In the aerospace industry, burr removal is a very important part of the manufacturing process. Stacks of material on sections of an aircraft are assembled and drilled by hand. Due to extensive burr formation the sheets must be destacked so that burrs can be removed and then the stacks are reassembled and fastened together. If burrs are minimized in the drilling process, this would reduce the necessity for the sheets to be destacked and deburred. One approach to minimizing burrs is to lower the thrust force in drilling through suitable modification of the drill geometry such as the use of a step drill. Although prior researchers have analyzed different drill geometries such as step drills and their effect on hole exit burr formation in the drilling process through experimentation, no work has been reported on modeling and analysis of step drilling forces and their relationship to burr formation as a function of the step drill geometry parameters. Consequently, this thesis focuses on the modeling of the thrust force and torque for step drills and analyzes their relationship with burr size as a function of the step drill geometry parameters. In the first step, a mechanistic model for thrust and torque in drilling is implemented for a standard twist drill. This mechanistic model is then adapted to predict the thrust and torque for a step drill. Subsequently, experiments are performed to validate the mechanistic model and to evaluate burr formation with standard and step drills. The influence of thrust and torque on hole exit burr formation is analyzed for different step drill geometries and experimental feeds and speeds. The results show that the predicted thrust and torque values for both drill geometries are in good agreement with measured values, although the torque model consistently underpredicts. For standard drill geometry in the calibration tests, the average error in the thrust prediction is 7.09% and the average error in the torque prediction is -18.05%. In validation tests, the average error for predicted thrust is 2.29% and the average error for predicted torque is -18.46%. For the step drill model the average error in thrust is 0.72% while the average error in torque is -8.72%. In addition, a reduction in the predicted thrust force for a step drill relative to the standard twist drill is found to correlate well with a reduction in the measured burr size. However, further reduction in the thrust force by varying the step angle and diameter ratio do not correlate well with the measured burr size. Likely reasons for these results are presented in this thesis.

Em um contexto global de aumento da demanda por produtos agrícolas, especialmente de proteína animal, concomitante às mudanças climáticas e escassez de áreas para expansão agrícola, o processo de intensificação sustentável da pecuária é colocado como um importante \"piloto\" frente às políticas públicas setoriais. O Plano ABC, exemplo deste tipo de política pública, destina recursos através do Programa ABC para a recuperação de pastagens degradadas. A identificação de pastagens degradadas se dá através de indicadores de baixa produtividade da pecuária, principalmente pautados na baixa lotação animal. Esta premissa nem sempre é verdadeira, pois há áreas com baixa capacidade de suporte, em função de limitações físicas, mas que não se encontram necessariamente degradadas. A utilização de modelos biofísicos de crescimento de forragem é uma possível alternativa de abordagem à problemática, que pode auxiliar na identificação espacial de áreas onde a pecuária poderia se intensificar sem implicar em impactos ambientais, de modo a melhor direcionar políticas públicas que busquem estimular este processo. O objetivo deste estudo foi determinar para o estado de São Paulo a eficiência na produção e o yield gap (lacuna de produtividade) explorável da pastagem via modelagem mecanística para identificar as áreas com maior potencial de intensificação da pecuária e compará-las com a geografia de aplicação do crédito agrícola do Programa ABC para recuperação de pastagens degradadas. O estudo foi organizado da seguinte forma: a primeira parte (Capítulo 2) apresenta uma discussão acerca das áreas de pastagem com maior potencial de incremento de produtividade com base nas simulações no CROPGRO Perennial Forage para crescimento da pastagem de modo a definir a eficiência produtiva e os yield gaps da pastagem para o estado de São Paulo; a segunda parte (Capítulo 3) traz a análise de desempenho do Programa ABC em relação à aplicação desse recurso para recuperação de pastagem degradada (RPD), ou seja, avaliar se o crédito está sendo aplicado em regiões de maior potencial para incremento de produtividade, com base nos valores de yield gap explorável encontrados anteriormente. A eficiência média atual da produção de pastagens (razão entre produtividade real e produtividade atingível) encontrada via simulação no estado de São Paulo foi de 46,4% e um yield gap médio explorável de 5,8 Mg.ha-1.ano-1 de biomassa úmida. Estes resultados indicam que há um significativo espaço para incremento de produtividade das pastagens neste estado. A priorização de áreas de maior yield gap para intensificação pode contribuir para a redução da pressão por abertura de novas áreas de pastagem em locais com baixo potencial produtivo, áreas estas que podem ser ocupadas para produção agrícola ou para a regularização ambiental, evitando a expansão do desmatamento. O presente estudo buscou contribuir, através de resultados em escala estadual, para futuros estudos em escala nacional, mostrando que o uso de modelos mecanísticos como o CROPGRO Perennial Forage podem ser instrumentos de grande utilidade na avaliação e no direcionamento de políticas públicas condizentes com a construção de diretrizes para o desenvolvimento rural sustentável. / In a global context of increasing demand for agricultural resources, especially animal protein, combined with climate change and scarcity of areas for agricultural expansion, the process of sustainable intensification of livestock farming is placed as an important public policie issue. The ABC Plan, an example of this type of public policy, allocates resources through the ABC Program for the recovery of degraded pastures. The identification of degraded pastures occurs through indicators of low productivity of livestock, mainly based on low animal stocking (UA.ha-1). This premise is not always true, as there are areas with low support capacity , due to physical limitations, but that are not necessarily degraded. The use of biophysical models of forage growth is an alternative approach to this problem, which may help in the spatial identification of areas where livestock could be intensified without environmental impacts, in order to guide public policies seeking to stimulate this process . The objective of this study was to determine pasture yield efficiency and yield gap for São Paulo State , using models to identify areas with the greatest potential for livestock intensification and compare them with the geography of the ABC program application. The study was organized as follows: the first part (Chapter 2) presents a discussion about pasture areas with the greatest potential for productivity increase, based on simulations in CROPGRO Perennial Forage for the growth of Brachiaria brizantha cv Marandu in order to define the productive efficiency and the yield gaps of the pasture for São Paulo State; the second part (Chapter 3) presents the performance analysis of the ABC Program in relation to the application of this resource for degraded pasture recovery (RPD), i.e., to evaluate if the credit is being applied in regions with greater potential for productivity increase, based on the explorable yield gap values previously found. The current average efficiency of pasture production (ratio between real productivity and attainable productivity) found by simulation in São Paulo State was 46.4% and an average yield gap of 5.8 Mg.ha-1.year-1 . These results indicate that there is a significant space to increase pasture productivity in this State. Prioritization of areas with a higher yield gap for intensification may contribute to the reduction of pressure for opening new pasture areas in areas with low productive potential, areas that can be occupied for agricultural production or for environmental regulation, avoiding the expansion of deforestation.The present study sought to contribute, through statewide results, to future studies on a national scale, showing that the use of mechanistic models, such as CROPGRO Perennial Forage, can be very useful for evaluating and directing public policies aimed at rural sustainable development.

Em um contexto global de aumento da demanda por produtos agrícolas, especialmente de proteína animal, concomitante às mudanças climáticas e escassez de áreas para expansão agrícola, o processo de intensificação sustentável da pecuária é colocado como um importante \"piloto\" frente às políticas públicas setoriais. O Plano ABC, exemplo deste tipo de política pública, destina recursos através do Programa ABC para a recuperação de pastagens degradadas. A identificação de pastagens degradadas se dá através de indicadores de baixa produtividade da pecuária, principalmente pautados na baixa lotação animal. Esta premissa nem sempre é verdadeira, pois há áreas com baixa capacidade de suporte, em função de limitações físicas, mas que não se encontram necessariamente degradadas. A utilização de modelos biofísicos de crescimento de forragem é uma possível alternativa de abordagem à problemática, que pode auxiliar na identificação espacial de áreas onde a pecuária poderia se intensificar sem implicar em impactos ambientais, de modo a melhor direcionar políticas públicas que busquem estimular este processo. O objetivo deste estudo foi determinar para o estado de São Paulo a eficiência na produção e o yield gap (lacuna de produtividade) explorável da pastagem via modelagem mecanística para identificar as áreas com maior potencial de intensificação da pecuária e compará-las com a geografia de aplicação do crédito agrícola do Programa ABC para recuperação de pastagens degradadas. O estudo foi organizado da seguinte forma: a primeira parte (Capítulo 2) apresenta uma discussão acerca das áreas de pastagem com maior potencial de incremento de produtividade com base nas simulações no CROPGRO Perennial Forage para crescimento da pastagem de modo a definir a eficiência produtiva e os yield gaps da pastagem para o estado de São Paulo; a segunda parte (Capítulo 3) traz a análise de desempenho do Programa ABC em relação à aplicação desse recurso para recuperação de pastagem degradada (RPD), ou seja, avaliar se o crédito está sendo aplicado em regiões de maior potencial para incremento de produtividade, com base nos valores de yield gap explorável encontrados anteriormente. A eficiência média atual da produção de pastagens (razão entre produtividade real e produtividade atingível) encontrada via simulação no estado de São Paulo foi de 46,4% e um yield gap médio explorável de 5,8 Mg.ha-1.ano-1 de biomassa úmida. Estes resultados indicam que há um significativo espaço para incremento de produtividade das pastagens neste estado. A priorização de áreas de maior yield gap para intensificação pode contribuir para a redução da pressão por abertura de novas áreas de pastagem em locais com baixo potencial produtivo, áreas estas que podem ser ocupadas para produção agrícola ou para a regularização ambiental, evitando a expansão do desmatamento. O presente estudo buscou contribuir, através de resultados em escala estadual, para futuros estudos em escala nacional, mostrando que o uso de modelos mecanísticos como o CROPGRO Perennial Forage podem ser instrumentos de grande utilidade na avaliação e no direcionamento de políticas públicas condizentes com a construção de diretrizes para o desenvolvimento rural sustentável. / In a global context of increasing demand for agricultural resources, especially animal protein, combined with climate change and scarcity of areas for agricultural expansion, the process of sustainable intensification of livestock farming is placed as an important public policie issue. The ABC Plan, an example of this type of public policy, allocates resources through the ABC Program for the recovery of degraded pastures. The identification of degraded pastures occurs through indicators of low productivity of livestock, mainly based on low animal stocking (UA.ha-1). This premise is not always true, as there are areas with low support capacity , due to physical limitations, but that are not necessarily degraded. The use of biophysical models of forage growth is an alternative approach to this problem, which may help in the spatial identification of areas where livestock could be intensified without environmental impacts, in order to guide public policies seeking to stimulate this process . The objective of this study was to determine pasture yield efficiency and yield gap for São Paulo State , using models to identify areas with the greatest potential for livestock intensification and compare them with the geography of the ABC program application. The study was organized as follows: the first part (Chapter 2) presents a discussion about pasture areas with the greatest potential for productivity increase, based on simulations in CROPGRO Perennial Forage for the growth of Brachiaria brizantha cv Marandu in order to define the productive efficiency and the yield gaps of the pasture for São Paulo State; the second part (Chapter 3) presents the performance analysis of the ABC Program in relation to the application of this resource for degraded pasture recovery (RPD), i.e., to evaluate if the credit is being applied in regions with greater potential for productivity increase, based on the explorable yield gap values previously found. The current average efficiency of pasture production (ratio between real productivity and attainable productivity) found by simulation in São Paulo State was 46.4% and an average yield gap of 5.8 Mg.ha-1.year-1 . These results indicate that there is a significant space to increase pasture productivity in this State. Prioritization of areas with a higher yield gap for intensification may contribute to the reduction of pressure for opening new pasture areas in areas with low productive potential, areas that can be occupied for agricultural production or for environmental regulation, avoiding the expansion of deforestation.The present study sought to contribute, through statewide results, to future studies on a national scale, showing that the use of mechanistic models, such as CROPGRO Perennial Forage, can be very useful for evaluating and directing public policies aimed at rural sustainable development.

Studies in grassland management and ecology have always been challenging because of the large amount and great variation of the entities representing and affecting the system. Despite that, we were able to progress significantly in range experimentation in the Campos, in Southern Brazil. Along past thirty years, a large amount of data and information was generated, from vegetation production to components of intake. In an attempt to integrate the information available, seeking for a deeper understanding of the functioning of native grasslands, we propose adapting a mechanistic vegetation model, aggregated of a spatialized grazing component to create PampaGraze. This model was developed for temperate perennial grasslands, and was adapted and tested for subtropical, C4-dominated grasslands of the Campos of Southern Brazil (Chapter III). Despite the limited capacity of field data for validating, the model was able to relatively well simulate the trends in vegetation production along the year and seasons, while overpredicting herbage production during peak growing season. The structure of the model as it is did not allow for an accurate simulation slow-growing, tussock-forming species. Further, we developed and integrated a grazing model, based on a hybrid approach of the classical mechanistic equations of the prey model (STEPHENS & KREBS, 1986), and experimental data on foraging behaviour measured on native grasslands (Chapter IV). The model was very successful on predicting the components of intake, and responded well to variation of components in relation to changes in vegetation and to selectivity pressures, compared to available literature. Regardless of the limitations on the vegetation model, we were able to further explore the relationships of components of intake, identifying possible major limitations for herbage consumption, thus animal performance, in native grasslands. A significant progress was achieved with this thesis, but still long ways to go with this project. A list of suggestions for further developments can be found in Chapter V. We identified the emergent needs for field studies on parameters and morphogenesis, for improving predictions of the vegetation model, as well as structural points of the model that could be addressed for better representation of natural phenomena. This thesis is the first step towards a more detailed and reliable tool for studying and predicting the behaviour of vegetation dynamics and animal production in sub-tropical grasslands. This can allow us to explore relationships and scenarios beyond our experimental capacity, and investigate the connectivity of the system, as well as each mechanism separately. The stage has been set, awaiting further developments.

Reliability has emerged as a first class design concern, as a result of an exponential increase in the number of transistors on the chip, and lowering of operating and threshold voltages with each new process generation. Radiation-induced transient faults are a significant source of soft errors in current and future process generations. Techniques to mitigate their effect come at a significant cost of area, power, performance, and design effort. Architectural Vulnerability Factor (AVF) modeling has been proposed to easily estimate the processor's soft error rates, and to enable the designers to make appropriate cost/reliability trade-offs early in the design cycle. Using cycle-accurate microarchitectural or logic gate-level simulations, AVF modeling captures the masking effect of program execution on the visibility of soft errors at the output. AVF modeling is used to identify structures in the processor that have the highest contribution to the overall Soft Error Rate (SER) while running typical workloads, and used to guide the design of SER mitigation mechanisms. The precise mechanisms of interaction between the workload and the microarchitecture that together determine the overall AVF is not well studied in literature, beyond qualitative analyses. Consequently, there is no known methodology for ensuring that the workload suite used for AVF modeling offers sufficient SER coverage. Additionally, owing to the lack of an intuitive model, AVF modeling is reliant on detailed microarchitectural simulations for understanding the impact of scaling processor structures, or design space exploration studies. Microarchitectural simulations are time-consuming, and do not easily provide insight into the mechanisms of interactions between the workload and the microarchitecture to determine AVF, beyond aggregate statistics. These aforementioned challenges are addressed in this dissertation by developing two methodologies. First, beginning with a systematic analysis of the factors affecting the occupancy of corruptible state in a processor, a methodology is developed that generates a synthetic workload for a given microarchitecture such that the SER is maximized. As it is impossible for every bit in the processor to simultaneously contain corruptible state, the worst-case realizable SER while running a workload is less than the sum of their circuit-level fault rates. The knowledge of the worst-case SER enables efficient design trade-offs by allowing the architect to validate the coverage of the workload suite and select an appropriate design point, and to identify structures that may potentially have high contribution to SER. The methodology induces 1.4X higher SER in the core as compared to the highest SER induced by SPEC CPU2006 and MiBench programs. Second, a first-order analytical model is proposed, which is developed from the first principles of out-of-order superscalar execution that models the AVF induced by a workload in microarchitectural structures, using inexpensive profiling. The central component of this model is a methodology to estimate the occupancy of correct-path state in various structures in the core. Owing to its construction, the model provides fundamental insight into the precise mechanism of interaction between the workload and the microarchitecture to determine AVF. The model is used to cheaply perform sizing studies for structures in the core, design space exploration, and workload characterization for AVF. The model is used to quantitatively explain results that may appear counter-intuitive from aggregate performance metrics. The Mean Absolute Error in determining AVF of a 4-wide out-of-order superscalar processor using model is less than 7% for each structure, and the Normalized Mean Square Error for determining overall SER is 9.0%, as compared to cycle-accurate microarchitectural simulation. / text

Studies in grassland management and ecology have always been challenging because of the large amount and great variation of the entities representing and affecting the system. Despite that, we were able to progress significantly in range experimentation in the Campos, in Southern Brazil. Along past thirty years, a large amount of data and information was generated, from vegetation production to components of intake. In an attempt to integrate the information available, seeking for a deeper understanding of the functioning of native grasslands, we propose adapting a mechanistic vegetation model, aggregated of a spatialized grazing component to create PampaGraze. This model was developed for temperate perennial grasslands, and was adapted and tested for subtropical, C4-dominated grasslands of the Campos of Southern Brazil (Chapter III). Despite the limited capacity of field data for validating, the model was able to relatively well simulate the trends in vegetation production along the year and seasons, while overpredicting herbage production during peak growing season. The structure of the model as it is did not allow for an accurate simulation slow-growing, tussock-forming species. Further, we developed and integrated a grazing model, based on a hybrid approach of the classical mechanistic equations of the prey model (STEPHENS & KREBS, 1986), and experimental data on foraging behaviour measured on native grasslands (Chapter IV). The model was very successful on predicting the components of intake, and responded well to variation of components in relation to changes in vegetation and to selectivity pressures, compared to available literature. Regardless of the limitations on the vegetation model, we were able to further explore the relationships of components of intake, identifying possible major limitations for herbage consumption, thus animal performance, in native grasslands. A significant progress was achieved with this thesis, but still long ways to go with this project. A list of suggestions for further developments can be found in Chapter V. We identified the emergent needs for field studies on parameters and morphogenesis, for improving predictions of the vegetation model, as well as structural points of the model that could be addressed for better representation of natural phenomena. This thesis is the first step towards a more detailed and reliable tool for studying and predicting the behaviour of vegetation dynamics and animal production in sub-tropical grasslands. This can allow us to explore relationships and scenarios beyond our experimental capacity, and investigate the connectivity of the system, as well as each mechanism separately. The stage has been set, awaiting further developments.

Studies in grassland management and ecology have always been challenging because of the large amount and great variation of the entities representing and affecting the system. Despite that, we were able to progress significantly in range experimentation in the Campos, in Southern Brazil. Along past thirty years, a large amount of data and information was generated, from vegetation production to components of intake. In an attempt to integrate the information available, seeking for a deeper understanding of the functioning of native grasslands, we propose adapting a mechanistic vegetation model, aggregated of a spatialized grazing component to create PampaGraze. This model was developed for temperate perennial grasslands, and was adapted and tested for subtropical, C4-dominated grasslands of the Campos of Southern Brazil (Chapter III). Despite the limited capacity of field data for validating, the model was able to relatively well simulate the trends in vegetation production along the year and seasons, while overpredicting herbage production during peak growing season. The structure of the model as it is did not allow for an accurate simulation slow-growing, tussock-forming species. Further, we developed and integrated a grazing model, based on a hybrid approach of the classical mechanistic equations of the prey model (STEPHENS & KREBS, 1986), and experimental data on foraging behaviour measured on native grasslands (Chapter IV). The model was very successful on predicting the components of intake, and responded well to variation of components in relation to changes in vegetation and to selectivity pressures, compared to available literature. Regardless of the limitations on the vegetation model, we were able to further explore the relationships of components of intake, identifying possible major limitations for herbage consumption, thus animal performance, in native grasslands. A significant progress was achieved with this thesis, but still long ways to go with this project. A list of suggestions for further developments can be found in Chapter V. We identified the emergent needs for field studies on parameters and morphogenesis, for improving predictions of the vegetation model, as well as structural points of the model that could be addressed for better representation of natural phenomena. This thesis is the first step towards a more detailed and reliable tool for studying and predicting the behaviour of vegetation dynamics and animal production in sub-tropical grasslands. This can allow us to explore relationships and scenarios beyond our experimental capacity, and investigate the connectivity of the system, as well as each mechanism separately. The stage has been set, awaiting further developments.

No description available.

Scientific machine learning (SciML) is a new branch of AI research at the edge of scientific computing (Sci) and machine learning (ML). It deals with efficient amalgamation of data-driven algorithms along with scientific computing to discover the dynamics of the time-evolving process. The output of such algorithms is represented in the form of a governing equation(s) (e.g., ordinary differential equation(s), ODE(s)), which one can solve then for any time point and, thus, obtain a rigorous prediction.  In this thesis, we present a methodology on how to incorporate the SciML approach in the context of clinical trials to predict IPF disease progression in the form of governing equation. Our proposed methodology also quantifies the uncertainties associated with the model by fitting 95\% high density interval (HDI) for the ODE parameters and 95\% posterior prediction interval for posterior predicted samples. We have also investigated the possibility of predicting later outcomes by using the observations collected at early phase of the study. We were successful in combining ML techniques, statistical methodologies and scientific computing tools such as bootstrap sampling, cubic spline interpolation, Bayesian inference and sparse identification of nonlinear dynamics (SINDy) to discover the dynamics behind the efficacy outcome as well as in quantifying the uncertainty of the parameters of the governing equation in the form of 95 \% HDI intervals. We compared the resulting model with the existed disease progression model described by the Weibull function. Based on the mean squared error (MSE) criterion between our ODE approximated values and population means of respective datasets, we achieved the least possible MSE of 0.133,0.089,0.213 and 0.057. After comparing these MSE values with the MSE values obtained after using Weibull function, for the third dataset and pooled dataset, our ODE model performed better in reducing error than the Weibull baseline model by 7.5\% and 8.1\%, respectively. Whereas for the first and second datasets, the Weibull model performed better in reducing errors by 1.5\% and 1.2\%, respectively. Comparing the overall performance in terms of MSE, our proposed model approximates the population means better in all the cases except for the first and second datasets, assuming the latter case's error margin is very small. Also, in terms of interpretation, our dynamical system model contains the mechanistic elements that can explain the decay/acceleration rate of the efficacy endpoint, which is missing in the Weibull model. However, our approach had a limitation in predicting final outcomes using a model derived from  24, 36, 48 weeks observations with good accuracy where as on the contrast, the Weibull model do not possess the predicting capability. However, the extrapolated trend based on 60 weeks of data was found to be close to population mean and the ODE model built on 72 weeks of data. Finally we highlight potential questions for the future work.

La libération des composés d'arôme détermine la qualité aromatique des produits alimentaires, et contribue ainsi aux choix et préférences des consommateurs. Dans ce contexte, la compréhension et la modélisation de la cinétique de libération est un défi scientifique et un enjeu de santé afin de pouvoir formuler des produits en intégrant des critères nutritionnels et sensoriels. Ce travail de thèse a permis d'étudier et de modéliser les mécanismes en bouche responsables de la dynamique de libération des stimuli olfactifs lors de consommation d'un aliment liquide ou solide chez l'homme.<br />• Dans un premier temps, un modèle mécanistique décrivant la libération des composés d'arôme au cours de la consommation d'un aliment liquide ou semi-liquide a été développé. Ces produits ont un temps de résidence en bouche très court et ne nécessitent pas de manipulation intra-orale complexe. Le modèle a été construit sur la base de bilans de matière prenant en compte des mécanismes de transfert entre certains sous-compartiments du système, ainsi que les conditions spécifiques aux différentes étapes de la consommation. Une comparaison du modèle avec des données de libération in vivo lors de la consommation de fluides newtnoniens aromatisés avec du diacétyle et de l'hexanoate d'éthyle a été effectuée. Cette étude nous a permis de comprendre le rôle du résidu post-pharyngé et de la viscosité sur la libération des composés d'arôme : • l'épaisseur du bol tapissant les muqueuses a pu être estimée à environ 15µm; • contrairement à l'hypothèse initiale, il a été mis en évidence que les propriétés pertinentes à prendre en considération pour la libération des composés d'arôme à partir d'un fluide newtonien sont celles d'un mélange de produit hautement dilué par la salive. • Dans un second temps, le modèle a été adapté à des produits nécessitant une mastication. Pour en rendre compte, de nouveaux mécanismes ont été intégrés: phénomènes de transfert de matière et de dissolution du produit dans la salive, génération d'une surface d'échange produit/fraction liquide du bol et l'ouverture vélopharyngienne lors de la mastication du produit. Le modèle a ensuite été confronté avec les données de libération du propanoate d'éthyle in vivo lors de la consommation de matrices fromagères modèles. Le modèle a pu être ajusté de façon satisfaisante à l'ensemble des données expérimentales et les deux paramètres inconnus de notre modèle (la vitesse d'incorporation moyenne de salive dans le bol au cours de la consommation et la fréquence d'ouverture du vélopharynx) ont pu être estimés. Cette étude nous a permis de comprendre le rôle de la mastication sur la libération des composés d'arôme lors de la consommation d'aliments solides. De plus, l'étude de la libération de la 2-nonanone a permis de mettre en évidence un phénomène d'adsorption sur les muqueuses pour cette molécule. • Enfin, il ressort de la comparaison des deux modèles que les paramèters clés gouvernant la libération des composés d'arôme ne sont pas les mêmes selon la catégorie de produit (liquide ou solide) considérée: • lors de la consommation d'aliments liquides ou semi-liquides, le coefficient de transfert de matière dans le bol, la fréquence respiratoire de l'individu et l'épaisseur du résidu post-pharyngé sont les trois facteurs clés gouvernant la libération des composés d'arôme; • en revanche, lors de la consommation de produits solides mastiqués, ce sont la vitesse d'incorporation moyenne de salive dans le bol, la fréquence d'ouverture du vélopharynx et la durée de mastication qui sont les trois paramètres ayant un effet majeur sur les cinétiques de libération. La démarche de modélisation nous a permis de mieux comprendre les parts relatives du produit, de l'individu, et de l'interaction produit-individu sur la libération des composés d'arôme au cours de la consommation d'un aliment. / Delivery of aroma compounds to olfactory receptors determines the aromatic quality of food products and contributes to consumer choices and preferences. Therefore, understanding and modelling the release kinetic is a scientific challenge and a health issue in order to formulate products of both high nutritional and sensory quality. This thesis studied in-mouth mechanisms responsible of the dynamics of olfactory stimuli release during food consumption. • First, a mechanistic model describing the aroma compounds release during consumption of a liquid or semi-liquid food has been developed. These products have a very short in-mouth residence time and do not require complex intra-oral manipulation. The model takes into account mass balances, transfer mechanisms occurring between some sub-compartments of the system, and the specific conditions at the different stages of consumption. A comparison of the model predictions with in vivo release data during the consumption of Newtnonien fluids flavored with diacetyl and ethyl hexanoate was performed. This study highlighted the role of post-pharyngeal residue and viscosity on the aroma compounds release: • the thickness of bolus covering the mucous membranes has been estimated at about 15μm; • it was found that the relevant properties to be considered for the release of aroma compounds from a Newtonian fluid are those of a mixture highly diluted by saliva. • Second, the model previously developed was adapted for products requiring chewing. It takes into account the phenomena of mass transfer and dissolution of the product in the saliva during chewing. The generation of a product/liquid contact surface as well as the velopharyngial opening that occurs during the mastication of the product were also integrated into the model. The model was then confronted with in vivo release data for ethyl propanoate during consumption of four cheese matrices. All simulations have been satisfactorily fitted to experimental data and the two unknown parameters of our model (the average rate of saliva incorporation into the bolus and the frequency of velopharyngial opening) could be estimated. This study has enabled us to understand the role of mastication on the release of aroma compounds during consumption of solid food: • the opening of velopharynx during intra-oral manipulation of the product produces a continuous supply of aroma compounds in the nose; • the residence time of solid product in the mouth are much longer than for the consumption of liquid and semi-liquid foods, allowing the secretion of significant volumes of saliva. In addition, the study of the release of 2-nonanone highlighted an adsorption phenomenon on the mucous membranes for this molecule. • Finally, sensitivity analysis of the two release models indicates that: • when eating a liquid or semi-liquid food, the mass transfer coefficient in the bolus, the breath rate and the thickness of post-pharyngeal residue are the three key factors governing the release of aroma compounds; • however, when eating a solid food product, it is the average rate of saliva incorporation into the bolus during consumption, the frequency and duration of velopharyngeal opening, and the mastication time which are the three parameters that have major effects on the kinetics of release. The modeling approach allowed us to better understand the relative effects of the product, the individual, and individual-product interaction on the release of aroma compounds during food consumption. The results of this work indicated that the most important parameters depend on the category of product (liquid or solid) under consideration.