à une description suffisamment fine de leurs morphologies dans la modélisation effectuée. Un modèle de matériau idéalisé ainsi qu'un modèle basé images ont été développés. La simulation d'essais macroscopiques a révélé que ces deux aspects avaient un effet non-négligeable sur le comportement effectif du composite C/C et ont permis le développement et la validation d'un modèle prédictif multi-échelle de ce matériau, prenant en compte les caractéristiques précédentes, et permettant le lien entre l'échelle de ses constituants élémentaires et celle macroscopique. / C/C composites are used in space and aeronautics for their excellent thermomechanical properties from room temperature to very high temperatures (> 3000°C). These materials have a complex architecture consisting of layers of laminated and needled fibers. Its use in highly stressed areas and at high temperature requires control of thermal and mechanical properties. Currently, the design of the material is done empirically and iteratively. To accelerate it, the development of a multi-scale digital model of the C/C composite is proposed. This material was first morphologically characterized at its different characteristic scales, the thermomechanical properties of its elementary constituents were also identified. At the microscopic scale, the wires are represented homogeneously and thermoelastically from the constituent levels associated with them. At the mesoscopic scale, two morphological aspects seem to predominate: its architecture as well as its porosities and damages. Their effect on the effective behavior of the C/C composite is studied in order to evaluate their relative influence and to arrive at a sufficiently fine description of their morphologies in the modeling carried out. An idealized material model as well as an image based model have been developed. The simulation of macroscopic tests revealed that both aspects had a non-negligible effect on the effective behavior of the C/C composite and allowed the development and validation of a multi-scale predictive model of this material, taking into account the preceding characteristics, and allowing the link between the scale of its elementary constituents and the macroscopic one.

De nombreuses études ont permis de développer des modèles de plasticité cristalline rendant compte de l’anisotropie d’un monocristal. Les matériaux à solidification dirigée (DS) peuvent être simulés avec des modèles semblables, sous réserve de connaître la morphologie et l'orientation cristallographique des grains contenus dans la pièce. Pour éviter ces analyses microstructurales coûteuses, il est possible de développer des approches, déterministes ou analysant les dispersions de la réponse sur un lot de << pièces synthétiques >> résolues par la méthode des éléments finis. Dans cette étude, nous avons tenté d'apporter les outils nécessaires aux deux types de modélisation. Avant tout, un modèle du monograin de DS200+Hf a été identifié pour une gamme de température allant de l’ambiante à 1200°C. Ensuite, plusieurs montées d'échelle ont été envisagées, d'abord sur un volume élémentaire représentatif (VER) puis sur une structure tridimensionnelle (éprouvette cruciforme). Sur le VER, la réponse de plusieurs modèles micromécaniques a été confrontée à des calculs de référence utilisant la méthode des éléments finis. Puis, le comportement mécanique d'une éprouvette cruciforme en DS200+Hf a été étudié, en réalisant des essais expérimentaux biaxiaux qui, ensuite, ont permis d'évaluer les prévisions du modèle. Ces résultats amènent à s'interroger sur la modélisation adaptée aux structures oligogranulaires (i.e. constituées d’un faible nombre de grains) : faut-il mailler explicitement l'échelle locale (les grains) dans la structure ou malgré la non-séparabilité des échelles, le modèle homogénéisé continue-t-il de fournir des résultats satisfaisants ? / Various studies were aimed at developing crystal plasticity models to account for the anisotropic mechanical behaviour of single crystals. Directionally solidified (DS) materials can be modeled using such approaches, taking into account the underlying crystallographic structure. It requires the knowledge of the position, shape and crystallographic orientations of grains. To prevent heavy microstructure analyses, other models have to be developed for industrial calculations, using homogenization theory or considering a batch of synthetic pieces calculated using Crystal Plasticity Finite Elements Method (CPFEM). The aim of this thesis is to bring computational tools to carry out the two types of modeling for industrial applications. First of all, a crystal plasticity model for one grain of DS200+Hf is defined ranging from room temperature to 1200°C. Some scale transition rules, using full-field or mean-field approaches, are studied first in the theoretical case of a representative volume element (RVE) and then on tri-dimensional structures in order to access overall and local responses. For the RVE responses, micromechanical models are compared with a reference produced by CPFEM for various loadings. Moreover, the mechanical behaviour of a DS200+Hf cruciform specimen is studied. Biaxial tests with digital image correlation allow us to check the model predictions. These results raise questions about the modeling of oligogranular structures (i.e. with a small number of grains): should it be accepted that the local scale must be explicitly meshed, or, despite the fact that scale separability is not ensured, can we consider that the homogenized model still produces reliable results?