Non-overflow flow-through rockfill structures are river engineering elements used to attenuate and delay inflow hydrographs. They represent expedient places to deposit rather enormous quantities of waste rock at mountainous mine sites. Their application has become so common that matters of safety regarding their design have been laid out in Section 8.5 of the Canadian Dam Safety Guidelines (CDA 2007). The research described herein was directed at investigating the different aspects of the hydraulics of these flow-through rockfill structures. In order to assess the potential for an unraveling failure of flow-through rockfill dams, a systematic study of the hydraulic design of these structures was conducted and the non-linear nature of flow through these structures was dealt with using a p-LaPlacian-like partial differential equation. Subsequently, factors of safety against this type of failure are presented for a range of downstream slopes, thus showing the unsafe combinations of embankment slope and particle diameter. Three different index gradients within the toe of such structures were investigated. In this regard, the gradient most suitable for independently computing the height of the point of first flow emergence on the downstream face is examined and a method for independently computing the variation in hydraulic head within that vertical (which allows for the toe of the structure to be isolated) is presented. An additional gradient that allows for the independent estimation of the default tailwater depth is proposed. In order to provide better tools to assess the behavior of these embankments at the toe, laboratory and analytical studies were undertaken. In this regard, the hydraulics associated with the zone of the downstream toe were studied. The depth variation of the seepage-face was computationally modeled, and two approaches for solving the spatially varied flow (SVF) condition problem within the toe region undertaken. The results show that a dual linear variation in depth can be used to good accuracy, without inducing any unrealistic exit gradients in the zone of primary concern with respect to unraveling. It is hoped that these techniques and computational tools provided herein will aid in facilitating the design and assessment of these flow-through rockfill structures.

Les structures en enrochements sont parmi les ouvrages les plus usuels de génie civil (barrages, murs de soutènement,. . . ). Des tassements importants peuvent apparaître tout au long de leur durée de vie et sont principalement dus à la rupture des blocs rocheux. Cette thèse propose un modèle numérique permettant de simuler le comportement de matériaux granulaires présentant des ruptures de grains. Afin de prendre en compte la nature discontinue de ces milieux, la méthode des éléments discrets est utilisée. La modélisation adoptée est de type "Non-Smooth Contact Dynamics", où les grains et particules sont supposés rigides. Afin de générer des blocs ayant des formes complexes, un modèle de grain 3D est proposé. Ce modèle de grain est ensuite discrétisé en sous-éléments de forme tétraédrique liés par des liaisons cohésives afin de pouvoir représenter la rupture. Un critère de rupture de Mohr-Coulomb est utilisé. Le modèle est implémenté sur la plateforme logicielle LMGC90. Le modèle est d’abord éprouvé lors de simulations d’écrasement de blocs cassables entre deux plaques. Plusieurs paramètres contrôlant la résistance du grain sont étudiés : cohésion intergranulaire, taille, discrétisation, forme et orientation du grain. L’effet d’échelle observé sur ce type de matériau est vérifié. Le modèle est ensuite testé lors de simulations numériques de compression œdométrique d’enrochements. L’effet des paramètres du modèle et de l’assemblage du milieu granulaire est également étudié. Les simulations œdométriques sont confrontées à des résultats expérimentaux et présentent une bonne concordance. Enfin, des expérimentations numériques sont menées afin d’étudier les énergies mises en jeu dans ces essais. L’énergie de création de surface est estimée pour ce type de matériau. Les résultats sont proches des données de la littérature. / Rockfill structures are very popular among civil engineering structures (dams, retaining walls, . . . ). Important settlements can take place during the lifetime of these structures, settlements mainly caused by the breakage of rockfill grains. This thesis proposes a numerical model that allows the simulation of the behavior of granular materials exhibiting grain breakage. To take into account the discrete nature of these media, the discrete element method is chosen. The adopted strategy is the Non-Smooth Contact Dynamics method, where grains are considered to be rigid. To generate blocks having complex shapes, a 3D grain model is suggested. This grain model is then discretized into tetrahedral subgrains, joined together using cohesive bonds so that breakage can be simulated. A Mohr-Coulomb failure criterion is used for the cohesive bonds. The model is implemented into the LMGC90 software platform. At first, the model is tested in single grain crushing simulations between two plates. Multiple parameters controling the strength of the grain are studied : the intra-granular cohesion, the size, the discretization and the orientation of the grain. The scale effect that characterizes this type of material is verified. Then the model is tested in numerical simulations of œdometric compression of rockfill. The influence of the parameters of the model and of those of the granular medium are studied. The results of œdometric simulations are compared to experimental results, and present a good agreement. Lastly, numerical experimentations are conducted in order to study the energies that are brought into play in the simulations. The surface creation energy is estimated for this type of material. Results are close to the data provided in the literature.