Numerical analysis of railway ballast behaviour using the Discrete Element Method

The development of high-speed train lines has increased significantly during the last twenty-five years, leading to more demanding loads in railway infrastructures. Most of these infrastructures were constructed using railway ballast, which is a layer of granular material placed under the sleepers whose roles are: resisting to vertical and horizontal loads and facing climate action. Moreover, the Discrete Element Method was found to be an effective numerical method for the calculation of engineering problems involving granular materials. For these reasons, the main objective of the thesis is the development of a numerical modelling tool based on the Discrete Element Method which allows the users to understand better railway ballast mechanical behaviour. The first task was the review of the specifications that ballast material must meet. Then, the features of the available Discrete Elements code, called "DEMPack", were analysed. After those revisions, it was found that the code needed some improvement in order to reproduce correctly and efficiently the behaviour of railway ballast. The main deficiencies identified in the numerical code were related to the contact between discrete element particles and planar boundaries and to the geometrical representation of such a irregular material as ballast. Contact interactions between rigid boundaries and Discrete Elements are treated using a new methodology called the Double Hierarchy method. This new algorithm is based on characterising contacts between rigid parts (meshed with a Finite Element-like discretisation) and spherical Discrete Elements. The procedure is described in the course of the thesis. Moreover, the method validation and the assessment of its limitations are also displayed. The representation of irregular particles using the Discrete Element Method is a very challenging issue, leading to different geometrical approaches. In this work, a deep revision of those approaches was performed. Finally, the most appropriate methods were chosen: spheres with rolling friction and clusters of spheres. The main advantage of the use of spheres is their low computational cost, while clusters of spheres stand out for their geometrical versatility. Some improvements were developed for describing the movement of each kind of particles, specifically, the imposition of the rolling friction and the integration of the rotation of clusters of spheres. In the course of this work the way to fill volumes with particles (spheres or clusters) was also analysed. The aim is to control properly the initial granulometry and compactness of the samples used in the calculations. After checking the correctness of the numerical code with simplified benchmarks, some laboratory tests with railway ballast were computed. The aim was to calibrate the ballast material properties and validate the code for the representation of railway ballast behaviour. Once the material properties were calibrated, some examples of a real train passing through a railway ballast track were reproduced numerically. This calculations allowed to prove the possibilities of the implemented tool.