This thesis highlights the role of phase interfaces on phasetransformations in metallic materials. The deviation from localequilibrium at the moving phase interface has been analysed interms of solute drag theory and finite interface mobility. Inparticular the planar growth of proeutectoid ferrite fromaustenite in steel has been studied. The deviation from localequilibrium is caused by dissipation of Gibbs energy bydiffusion inside the phase interface and interface friction. Inthe analysis the interface is divided into three zones and thethermodynamic and kinetic properties are assumed to varycontinuously across the interface. A new model suitable formulticomponent alloys is developed. The model reduces to thefamiliar solute drag model by Cahn under simplifyingconditions. It was demonstrated how the interface model couldbe combined with a method for calculating the volume diffusionin both the growing and parent phases. With this combination ofprocedures the changes in local conditions at the interface, asthe growth rate changes due to long-range diffusion, could bedemonstrated for the case of continuous cooling in an Fe-Nialloy. The critical limit for massive transformation in the Fe-Niand Fe-C systems was calculated and found to lie well below theT0 line for both systems. The calculated limit for Fe-Ni wascompared with a recent experimental study and reasonableagreement was found. For the Fe-C system the limit calculatedwith the present model was compared with a phase-field model.The two approaches showed qualitatively the same behaviour andthe quantitative difference was due to different assumptions onhow properties vary across the interface. Finally, an attempt to simulate the partitionless growth offerrite in austenite in the Fe-Ni-C system was performed. Inthe applied model the dissipation of Gibbs energy inside theinterface and in the nickel spike ahead of the migratinginterface were accounted for. The long-range diffusion ofcarbon in austenite was treated with an approximate analyticalgrowth equation. A continuous change from paraequilibriumconditions and quasi-paraconditions could be shown in anisothermal section of the Fe-Ni-C phase diagram. Partitionlessgrowth starts in a parabolic fashion but slows down. For alloysoutside the limit for quasiparaconditions partitionless growthis predicted to stop abruptly while for alloys inside thatlimit growth approaches a second parabolic growth law. However,the latter case should not be expected in practise because ofimpingement effects.

<p>This thesis highlights the role of phase interfaces on phasetransformations in metallic materials. The deviation from localequilibrium at the moving phase interface has been analysed interms of solute drag theory and finite interface mobility. Inparticular the planar growth of proeutectoid ferrite fromaustenite in steel has been studied. The deviation from localequilibrium is caused by dissipation of Gibbs energy bydiffusion inside the phase interface and interface friction. Inthe analysis the interface is divided into three zones and thethermodynamic and kinetic properties are assumed to varycontinuously across the interface. A new model suitable formulticomponent alloys is developed. The model reduces to thefamiliar solute drag model by Cahn under simplifyingconditions. It was demonstrated how the interface model couldbe combined with a method for calculating the volume diffusionin both the growing and parent phases. With this combination ofprocedures the changes in local conditions at the interface, asthe growth rate changes due to long-range diffusion, could bedemonstrated for the case of continuous cooling in an Fe-Nialloy.</p><p>The critical limit for massive transformation in the Fe-Niand Fe-C systems was calculated and found to lie well below theT0 line for both systems. The calculated limit for Fe-Ni wascompared with a recent experimental study and reasonableagreement was found. For the Fe-C system the limit calculatedwith the present model was compared with a phase-field model.The two approaches showed qualitatively the same behaviour andthe quantitative difference was due to different assumptions onhow properties vary across the interface.</p><p>Finally, an attempt to simulate the partitionless growth offerrite in austenite in the Fe-Ni-C system was performed. Inthe applied model the dissipation of Gibbs energy inside theinterface and in the nickel spike ahead of the migratinginterface were accounted for. The long-range diffusion ofcarbon in austenite was treated with an approximate analyticalgrowth equation. A continuous change from paraequilibriumconditions and quasi-paraconditions could be shown in anisothermal section of the Fe-Ni-C phase diagram. Partitionlessgrowth starts in a parabolic fashion but slows down. For alloysoutside the limit for quasiparaconditions partitionless growthis predicted to stop abruptly while for alloys inside thatlimit growth approaches a second parabolic growth law. However,the latter case should not be expected in practise because ofimpingement effects.</p>

Le compostage constitue un phénomène simple qui nécessite cependant un savoir-faire important puisque la réaction biologique est orientée par le contrôle de certains paramètres (température, oxygène, humidité) faisant intervenir de nombreux phénomènes couplés. Face à la complexité des mécanismes à étudier et dans le but d'optimiser le procédé industriel, l'utilisation d'un modèle de compostage semble donc pertinente afin de mieux comprendre les mécanismes mis en jeu, d'identifier les effets des couplages entre les mécanismes et mettre en avant certains facteurs clés ou encore comparer différents scénarios. L'état de l'art des modèles de compostage montre qu'aujourd'hui, malgré le nombre important de modèles de compostage, il n'en existe aucun capable de prédire, avec une formulation en temps et en espace, la température, la concentration des gaz (oxygène, dioxyde de carbone, diazote, …), la teneur en eau, les transferts entre phase, la dégradation du substrat, et prenant en compte les modifications d'aération. Le développement d'un nouveau modèle a donc été nécessaire pour pouvoir prédire l'évolution de ces variables et étudier leur couplage dans le procédé. Pour cela, la technique de moyenne volumique, appliquée aux équations à l'échelle du pore, a été utilisée et a permis d'obtenir un modèle de compostage à l'échelle de Darcy. Ce modèle prend en compte une phase gazeuse, une phase liquide et une phase solide. La phase gazeuse comprend quatre espèces : oxygène, dioxyde de carbone, azote, et vapeur d'eau. Dans la phase liquide, seule l'eau est considérée. Le séchage est intégré au modèle sous la forme d'un terme d'échange entre les phases gazeuse et liquide. Enfin, un modèle biologique, inclus dans le modèle de compostage, permet de prendre en compte la dégradation du substrat. Celui-ci est divisé en trois fractions : rapidement biodégradable, lentement biodégradable et inerte. Les deux premières fractions sont solubilisées, fournissant une fraction rapidement hydrolysable. Cette fraction est supposée directement consommée par les bactéries. En compostage, procédé aérobie, la dégradation de la matière organique est associée à une consommation d'oxygène et production de dioxyde de carbone, d'eau et de chaleur. L'hypothèse d'équilibre local thermique et chimique a été supposée ici. Cependant, pour l'eau, les 2 approches (Equilibre Local (EL) et Non-Equilibre Local (NEL)) ont été testées numériquement. Les résultats ont montré que lorsque σ, le coefficient d'échange de masse entre la phase gaz et la phase liquide pour l'eau, est compris dans l'intervalle [1, 4], les approches EL et NEL sont équivalentes, avec des temps de calcul moindres pour le cas NEL. Ainsi, pour toutes les simulations, une écriture NEL a été adoptée avec une valeur de σ de 2.5. Des tests ont ensuite permis de montrer la consistance du modèle. Au vu du nombre important de paramètres, une analyse de sensibilité a ensuite été réalisée afin de déterminer quels sont les paramètres qui ont l'impact le plus important sur le procédé. Ainsi, l'analyse a mis en évidence qu'il faut être prudent quant aux valeurs utilisées pour la capacité calorifique, un coefficient de l'isotherme de sorption, de nombreux paramètres du modèles biologiques (ksH, krH, µmax, Xa,0, Tmax, Topt, Xi,0, Xrb,0) et la porosité. Enfin, les résultats fournis par le modèle ont été comparés aux résultats expérimentaux obtenus à l'échelle pilote 1/1000 en usant des conditions opératoires identiques. Les essais de compostage réels ont été réalisés par Veolia Environnement Recherches et Innovation sur un mélange de biodéchets des ménages et de déchets verts... / Composting may look like a simple process. However, it requires an important expertise, as the biological response is governed by the control parameters (temperature, oxygen, moisture content), involving many coupled phenomena. Given the complexity of the studied mechanisms and in order to optimize the process, using a composting model seems relevant to understand the mechanisms involved, identify the effects of coupling between these mechanisms, highlight some key factors or compare different scenarios, in order to optimize the industrial process. The state of the art of composting models in the literature shows that today, despite the large number of composting models, there is none that can predict, with a formulation in time and space, temperature, concentration of gases (oxygen, carbon dioxide, nitrogen, ...), moisture content, transfers between phases, degradation of the substrate, and take into account the changes in aeration. The development of a new model was therefore necessary to predict the evolution of these variables and study their coupling in the process. The technique of volume averaging applied to the pore scale equations has led to a composting model at the Darcy-scale. This model takes into account a gas phase, a liquid phase and a solid phase. The gas phase includes four species: oxygen, carbon dioxide, nitrogen and water vapor. In the liquid phase, only water is considered. Drying is integrated into the model as an exchange term between gas and liquid phases. Finally, the biological model, included in the composting model, allows to take into account the degradation of the substrate. It is divided into three fractions: readily hydrolysable, slowly hydrolysable and inert. The first two fractions are hydrolized, providing a readily assimilable soluble fraction. It is this fraction that is directly consumed by bacteria. In a composting process, degradation of organic matter is associated with oxygen consumption and production of carbon dioxyde, water and heat. The assumption of thermal and chemical local equilibrium was assumed in this work. However, for water, the two approaches (Local Equilibrium (LE) and Local Non-Equilibrium (LNE)) have been numerically tested. The results showed that when , the water mass exchange coefficient between gas and liquid phases, ranges from 1 to 4 s-1, the LE and LNE approaches are equivalent, with less computing time for the LNE case. Thus, for all future simulations, it was decided to adopt a LNE approach with a value of equal to 2.5 s-1. Tests were then carried out to show the consistency of the model. Given the large number of parameters, a sensitivity analysis was performed to determine the parameters that have the greatest impact on the process. This analysis showed that one must be cautious about the values used for the heat capacity, a coefficient of the sorption isotherm, many parameters from the biological model (ksH, krH, μmax, Xa,0, Tmax, Topt, Xi,0, Xrb,0) and porosity, because these are the parameters that affect mainly the process. Finally, the results provided by the model were compared with experimental results obtained at a pilot scale of 1/1000 using identical operating conditions. The composting experiments were carried out by Veolia Environment Research and Innovation with a mixture of household biowaste and green waste. The results on the 1/1000 scale pilot showed that the model is good at capturing the average change in temperature and concentration during the process. The temperature at the central point in particular is very well reproduced by the model. The same applies to the assessment of organic matter degradation. Simulations at industrial scale (1/1) have also been carried out. They have given promising results.