Abstract : The objective of this thesis work is to model the high-strain rate and dynamic fragmentation of brittle materials using the Discrete Element Method. Fragmentation is an irreversible, nonlinear and random phenomenon.It can be found in many practical applications in engineering and can take place at various length scales. This research work takes advantages of computer simulations to model this phenomenon and to predict a few statistical parameters related to fragmentation including number, size, and size distribution of fragments. To this effect, the Discrete Element Method was found to simulate efficiently fracturing, which is a discrete phenomenon by nature. However, an efficient computer simulation is not sufficient for representing fragmentation. It also needs to account for a rupture criterion and a damage criterion. This rupture criterion is defined at the contact points between particles where it generates a local damage that decreases the local stress until a discrete crack appears. In a first step, the rupture criterion of Camacho-Ortiz |24| has been introduced in the Discrete Element Method. This criterion expresses damage as a function of crack opening. When the local stress reaches a rupture threshold, it decreases linearly with the crack opening until the rupture is obtained. This first criterion gives good results on the convergence of fragmentation parameters in simple cases |69, 88, 97, 143-147|, but requires a great number of particles. In a second step, another rupture criterion has been introduced for simulating the fragmentation of more complex three-dimensional structures for high-strain rates. This rupture criterion is based on a different physical approach that accounts for heterogeneous brittle materials with defects. These defects can evolve and cause local failure. They are introduced per unit volume elementusing a Weibull probability distribution |29, 33, 42-44|. This distribution depends on the local stress until the local stress reaches an activation threshold. After that, the defects propagate and form areas of relaxation in which defect cannot evolve. The damage evolves as these areas of relaxation evolve. This second rupture criterion has been validated in simple cases by examining the convergence of the statistical parameters of fragmentation. Compared to the first criterion, the second criterion requires ten times fewer particles. After, a more complex three-dimensional case, dynamic tensile tests in Hopkinson bars, has been treated.