Ces travaux portent sur la modélisation et la simulation numérique de la dispersion de matériaux nocifs (pulvérisations liquides ou gazeuses) en milieu urbain ou naturel (attentat ou explosion accidentelle survenant en zone peuplée, fuites de produits toxiques gazeux ou liquides, éclatement de réservoir..). Afin de prédire ces risques un outils de simulation tridimensionnel a été développé. Celui-ci est basé sur un modèle de milieu hétérogène afin de traiter des phénomènes dont la durée et les distances associées peuvent être très grandes. La topographie des milieux étudiées est prise en compte grâce à des données numériques d'´elévation ainsi que les conditions météo permettant l'utilisation de profils de température et de vent complexes. Les transferts de chaleur et de masse sont considérés, notamment au niveau des obstacles. Un schéma numérique d'ordre élevé en temps et en espace est utilisé pour calculer les concentrations massiques de polluants. Par ailleurs, un modèle d'écoulement gaz-particule a été développé et implémenté dans le code de calcul. L'instabilité d'une couche de fluide soumise à un important gradient de pression est également étudiée, ceci afin de mieux comprendre et de caractériser les conditions initiales à utiliser pour ce type d'écoulement, impliquant des couches de particules. / This work deals with the modeling and the numerical simulation of the dispersion of toxic cloud of dropplets or gas in uneven geometry such as urban environment, industrial plants and hilly environment. Examples of phenomena under study are the dispersion of chemical products from damaged vessels, gas diffusion in an urban environment under explosion conditions, shock wave propagation in urban environment etc. A 3D simulation code has been developed in this aim. To simplify the consideration of complex geometries, a heterogeneous discrete formulation has been developed. When dealing with large scale domains, such as hilly natural environment, the topography is reconstructed with the help of numerical elevation data. Meteorological conditions are also considered, concerning temperature and wind velocity profiles. Heat and mass transfers on subscale objects, such as buildings are studied. A high order numerical scheme in space and time is used to compute mass concentration of pollutant. A two-phase model for dilute gas-particles flow has been developed and implemented in the 3D simulation code. The instability of a fluid layer appearing under high pressure gradient is also studied. This analysis allows us a better understanding of initial conditions for similar problems involving particles layer.

Dans les analyses d'interaction sol-structure (ISS), la pratique commune en génie civil est de considérer un mouvement uniforme du champ libre à tous les points situés à la surface du sol. Néanmoins, cette considération n'est pas tout à fait réaliste parce que les signaux sismiques sont spatialement différents grâce à l'effet de passage d'ondes, à l'effet de site et aussi aux dispersions et réflexions des ondes qui propagent dans des milieux hétérogènes aléatoires ("incohérence pure"). Ainsi, pour répondre aux problèmes de sécurité des bâtiments et équipements, il est important de faire une analyse d'interaction sol-structure dans la manière plus réaliste. Cela peut être acquis par prendre en compte la variabilité spatiale du champ sismique dans les études d'ISS. Un grand nombre d'études dans la littérature montrent que la prise en compte de la variabilité spatiale du champ sismique dans les études d'ISS peut avoir des effets importants sur la réponse de structures. L'incohérence spatiale du champ sismique due aux dispersions et réflexions des ondes (incohérence pure) peut généralement être modélisée pour ce genre d'études dans le cadre probabiliste par une fonction de cohérence. Le but principal des études réalisées dans cette thèse de doctorat est de construire une description stochastique de la variabilité spatiale du champ sismique par un modèle de cohérence. Ce modèle devrait avoir une relation avec les propriétés physiques et statistiques de milieux considérés. En s'appuyant sur les analyses théoriques de la propagation des ondes sismiques dans des milieux hétérogènes aléatoires, les analyses des données expérimentales obtenues par des enregistrements sur des sites sismiques, ainsi que sur les modélisations numériques de propagation des ondes sismiques dans des milieux hétérogènes aléatoires, un modèle de cohérence est validé dans le cadre des études de cette thèse de doctorat pour représenter la variabilité spatiale du champ sismique dans les études d'interaction sol-structure. L'influence de la variabilité spatiale du champ sismique sur la réponse de structure est également analysée. / In seismic soil-structure interaction studies (SSI), the common practice in Civil Engineering is to consider a uniform movement of free field at any point on the ground surface. However, that assumption is not completely realistic since the seismic ground motions can vary spatially due to wave passage effects, dispersions and reflections of wave propagating in the random heterogeneous media "pure incoherence" and site effects. Therefore, in order to increase the security of buildings and equipment, it is important to do an analysis of seismic soil-structure interactions in the most realistic way. This can be achieved by taking into account the spatial variability of seismic ground motions. Several studies in the literature show that taking into account the spatial variability of seismic ground motions in SSI analyses can have remarkable effects on the structural responses. The spatial incoherence of seismic ground motions due to dispersions and reflections of wave "pure incoherence" can generally be modelled in such analysis by a "coherency function" in frequency domain. The principal goal of this Ph.D thesis is to construct a stochastic description of spatial variability of seismic ground motions by means of coherency functions. Accurately, it aims to propose a parametrical coherency model of spatial variability of seismic ground motions. This later should be related to some physical and statistical properties of the soil at the application sites so that it can be applied in any types of sites. Based on theoretical considerations on coherency of seismic wave propagation in random heterogeneous media, on experimental data analyses, and on numerical modelling of seismic wave propagation in random heterogeneous media, a coherency model is validated and proposed for the analyses of soil-structure interactions. The influence of spatial variability of seismic ground motions on the structural responses are also pointed out by using the validated coherency model.

Arbitrary high order numerical methods for time-harmonic acoustic scattering problems originally defined on unbounded domains are constructed. This is done by coupling recently developed high order local absorbing boundary conditions (ABCs) with finite difference methods for the Helmholtz equation. These ABCs are based on exact representations of the outgoing waves by means of farfield expansions. The finite difference methods, which are constructed from a deferred-correction (DC) technique, approximate the Helmholtz equation and the ABCs to any desired order. As a result, high order numerical methods with an overall order of convergence equal to the order of the DC schemes are obtained. A detailed construction of these DC finite difference schemes is presented. Details and results from an extension to heterogeneous media are also included. Additionally, a rigorous proof of the consistency of the DC schemes with the Helmholtz equation and the ABCs in polar coordinates is also given. The results of several numerical experiments corroborate the high order convergence of the proposed method. A novel local high order ABC for elastic waves based on farfield expansions is constructed and preliminary results applying it to elastic scattering problems are presented.

京都大学 / 0048 / 新制・課程博士 / 博士(医学) / 甲第20248号 / 医博第4207号 / 新制||医||1020(附属図書館) / 京都大学大学院医学研究科医学専攻 / (主査)教授 鈴木 実, 教授 別所 和久, 教授 大森 孝一 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

Les voies ferrées sont endommagées par les chargements dynamiques répétés issus du passage des trains, en particulier pour les trains à grandes vitesses. Structures multicouches complexes, ces voies sont constituées : de rails en acier, de semelles en élastomère, des traverses de béton, d’une couche de ballast et d’une sous-couche. L’étude du comportement mécanique d’une voie ferrée (de chaque composant à la structure entière) est donc étroitement liée à l’amélioration de la sécurité ferroviaire, ainsi qu’à l’efficacité de ce mode de transports.Ce travail de thèse se focalise sur l’étude des semelles sous rail et de la sous-couche en grave bitumineuse. Ces deux composants ont été choisis pour leurs similarités en termes de comportement mécanique non linéaire et capacité de dissipation d’énergie. Ce manuscrit est divisé en trois chapitres.Dans le cadre de ce premier chapitre le modèle d’intégrale par convolution (modèle-CI) est choisi pour modéliser le comportement mécanique du matériau élastomère de la semelle. Le modèle-CI est une extension naturelle de la théorie de la viscoélasticité linéaire, car basé sur l'extension du principe de superposition Boltzmann ; la séparation des contraintes proposée par ce modèle, a été observée expérimentalement par de nombreux auteurs.Le deuxième chapitre concerne l’étude du comportement mécanique du matériau élastomérique qui compose les semelles sous-rail et de modélisation pour prédire le comportement non-linéaire et la capacité d'absorption d'énergie d'une structure semelle.Le modèle de comportement (modèle-CI) permet de représenter de façon très fiable la partie chargement de la semelle (erreur de 1 % pour la rigidité). Pour la partie déchargement, la représentation est un peu moins bonne : la déformation résiduelle "numérique" est de 2,2 % alors qu'expérimentalement elle n'est que de 0,4 %, ce qui conduit à une erreur de prédiction sur l'énergie dissipée de 37.5 %. La comparaison entre les résultats numériques et expérimentaux in-situ montrent que le modèle utilisé permet de décrire assez correctement la réponse de la semelle au passage d'un train dans les limites d’erreur de prédiction de la déformation résiduelle.Ce modèle-CI doit être utilisé sur le modèle géométrique 3D complet de la semelle, les approches simplifiées (2D, semelle sans rainure) conduisent à des prédictions fortement erronées.Le troisième chapitre se focalise sur l’étude du matériau de type asphalte utilisé pour la couche sous-ballast des voies ferrées. Des cubes de “Matériaux Virtuels” ont été réalisés en disposant aléatoirement des inclusions sphériques monodisperses rigides dans un volume de matrice au comportement hyper-élastique. L’influence du diamètre et de la fraction volumique de ces inclusions sur le comportement mécanique d’une structure a été étudiée numériquement et expérimentalement en utilisant un plan d’expérience de type Doehlert. Cette approche de « Matériaux Virtuels » a permis d’avoir une correspondance exacte entre les géométries des spécimens numériques et expérimentaux sur les 7 échantillons testés.L’analyse des surfaces de réponses a montré que les deux paramètres observés F_max et E_% sont fortement corrélés aux valeurs de V_fr. L’influence du diamètre des inclusions, par contre, est très faible.Enfin, les simulations par éléments finis ont permis d’étudier la répartition interne des contraintes et déformations. Les résultats ont été présentés pour l’échantillon V0225-D08 : la chaine d’effort a été visualisée à l’intérieur de la matrice et présente des contraintes de Von Mises jusqu’à 8 fois celles obtenues dans la matrice.Dans l’étude proposée, le diamètre et la forme des inclusions ont été fixés. Il serait intéressant de faire varier ces paramètres en utilisant la même méthodologie. Par ailleurs, les récentes avancées en termes de fabrication additive permettent d’imaginer la construction d’échantillons hétérogènes complexes. / Repetitive dynamic loads caused by passing trains can damage a railway track, especially at high speeds. The complex multilayer structure of the modern track consists of: stainless steel rails, elastomeric rail pads, concrete sleepers, track ballast and sub-ballast layers. Investigation of the mechanical behaviour of the railway track structure (as the whole and by parts) can have a great importance for the improvement of safety and efficiency of railway transportation.In the present study rail pad and bituminous mixture (BM) sub-ballast layers of a standard ballasted railway track were considered for investigation. These parts of the track were chosen for their similarities in the mechanical behaviour (nonlinearity and energy dissipation) and function (reduction of the dynamic part of load, an influence on the load distribution).The first chapter reviews the main aspects of the mechanical behaviour of elastomeric materials and covers the common theoretical approaches, appropriate for the modelling of this behaviour. The Convolution Integral approach (CI-model) was chosen to represent the mechanical behaviour of a rail pad material as a natural extension of theory of linear viscoelasticity, based on extension of the well-established Boltzmann Superposition Principle.The second chapter is devoted to study of elastomeric material of a rail pad and to numerical modelling of a whole elastomeric rail pad structure subjected to common track loads. Special attention was given to possibility of the model to describe the nonlinearity of the mechanical behaviour and capability of energy dissipation.Sufficient conformity between experimental and numerical results was established on loading part of a Force vs Displacement curve (an error of 1 % was obtained for the stiffness value) for the quasi-static loading, while prediction of the residual compression displacement remains poor, especially in the first loading cycle (2.2% of the macroscopic strain against 0.4% in experiment). The observed discrepancy led to poor prediction of the dissipated energy (an error of 37.5 % was found). Comparison between results of the numerical simulation and in situ experimental measurements has shown that the FE model is capable to describe dynamic behaviour of a rail pad structure to within the error of prediction of the residual compression displacement.Possible ways to simplify the numerical model, discussed in the second chapter, generally lead to high overestimation (2D plain strain and 3D grooveless models) or underestimation (2D plain stress model) of the rail pad mechanical behaviour.The third chapter of the thesis is connected to the study of a BM material, used on a railway track as a sub-ballast layer. Influence of size and volume fraction of monodisperse spherical inclusions, randomly packed into a cubic matrix, on the mechanical behaviour of obtained composite structure were investigated using “Virtual Material” approach. This approach allows numerical study of a theoretical case without losing connection with a real experiment (by means of direct geometrical correspondence). Parameters of 7 specimens were chosen in accordance with Doehlert experimental design.Analysis of “response surfaces” has shown that both F_max and E_% have a strong dependence on the value of V_fr and almost no dependence on the value of D.Stress/strain concentrations were analyzed using FE method on example of V0225-D08 specimen. This allows to find and to visualize load-bearing chains going through the matrix. Von Mises stress in load-bearing chains is almost 8 times higher than the average in the matrix.More complex models (real and numerical) in terms of problem discretization (more than one inclusions’ fraction, different inclusions’ shapes, etc.) can be developed and studied in the similar way. Moreover, the recent progress in additive manufacturing technologies shows potential to create complex heterogeneous specimens with an increased precision.

In this work, an analytical approach is used along with nodal schemes for the solution of xed source two-dimensional neutron transport problems, in Cartesian geometry, de ned in heterogeneous medium, with anisotropic scattering. The methodology is developed from the discrete ordinates version of the two-dimensional transport equation along with the level symmetric angular quadrature set. One-dimensional equations for the averaged angular uxes are obtained by transverse integration of the original problem. Such equations are solved by the ADO method. Explicit expressions in spatial variables are derived for averaged uxes in each region in which the domain is subdivided. The solution in each region is coupled with that of its neighbouring regions to provide the solution in the whole domain, without resorting to using iterative methods. As usual in nodal schemes, auxiliary equations are needed. Here two di erent treatments were given to this issue: one based on relations between the unknown ows in the contours of the regions and the average angular uxes, and another in which these ows are approximated by polynomials of order zero being in this case, incorporated into the source term. Numerical results were compared with available literature showing the solution preserve the computational e ciency which has been a good feature of the ADO method when applied to different problems. / Neste trabalho, uma abordagem analítica é utilizada juntamente com esquemas nodais na resolução de problemas bidimensionais de transporte de nêutrons de fonte fixa, em geometria cartesiana, definidos em meio heterogêneo, com espalhamento anisotrópico. A metodologia proposta é desenvolvida a partir da versão em ordenadas discretas da equação de transporte bidimensional, juntamente com o esquema de quadratura simétrica de nível. As equações em ordenadas discretas são integradas transversalmente, originando equações unidimensionais para os fluxos angulares médios. Tais equações unidimensionais são resolvidas pelo método ADO (Analytical Discrete Ordinates). Expressões explícitas nas variáveis espaciais são derivadas para os fluxos angulares médios em cada região em que o domínio foi subdividido. A solução em cada região é acoplada às regiões vizinhas, para fornecer a solução no domínio todo, sem a utilização de métodos iterativos. Como usual em esquemas nodais, equações auxiliares são necessárias, recebendo neste estudo dois tratamentos distintos: um em que os fluxos desconhecidos nos contornos das regiões assumem relações de proporcionalidade, com os fluxos angulares médios; e, outro, em que esses fluxos são aproximados por polinômios de ordem zero sendo, nesse caso, incorporados ao termo fonte. Resultados numéricos obtidos e comparados com disponíveis na literatura mostram a viabilidade da formulação, mantendo a eficiência computacional já verificada no tratamento de outros problemas, com o uso do método ADO.

In this work, an analytical approach is used along with nodal schemes for the solution of xed source two-dimensional neutron transport problems, in Cartesian geometry, de ned in heterogeneous medium, with anisotropic scattering. The methodology is developed from the discrete ordinates version of the two-dimensional transport equation along with the level symmetric angular quadrature set. One-dimensional equations for the averaged angular uxes are obtained by transverse integration of the original problem. Such equations are solved by the ADO method. Explicit expressions in spatial variables are derived for averaged uxes in each region in which the domain is subdivided. The solution in each region is coupled with that of its neighbouring regions to provide the solution in the whole domain, without resorting to using iterative methods. As usual in nodal schemes, auxiliary equations are needed. Here two di erent treatments were given to this issue: one based on relations between the unknown ows in the contours of the regions and the average angular uxes, and another in which these ows are approximated by polynomials of order zero being in this case, incorporated into the source term. Numerical results were compared with available literature showing the solution preserve the computational e ciency which has been a good feature of the ADO method when applied to different problems. / Neste trabalho, uma abordagem analítica é utilizada juntamente com esquemas nodais na resolução de problemas bidimensionais de transporte de nêutrons de fonte fixa, em geometria cartesiana, definidos em meio heterogêneo, com espalhamento anisotrópico. A metodologia proposta é desenvolvida a partir da versão em ordenadas discretas da equação de transporte bidimensional, juntamente com o esquema de quadratura simétrica de nível. As equações em ordenadas discretas são integradas transversalmente, originando equações unidimensionais para os fluxos angulares médios. Tais equações unidimensionais são resolvidas pelo método ADO (Analytical Discrete Ordinates). Expressões explícitas nas variáveis espaciais são derivadas para os fluxos angulares médios em cada região em que o domínio foi subdividido. A solução em cada região é acoplada às regiões vizinhas, para fornecer a solução no domínio todo, sem a utilização de métodos iterativos. Como usual em esquemas nodais, equações auxiliares são necessárias, recebendo neste estudo dois tratamentos distintos: um em que os fluxos desconhecidos nos contornos das regiões assumem relações de proporcionalidade, com os fluxos angulares médios; e, outro, em que esses fluxos são aproximados por polinômios de ordem zero sendo, nesse caso, incorporados ao termo fonte. Resultados numéricos obtidos e comparados com disponíveis na literatura mostram a viabilidade da formulação, mantendo a eficiência computacional já verificada no tratamento de outros problemas, com o uso do método ADO.

In this work, an analytical approach is used along with nodal schemes for the solution of xed source two-dimensional neutron transport problems, in Cartesian geometry, de ned in heterogeneous medium, with anisotropic scattering. The methodology is developed from the discrete ordinates version of the two-dimensional transport equation along with the level symmetric angular quadrature set. One-dimensional equations for the averaged angular uxes are obtained by transverse integration of the original problem. Such equations are solved by the ADO method. Explicit expressions in spatial variables are derived for averaged uxes in each region in which the domain is subdivided. The solution in each region is coupled with that of its neighbouring regions to provide the solution in the whole domain, without resorting to using iterative methods. As usual in nodal schemes, auxiliary equations are needed. Here two di erent treatments were given to this issue: one based on relations between the unknown ows in the contours of the regions and the average angular uxes, and another in which these ows are approximated by polynomials of order zero being in this case, incorporated into the source term. Numerical results were compared with available literature showing the solution preserve the computational e ciency which has been a good feature of the ADO method when applied to different problems. / Neste trabalho, uma abordagem analítica é utilizada juntamente com esquemas nodais na resolução de problemas bidimensionais de transporte de nêutrons de fonte fixa, em geometria cartesiana, definidos em meio heterogêneo, com espalhamento anisotrópico. A metodologia proposta é desenvolvida a partir da versão em ordenadas discretas da equação de transporte bidimensional, juntamente com o esquema de quadratura simétrica de nível. As equações em ordenadas discretas são integradas transversalmente, originando equações unidimensionais para os fluxos angulares médios. Tais equações unidimensionais são resolvidas pelo método ADO (Analytical Discrete Ordinates). Expressões explícitas nas variáveis espaciais são derivadas para os fluxos angulares médios em cada região em que o domínio foi subdividido. A solução em cada região é acoplada às regiões vizinhas, para fornecer a solução no domínio todo, sem a utilização de métodos iterativos. Como usual em esquemas nodais, equações auxiliares são necessárias, recebendo neste estudo dois tratamentos distintos: um em que os fluxos desconhecidos nos contornos das regiões assumem relações de proporcionalidade, com os fluxos angulares médios; e, outro, em que esses fluxos são aproximados por polinômios de ordem zero sendo, nesse caso, incorporados ao termo fonte. Resultados numéricos obtidos e comparados com disponíveis na literatura mostram a viabilidade da formulação, mantendo a eficiência computacional já verificada no tratamento de outros problemas, com o uso do método ADO.

Submitted by Geandra Rodrigues (geandrar@gmail.com) on 2018-10-24T12:40:05Z No. of bitstreams: 1 filipefernandesdepaula.pdf: 19865574 bytes, checksum: dfebb62a2a39cd7c70ab5c775d5441ce (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2018-11-23T12:10:30Z (GMT) No. of bitstreams: 1 filipefernandesdepaula.pdf: 19865574 bytes, checksum: dfebb62a2a39cd7c70ab5c775d5441ce (MD5) / Made available in DSpace on 2018-11-23T12:10:30Z (GMT). No. of bitstreams: 1 filipefernandesdepaula.pdf: 19865574 bytes, checksum: dfebb62a2a39cd7c70ab5c775d5441ce (MD5) Previous issue date: 2018-09-12 / O desenvolvimento de técnicas adequadas para extração eficiente de óleo de reservatórios de petróleo passa pela simulação precisa de tais fenômenos, que é alcançada através do estudo de modelos matemáticos e métodos computacionais robustos, eficientes e precisos. Neste contexto, este trabalho visa o estudo de métodos numéricos para a simulação de escoamentos bifásicos em meios porosos heterogêneos. Para tanto, propomos uma abordagem numérica do tipo staggered para estes modelos, que se baseia na aproximação de forma desacoplada dos sistemas de equações diferenciais parciais referentes aos problemas de Darcy e da saturação das fases. Dessa forma, podem ser empregados métodos numéricos específicos para cada sistema, que melhor se adequem às suas caractrerísticas. Assim, propomos o estudo de métodos de elementos finitos mistos, estáveis e estabilizados, clássicos e híbridos e localmente conservativos para o cômputo da velocidade da mistura e de um método de volumes finitos não-oscilatório de alta ordem, baseado em esquemas centrais, para a equação hiperbólica não-linear que governa o transporte da saturação das fases. Resultados numéricos comprovam a flexibilidade, a taxa de convergência e o custo computacional dos métodos adotados, além de demonstrar a eficácia dos métodos quando aplicados a simulação de problemas associados a extração de petróleo em cenários fortemente heterogêneos. / The development of techniques for efficient oil extraction from reservoirs passes through the simulation of such phenomena, which is achieved by the study of mathematical models and robust, precise and efficient computational methods. This dissertation studies methods for the simulation of two-phase flows in heterogeneous porous media. To do so, we propose a “staggered” numerical approach for the numerical methods, that is based on the approximation of uncoupled systems of differential equations related to Darcy’s problems and saturation of the phases. Then, appropriate methods for each system, that best suit its characteristics can be applied. Therefore, we propose studying locally conservative finite element methods, stable and stabilized, classical and hybrid to approximate the velocity field and a non-oscillatory high order finite volume method, based on central schemes, to approximate the non-linear hyperbolic equation that governs the transport of phases. Numerical results attest to the flexibility, convergence rate and computational cost of the adopted methods, and demonstrate the effectiveness of such methods when applied to oil extraction in various heterogeneous porous media scenarios.

Les mécanismes responsables du transport hydrodynamique anormal (non-Fickéen) peuvent être rattachés à la complexité de la géométrie du milieu à l'échelle des pores. Dans cette thèse, nous étudions la dynamique des vitesses de particules à l'échelle des pores. À l'aide de simulations de suivi de particules effectuées sur un échantillon numérisé de grès de Berea, nous présentons une analyse détaillée de l'évolution Lagrangienne et Eulérienne et de leur dépendance aux conditions initiales. Le long de leur ligne de courant, la vitesse des particules montre un signal intermittent complexe, alors que leur sériede vitesses spatiales présente des fluctuations régulières. La distribution spatiale des vitesses des particules converge rapidementvers l'état stationnaire. Ces résultats dénotent un processus Markovienqui permet de prédire les fluctuations de vitesse dans le réseau poral.Ces processus, associés à la tortuosité et à la distance de corrélation de vitesse permettent de paramétrer un modèle de marche aléatoire dans le temps (CTRW) et de réaliser le changement d’échelle pour simuler le transport à l’échelle de Darcy. Le modèle, comme tout modèle issu d’un changement d'échelle, repose sur la définition d'un volume élémentaire représentatif (VER). Nous montrons qu’un VER basé sur les statistiques de vitesse permet de définir un support pertinent pour la modélisation du transport hydrodynamique pré-asymptotique à asymptotique, et ainsi d’éviter les limitations associées à l’équation d’advection-dispersionFickéenne. Cette approche est utilisée pour étudier l’impact de l’hétérogénéité du réseau poral sur le volume de mélange et la masse du produit d’une réaction bimoléculaire. / The mechanisms responsible for anomalous (non-Fickian) hydrodynamictransport can be traced back to the complexity of the medium geometry atthe pore-scale. In this thesis, we investigate the dynamics of pore-scaleparticle velocities. Using particle tracking simulations performed on adigitized Berea sandstone sample, we present a detailed analysis of theevolution of the Lagrangian and Eulerian evolution and their dependenceon the initial conditions. The particles experience a complexintermittent temporal velocity signal along their streamline while theirspatial velocity series exhibit regular fluctuations. The spatialvelocity distribution of the particles converges quickly to thesteady-state. These results lead naturally to Markov processes for theprediction of these velocity series.These processes, together with the tortuosity and the velocitycorrelation distance that are properties of the medium, allow theparameterization of a continuous time random walk (CTRW) for theupscaling of the transport. The model, like any upscaled model, relieson the definition of a representative elementary volume (REV). We showthat an REV based on the velocity statistics allows defining a pertinentsupport for modeling pre-asymptotic to asymptotic hydrodynamictransport at Darcy scale using, for instance, CTRW, thus overcomingthe limitations associated with the Fickian advection dispersionequation. Finally, we investigate the impact of pore-scale heterogeneityon a bimolecular reaction and explore a methodology for the predictionof the mixing volume and the chemical mass produced.