É uma verdade incontestável que a fadiga constitui um dos problemas mais críticos em engenharia, especialmente em estruturas formadas por materiais dúcteis. Por essa razão, uma grande quantidade de métodos e estudos tem sido desenvolvida para tratar deste problema. No entanto, no caso de materiais quase frágeis como concreto, rochas cerâmicas e alguns tipos de materiais compostos, o efeito que cargas oscilantes produzem sobre estes materiais é menos conhecido e aparentemente também menos crítico. No presente trabalho, se utiliza uma versão do método dos elementos discretos formado por barras para explorar as possibilidades do mesmo na simulação do efeito de fadiga em materiais quase frágeis. Simulações sobre corpos de prova simples são apresentadas e vários aspectos deste estudo são discutidos, entre eles: influência da escala, influência da aleatoriedade nas propriedades do material simulado e se a lei de crescimento prevista por Paris (1961) se apresenta nas simulações realizadas. Finalmente, nas considerações finais, são salientadas as possibilidades que se abrem ao aplicar o modelo de elementos discretos apresentado no estudo de materiais quase frágeis submetidos à ação de cargas oscilantes. / It is an unquestionable truth that fatigue consists in one of the most critical problems of engineering, especially in ductile material structures. For that reason, a great amount of methods and studies has been developed to deal with this matter. However, when it comes to quasi brittle materials like concrete, ceramic stones and a few kinds of composites, the effect of cyclic loading on these materials is less well known, and apparently also less critical. In this work, a version of the discrete elements method formed by bars is applied to explore the possibilities of its use on simulating the effect of fatigue over quasi brittle materials. Simulations are presented over simple test specimens and several features of this study are discussed, among them: the influence of specimen scale, the influence of random distribution on material properties and if crack growth laws previewed by Paris, (Paris et al., 1961) are verified in the performed simulations. In the final considerations, the possibilities brought by applying the discrete elements method in this study of quasi brittle materials submitted to cyclic loading are highlighted.

É uma verdade incontestável que a fadiga constitui um dos problemas mais críticos em engenharia, especialmente em estruturas formadas por materiais dúcteis. Por essa razão, uma grande quantidade de métodos e estudos tem sido desenvolvida para tratar deste problema. No entanto, no caso de materiais quase frágeis como concreto, rochas cerâmicas e alguns tipos de materiais compostos, o efeito que cargas oscilantes produzem sobre estes materiais é menos conhecido e aparentemente também menos crítico. No presente trabalho, se utiliza uma versão do método dos elementos discretos formado por barras para explorar as possibilidades do mesmo na simulação do efeito de fadiga em materiais quase frágeis. Simulações sobre corpos de prova simples são apresentadas e vários aspectos deste estudo são discutidos, entre eles: influência da escala, influência da aleatoriedade nas propriedades do material simulado e se a lei de crescimento prevista por Paris (1961) se apresenta nas simulações realizadas. Finalmente, nas considerações finais, são salientadas as possibilidades que se abrem ao aplicar o modelo de elementos discretos apresentado no estudo de materiais quase frágeis submetidos à ação de cargas oscilantes. / It is an unquestionable truth that fatigue consists in one of the most critical problems of engineering, especially in ductile material structures. For that reason, a great amount of methods and studies has been developed to deal with this matter. However, when it comes to quasi brittle materials like concrete, ceramic stones and a few kinds of composites, the effect of cyclic loading on these materials is less well known, and apparently also less critical. In this work, a version of the discrete elements method formed by bars is applied to explore the possibilities of its use on simulating the effect of fatigue over quasi brittle materials. Simulations are presented over simple test specimens and several features of this study are discussed, among them: the influence of specimen scale, the influence of random distribution on material properties and if crack growth laws previewed by Paris, (Paris et al., 1961) are verified in the performed simulations. In the final considerations, the possibilities brought by applying the discrete elements method in this study of quasi brittle materials submitted to cyclic loading are highlighted.

[pt] Considerando o investimento do governo federal na expansão da malha ferroviária com o programa de aceleração do crescimento (PAC2) e com o programa de investimento em logística (PIL), vem sendo desenvolvidas no país inúmeras pesquisas envolvendo ferrovias, inclusive na área de materiais, como o lastro ferroviário, objeto de estudo deste projeto. Paralelamente, pesquisas com foco em modelagem computacional vem sendo desenvolvidas a fim de otimizar a avaliação do comportamento dos materiais. O projeto tem como objetivo desenvolver modelos computacionais para avaliar a capacidade do método dos elementos discretos em reproduzir o comportamento mecânico de lastro ferroviário submetido a carregamentos cíclicos, com base em um ensaio triaxial cíclico. O ensaio analisou o comportamento mecânico de duas distribuições granulométricas de lastro ferroviário, sendo a primeira sugerida por Indraratna et al. (2004), referente a alterações para melhorias da normativa australiana, enquanto a segunda, utiliza a normativa brasileira. O modelo computacional foi desenvolvido no software WooDEM (2012). Foram realizadas análises mecânicas do comportamento do lastro ferroviário considerando as coordenadas das partículas em relação ao tempo e a profundidade, a fim de determinar o deslocamento gerado pelo carregamento cíclico. / [en] Considering the federal government s investment in the expansion of the railway network with the growth acceleration program (PAC2) and with the investment in logistics program (PIL), numerous surveys involving railways have been developed in the country, including in the area of materials, such as the rail ballast, object of study of this project. In parallel, research focused on computational modeling has been developed in order to optimize the evaluation of the behavior of materials. This dissertation aims to develop computational models to evaluate the ability of the discrete element method to reproduce the mechanical behavior of cyclically loaded rail ballast, based on a cyclic triaxial test. The study analyzed the mechanical behavior of two rail ballast granulometric distributions, the first being suggested by Indraratna et al. (2004), regarding changes to improve Australian regulations, while the second, uses the Brazilian legislation. The computational model was developed in WooDEM software (2012). Mechanical analysis of the rail ballast behavior was performed considering the coordinates of the particles in relation to time and depth in order to determine the displacement generated by the cyclic loading.

Les régions de montagne sont parfois le siège de phénomènes dévastateurs brutaux, tels que les éboulements, les glissements de terrains, ou encore les avalanches. Fondamentalement, ces phénomènes sont issus de la rupture d’un volume de matériau qui va alors s’écouler sous l’effet de son poids propre. Prévoir ces phénomènes impose donc de mieux décrire la phase initiale de rupture. Ce travail de thèse s’inscrit dans cette volonté, en considérant le cas très général de la rupture dans un milieu granulaire. Pour cela, des simulations numériques utilisant une méthode aux éléments discrets ont été mises en œuvre, afin de vérifier la pertinence des critères usuels de rupture et d’instabilité. Dans un second temps, des analyses plus fines basées sur le travail du second ordre calculé à partir des variables microscopiques ont été menées afin de cerner comment évolue la microstructure de l’assemblage granulaire juste avant et pendant la rupture. Finalement, l’influence du chemin de chargement et des paramètres de contrôle sur la nature (effective ou non effective) et le mode (diffus ou localisé) de rupture a été examinée à travers une série d’essais biaxiaux sur deux échantillons numériques bidimensionnels (dense et moyennement dense). / Landslides and avalanches are the most common brutal and devastating natural hazards in mountain regions. Basically, these phenomena are derived from the failure of a material volume which will then flow under the effect of its own weight. Therefore, predicting these phenomena requires a better description of the initial phase of failure. This work comes within this willingness, considering the very general case of failure in granular media. For this purpose, numerical simulations using a discrete element method were carried out in order to investigate the relevance of the common failure and instability criteria. In a second step, more detailed analyses based on the second order work computed from the microscopic variables were conducted in order to describe the evolution of the microstructure of a granular assembly just before and during the failure. Finally, the influence of the loading path and the control parameters on the nature (effective or non-effective) and the mode (diffuse or localized) of failure was discussed through a series of biaxial tests performed on two-dimensional numerical samples depicting respectively a dense and a medium dense dry sand.

[pt] Este trabalho tem como finalidade comparar e analisar diferentes modelos numéricos na análise do processo de queda de blocos com diferentes geometrias, para que seja possível determinar os riscos à infraestrutura e à população local que este evento pode causar. Foram utilizados o software RAMMS e o programa de modelagem PFC3D para simulação de queda de blocos. O primeiro foi desenvolvido pelo instituto suíço WSL, o software utiliza a dinâmica não suave (nonsmooth), não havendo penetração dos blocos com a superfície, e as leis de contato são feitas por restrições. O segundo utiliza o método dos elementos discretos através de dinâmica de contato suave. Na primeira parte do trabalho foram realizadas simulações de casos simples de queda de blocos para ambos os modelos, a fim de analisar o principal parâmetro do contato rocha-solo de cada programa com o alcance das rochas. O modelo de contato no programa PFC3D utilizado foi o método linear, e depois foi adicionada uma força de resistência ao rolamento para que o modelo conseguisse reproduzir o comportamento de solos mais macios. Os modelos foram simulados para casos reais, primeiro foram aplicados para uma encosta na BR116, onde um evento de deslizamento já tinha ocorrido. Foram realizadas simulações com os dois modelos na região, comparando-os e analisando com os vestígios deixados pelo evento. Outro caso estudado foi o de blocos susceptíveis ao deslizamento no bairro Glória, no município de Petrópolis. E por último, foram criados modelos para analisar o comportamento da vegetação em ambos os modelos. Com os resultados obtidos, conclui-se que o PFC3D consegue simular o processo de queda com melhor precisão através do auxílio da força de resistência ao rolamento. Para ambas simulações, o caso da BR116 e o do bairro Glória, os modelos apresentarem uma mesma tendência quanto a trajetória do bloco, mas melhorias na calibração do modelo linear com resistência ao rolamento precisam ser realizadas para que melhores resultados sejam obtidos. A presença de vegetação se mostrou como um mecanismo muito eficiente para perdas de energia no sistema, diminuindo significativamente a velocidade dos blocos. / [en] The aim of this work is to compare and analyze different numerical models on the rockfall process with different geometries, to determine the risks to the local infrastructure and population. It was used the software RAMMS and the modeling program PFC3D to simulate rockfall events. The first software was developed by the Swiss institute WSL, it uses the nonsmooth dynamics system, it is a dynamic behavior where there is no penetration between the rock and the surface and the contact laws are created by restrictions laws. And the second one uses the discrete element method applying the smooth dynamics system to simulate the event. On the first part of the work, numerical models were created on both models to simulate simple cases of rockfall events, to compare the main soil-rock parameter of both modeling program with the distance reached of the rocks. The contact model used on the PFC3D program was the linear method, and after the first simulations was added a rolling resistance force to the model reproduce soft. Both models were applied to real cases, the BR116 was the first case modeled, where some natural hazards have already happened. The results of the numerical simulations were compared and analyzed. It was also created a model to simulate the rockfall event of the Gloria neighborhood, in Petropolis; rocks that could fall were determined and simulations for those blocks were performed, the deposition area and the blocks trajectory of each model were examined. And finally, it was studied the vegetation effect of each model and its particularities. Through the results, it is possible to conclude that the PFC3D can simulate the rockfall process with better accuracy using the rolling resistance method. For both simulation, the BR116 and the Gloria neighborhood, the models showed the same tendency of the blocks trajectory, but it is necessary to improve the calibration of the rolling resistance coefficient. The existence of the vegetation revealed as a very effect lose energy mechanism on the system, decreasing significantly the velocity of the blocks.

Contact detection algorithms are needed in different areas of science and technology. From digital games and computer graphics to high-performance simulations and robotics. These algorithms require great computational effort and are prone to become the bottlenecks of its applications, even more when this computation must be done in real-time or large-scale systems. With the popularization of GPU cards use for both science and business, it is only natural that parallel implementations for this problem arise in the scientific community. In this work the main contact detection algorithms are analyzed and a numerical experiment is performed, with the goal of finding out which algorithm has better computational performance and memory use, or if they efficiency depends on different scenario features. For performing the experiment, a parallel Discrete ElementMethod application was developed using CUDA/C++ with the main algorithms presented in literature, besides these, the author proposes and implements the Sorting Contact Detection algorithm parallelization, that hadn’t been parallelized until now. The tests have found that the parallel Sorting Contact Detection algorithm is the most efficient in all studied scenarios, achieving a good performance and a superiormemory usage than its peers. / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Algoritmos de detecção de contatos são necessários em diferentes áreas da ciência e tecnologia, de jogos digitais e computação gráfica à simulações de alto desempenho e robótica. Esses algoritmos exigem grande esforço computacional e tendem a ser os gargalos das aplicação as quais fazem parte, principalmente em sistemas de grande escala ou em tempo real. Com a popularização das placas GPUs para uso científico e comercial, é natural que surjam implementações paralelas para esse problema. Nesse trabalho os principais algoritmos de detecção de contatos para GPU são analisados e é realizado umexperimento numérico, com objetivo de descobrir qual algoritmo é o melhor emtermos de desempenho computacional e uso de memória, ou se a eficiência de cada umdepende das diferentes características do cenários. Para a realização do experimento, foi implementado em CUDA/C++ uma aplicação paralela doMétodo dos Elementos Discretos comos principais algoritmos apresentados na literatura, além desses o autor propõe e implementa a paralelização do algoritmo de detecção com ordenação e busca binária que ainda não havia sido paralelizado. Após os testes é constatado que o algoritmo com ordenação e busca é o mais eficiente para todos os cenários estudados, obtendo nos resultados um bom desempenho em tempo de execução e com uso de memória muito superior aos outros.

[pt] O método de exploração Sublevel Caving é um dos métodos de extração massiva mais usados na indústria mundial de exploração subterrânea, sendo considerado pela indústria de mineração, num futuro próximo dentre os substitutos naturais das atuais minas a céu aberto. Uma operação Sublevel Caving requer que o maciço rochoso circundante ao minério rompa continuamente e se movimente para dentro do espaço criado pela extração do minério. Análises existentes na literatura consideram apenas configurações parciais do processo Sublevel Caving sem considerar o processo evolutivo da extração, e o dano induzido ao maciço rochoso decorrente deste processo. Esta dissertação desenvolve uma modelagem numérica utilizando o método dos elementos discretos para simular o mecanismo de ruptura e a subsidência causada pelo método de exploração Sublevel Caving, analisando o referido efeito e suas consequências na evolução do mecanismo de ruptura e subsidência no Sublevel Caving. O software comercial Particle Flow Code (PFC2D) foi selecionado para esta modelagem devido à capacidade de simular, em um evento de excesso de tensão, o fraturamento do maciço rochoso e sua desintegração, desta forma, originam-se o fluxo do material e os deslocamentos em grande escala, os quais são considerados fenômenos físicos dominantes que regem a formação da subsidência e fraturamento num processo Sublevel Caving. Os resultados obtidos nesse estudo mostraram-se satisfatórios, reproduzindo adequadamente a superfície de subsidência induzida por Sublevel Caving, conseguindo-se uma simulação física realista da sua evolução desde o início do fraturamento até à subsidência final. / [en] The Sublevel Caving Method is one of the most massive extraction methods used in underground mining industry worldwide and is considered by the mining industry as one of the natural replacements of the current open cut mines in the near future. A Sublevel Caving operation requires that the rockmass surrounding the orebody continually fails and moves into the void created by ore extraction. This dissertation develops a modeling using the discrete element method to simulate the failure mechanism and subsidence caused by Sublevel Caving method. Analyses reported in the literature consider only partial configurations of process Sublevel Caving, without taking into consideration the excavation evolution process, and damage induced to the rock mass resulting from this process. This dissertation analyzes this effect and its consequences on the evolution of failure mechanism and subsidence in Sublevel Caving. Particle Flow Code (PFC2D) was selected for modeling because of its ability to simulate, if the event of excess stress, fracturing and disintegration of the rock mass and large-scale deformation and material flow, to be simulated, which are believed to be the dominant physical phenomena governing the formation of subsidence and fracturing of Sublevel Caving. The results obtained in this study were satisfactory, reproducing properly the surface subsidence induced by Sublevel Caving, allowing physically realistic simulation of its evolution since the beginning of the fracturing to final subsidence.

Ce travail de thèse est consacré à l'étude des milieux granulaires humides. On étudie principalement le cas de faibles teneurs en eau. L'influence du phénomène de capillarité sur la rétention d'eau et sur le comportement mécanique du milieu est analysée. A chaque état d'équilibre, la distribution d'eau dans le milieu granulaire est obtenue en appliquant une succion homogène dans un volume élémentaire représentatif. La méthode multi-échelles utilisée se nourrit d'expérimentations/modélisation à l'échelle des interactions entre les grains, et produit des simulations en éléments discrets du comportement macroscopique qui sont comparées aux résultats d'expérimentation. A l'échelle des interactions capillaires, une étude expérimentale du pont liquide pilotée en succion permet la validation d'un modèle basé sur l'approximation toroïdale du profil du pont liquide. A l'échelle macroscopique, les courbes de rétention d'eau simulées sont proches des courbes de rétention d'eau expérimentales réalisées sur des milieux modèles composés de billes de verre. Enfin, le comportement mécanique macroscopique est simulé. On note une fragilisation du matériau en fonction de la succion dans le domaines des faibles teneurs en eau qui s'explique par une diminution de la densité des ponts liquides. La prise en compte d'une rugosité des grains permet une meilleure description de la transition entre un état humide et un état sec. L'analyse des contraintes montre la pertinence du tenseur des contraintes associé aux interactions attractives dans l'étude de l'évolution de la contrainte à la rupture (compression simple) et de la compressibilité (compression oedométrique) en fonction de la succion. / This work deals with study of humid granular media. Weak water percent media are mainly studied. The influence of capillarity on water retention and mechanical behaviour is analyzed. At each state of equilibrium, water distribution is obtained by applying a homogeneous suction in the entire elementary representative volume. The multi-scale approach used, is based on experimental/modelling at the local scale (interaction between grains) and produce discrete elements simulation of the macroscopic behaviour which are compared to experimental results. At the scale of the interactions, an experimental study of the liquid bridge piloted by suction validates a model based on the toroidal approximation of the liquid bridge shape. At the macroscopic scale, simulated water retention curves are near of experimental water retention curves made on a model sample composed of glass beads. Then, the mechanical behaviour is simulated. We notice a loss of cohesi on in function of suction because of a reduction of the liquid bridge density. The introduction of a roughness allows a best representation of the transition between the non saturation state and the dry state. The analysis of the stress reveal the pertinence of the stress tensor associated with attractives interactions in the evolution of rupture stress (simple compression) and compressibility (oedometric compression) in function of suction.

Au cours du temps, les cavités souterraines sont soumises à un vieillissement et plusieurs types de dégradation peuvent apparaitre. Les anciennes exploitations souterraines, parfois constituées d’un ou plusieurs niveaux, n’ont, sans doute, pas été conçues pour être stables à long terme. Elles ont été réalisées à une époque où n'existaient pas d'enjeux en surface, de zones de travaux, ce qui permettait d’éviter de se préoccuper des mouvements de sol induits. Elles ont pu quelquefois être totalement ou partiellement remblayées, mais pas de manière systématique. L’effondrement d'une cavité souterraine engendre la déconsolidation des niveaux supérieurs des terrains de recouvrement. Ces mécanismes peuvent provoquer en surface deux types de désordres : un affaissement ou un fontis. L'affaissement et le fontis peuvent provoquer des graves dommages aux structures et aux infrastructures en surface mais aussi mettre en péril la sécurité des populations. Les travaux de cette thèse s’articulent autour de l’étude du foisonnement, du fontis et de sa propagation dans les terrains de recouvrement.Les objectifs de cette thèse sont doubles : tout d’abord il s’agit d’étudier le foisonnement de la roche lors d’un effondrement de toit de carrières souterraines ; ensuite il s’agit de modéliser la propagation du fontis dans les terrains de recouvrement et ainsi hiérarchiser les paramètres associés à ce phénomène.La première partie de cette thèse repose sur une étude bibliographique qui récapitule les méthodes d’exploitation, méthodes d’analyse de stabilité de carrières souterraines, méthodes de prévision de la hauteur d’effondrement et estimation de foisonnement. A l’issue de cette synthèse bibliographique l’étude s’est focalisée sur les carrières souterraines à faible profondeur exploitées par chambres et piliers. Ainsi, la modélisation numérique par la méthode des éléments discrets (MED) a été choisie pour analyser l’instabilité des toits de carrières souterraines.La seconde partie s’intéresse au développement d’un modèle numérique qui a pour objectifs : d’une part, le développement d’un Programme de Discrétisation des Massifs Rocheux (PDMR) qui constitue le préprocesseur du logiciel STTAR3D et le développement d’un code permettant le calcul du coefficient de foisonnement des débris de l’effondrement. D’autre part, l’implémentation des lois de comportement sur STTAR3D.La troisième partie consiste à déterminer, d’une part les caractéristiques physico-mécaniques d’échantillons prélevés dans la carrière de la Brasserie (Paris-France), qui a été choisie pour une tester le modèle développé et d’autre part, les deux paramètres de la loi de comportement utilisée pour modéliser les contacts à savoir  et µ.Enfin, la dernière partie de ce travail est constituée des simulations numériques pour lesquelles les paramètres de la loi de comportement mesuré expérimentalement ont été introduits dans STTAR3D. Dans la première étude numérique menée, on s’intéresse à l’effet de la hauteur de chute, du rayon de l’ouverture initiale du fontis et du degré de fracturation sur le foisonnement des décombres, ainsi qu’à l’effet de la variation du foisonnement sur la hauteur de l’effondrement et sur l’affaissement. Dans un second temps, on réalise un modèle de la carrière de la Brasserie dont on calcule le comportement par simulation numérique afin d’obtenir l’affaissement en surface et la hauteur de l’effondrement qui sont comparés aux observations in-situ / Over time, the underground cavities are subjected to aging and several types of degradation can occur. The old underground cavities have probably not been designed to be stable over the long term. They have sometimes been totally or partially backfilled, but not in a systematic way. The collapse of a mine causes deconsolidation of the upper levels of the overburden. These mechanisms can cause two types of disorders on the surface: subsidence or sinkhole. Subsidence and sinkhole can cause severe damage to structures and infrastructures in surface, but also jeopardize the safety of the population.The work of this thesis revolves around the study of rock's bulking, sinkhole and its spread in the overburden. The aims of this thesis are twofold: firstly, to study the bulking of rock during the roofs mine collapse; Secondly, modeling the spread of the sinkhole in the overburden and thus to prioritize the parameters associated with this phenomenon.The first part of this thesis is a bibliographical study which summarizes the methods of exploitation, methods of analysis of stability of underground quarries, methods of prediction of the height of collapse and estimation of the bulking factor. At the end of this bibliographic synthesis, the study focused on shallow underground quarries operated by rooms and pillars. Thus, numerical modeling using the discrete element method (MED) was chosen to analyze the instability of roofs of underground quarries.The second part focuses on the development of a numerical model with the following objectives: on the one hand, the development of a Rock Mass Discretization Program (RMDP) which constitutes the preprocessor of the STTAR3D software and the development of a Code allowing calculation of the bulking factor of the rubble of collapse. On the other hand, implementation of the behavior laws on STTAR3D.The third part consists of determining, on the one hand, the physicals and mechanicals characteristics of samples taken from the quarry of the Brasserie (Paris-France), which was chosen to test the model developed. On the other hand, determining of parameters of the behavior law used for modeling the contacts, namely “” and “μ”.Finally, the last part of this work is made of numerical simulations for which the parameters of the behavior law measured experimentally have been introduced in STTAR3D. In the first numerical study, we investigate the effect of fall height, the radius of the initial opening of the sinkhole and the fracturing degree on the bulking of the rubble, as well as the effect of variation of the bulking on the collapse height and on the subsidence. In a second step, a model of the Brasserie’s mine is realized, the behavior of which is studied by numerical simulation in order to obtain the subsidence on the surface and the collapse height, which are compared with the in-situ observations

Pour voies ferrées à grandes vitesses (LGV, Lignes à Grande Vitesse), la durabilité des performances du ballast de chemin de fer n’est pas aussi importante qu’attendu. Le comportement mécanique de cette couche granulaire mince dépend fortement de la forme, la taille et la minéralogie des grains. Sur les LGV, les grains s’usent plus vite qu’attendu, essentiellement à cause de l’accumulation des opérations de maintenance appelées bourrage. Une conséquence à cela est un renouvellement complet du ballast avec une fréquence largement supérieure à ce qui était initialement prévu à la création de ces lignes.Soumis à des contraintes dynamiques combinées (trafic ferroviaire et des opérations de bourrage), les grains de ballast se dégradent par fragmentation et par attrition aux contacts. Les conséquences directes de cette dégradation progressive sont l’évolution de la taille et de la forme des grains. La courbe granulométrique est alors translatée vers les petits éléments, avec une présence notable de particules très fines résultant de l’usure des grains. De plus, l’angularité des grains est progressivement diminuée. Au-delà d’un certain temps, le cumul de dégradation se traduit par une chute des performances mécaniques du ballast. Le ballast ne remplit plus efficacement ses fonctions. La résistance latérale de la voie est réduite, limitant ainsi la répartition des contraintes sur la plateforme et l’ancrage des traverses. La présence excessive de fines rend le bourrage inefficace et diminue la perméabilité de la voie. Par conséquent, pour trouver des solutions optimales pour prolonger la durée de vie du ballast, il est nécessaire d’abord de bien comprendre les origines et mécanismes menant à l’usure des grains, pour finalement construire un modèle prédictif de dégradation.La dégradation des interfaces au contact génère de particules fines. La quantité de fines produite, laquelle dépend des conditions de chargement, est classiquement prédite par l’équation d’Archard. Ce modèle part du principe que le volume d’usure généré est proportionnel à la force normale et au déplacement relatif entre les surfaces en contact. La simulation numérique par éléments discrets (NSCD) d’une portion de voie de chemin de fer soumis à un chargement cyclique est un outil nécessaire pour réaliser la transition entre l’échelle de la voie et l’échelle du contact, fournissant les informations sur le ballast en tant que couche granulaire, depuis son comportement global jusqu’aux forces de contact et les déplacements relatifs entre les grains. Les contacts montrant un grand potentiel de génération de fines (selon le modèle d’Archard) sont identifiés et reproduits expérimentalement avec des essais de cisaillement entre deux grains. Parallèlement, l’essai d’attrition Micro-Deval est utilisé pour relier les résultats numériques et expérimentaux, et ainsi valider le modèle d’Archard, et pour suivre l’évolution de la forme des grains avec l’aide des scans d’un échantillon de grains par tomographie RX à différents états d’usure. Les deux campagnes d’essais montrent la faiblesse des aspérités les plus aiguisées, dont spécialement celles sur les arêtes et sommets.Un modèle prédictif d’usure en deux phases est donc proposé. La première phase décrit une usure rapide due aux fortes contraintes normales à l’interface de contact, et la deuxième phase décrit un taux d’usure plus modéré. Une contrainte seuil permet d’identifier clairement le passage d’une phase à une autre. Sur la base des déplacements relatifs intergranulaires observés dans la simulation numérique discrète, ce modèle est appliqué pour chaque contact dont l’histoire de chargement est variable. Une estimation de la courbe de génération de fines dans la voie est ainsi proposée. / After some years of high-speed lines in France (HSL), ballast has proven not to be resistant enough. The performance of ballast, as a thin layer of coarse grains, strongly depends on the shape, size and mineralogical nature of the grains composing it. However, in HSL, grains wear faster than expected due to the traffic of trains at high speeds and the accumulation of maintenance operations (tamping). Ballast replacement has therefore been required much before than its originally expected lifespan.Under the dynamic stresses imposed by the circulation of trains and tamping operations, ballast is gradually worn by fragmentation of grains and attrition at the contacts. The direct consequence of this degradation is the evolution of grain size and shape: the grading curve is shifted towards small and fine particles and the grains progressively lose their angularity. Eventually, the cumulated wear will no longer allow ballast to perform properly: the shear resistance of the layer is reduced limiting both the anchorage of sleepers and the distribution of loads to the platform. In addition, the presence in excess of fine particles renders tamping ineffective (fast evolution of track defaults) and reduces the permeability of the track. Thus, in order to search for optimized solutions for prolonging ballast lifespan, it is crucial to first understand the origins and mechanisms leading to ballast degradation when it is subjected to complex loading, for building a predictive model of ballast wear.The degradation of contact interfaces generates fine particles. The associated mass flux, which depends on the loading conditions, has been classically predicted by Archard equation. The model assumes that the generated volume of wear is proportional to the normal force and the relative displacement between the surfaces. Therefore, it is crucial to quantify the forces at the contact scale and the relative displacements between ballast grains in sliding contact. Discrete elements simulations by NSCD are used as a tool for performing a change in scale from the track scale to the contact scale, giving information of ballast as a granular layer, from its global behaviour down to the contact forces and relative displacements between grains. Contacts with a higher potential of generating fine particles (according to Archard model) are then identified and reproduced experimentally by two-grain shearing tests. In parallel, the Micro-Deval standard attrition test is used as a link between numerical and experimental results to validate Archard model, and to study the evolution of grain morphology by scanning a sample of grains using X-ray tomography at different stages of the test. Both experimental campaigns show the weakness of sharp asperities, especially on edges and vertexes.A model in two phases is proposed, accounting for a first phase of fast and aggressive degradation due to the high stress at the contact interface and a more stable second phase with a lower wear rate. A critical stress is identified as a threshold between phases. This model is then applied at each individual contact on the numerical simulations, resulting in a first approach of the production curve of fine particles within the track.