Abstract
Ballast gluing is a process that consists in spreading a binder on the ballast as a set of rocky aggregates in order to limit its movements, and consequently the maintenance operations, which are very costly in time and resources. To understand and optimize the ballast gluing process used in railway operations by the SNCF, the development of a numerical model of ballast glue is proposed as well as its exploitation in a discrete element calculation code LMGC90, based on the NonSmooth Contact Dynamics approach.To develop this model, an approach combining laboratory and numerical experiments is proposed, experiments conducted at different scales. In a first step, experimental tensile tests are carried out on reproductions of contacts between bonded ballast grains. These tests provided local data used to calibrate the two parameters of a first simple numerical cohesive contact law, of Dugdale type: the maximum cohesive resistance of a contact, and the separation distance beyond which the detached contact is considered.A second law developped by Venzal, more precise, is also calibrated. It is described by an initial stiffness, a maximal strain and allows the contact to damage.These models are then exploited thanks to the LMGC90 platform dedicated to the modeling of divided media to carry out "model" simulations, and then simulations applied to railway issues.Thus, numerical tests of triaxial compression carried out on a cohesive polyhedral granular medium have highlighted the importance of the cohesion law used : simulations executed with Dugdale law and important cohesion intensity cannot respect the application of an isotropic pressure on the sample because of the importance of its strain. Tests realized with Venzal law are meaningful, and link the evolution of sample shearing resistance to the evolution of contact cohesion damage. At first, the sample quickly damage itself, because its volumic expansion breaks the cohesion of contacts which actively take part of the medium resistance ; at the same time, the system shearing resistance increase considerably. It decreases and stabilizes when the cohesion damage is sufficient.Experimental and numerical lateral resistance tests have also been conducted, with different bonding configurations. Experimental tests, in addition to providing reference data, prove that adding cohesion to the ballasted mass increases its lateral strength, and quantify this reinforcement (up to a 43% improvement for deep bench bonding). A surface gluing of the seat offers a 22% reinforcement for a lower quantity of glue. Numerical lateral strength tests show that cohesion can act on two different aspects: reinforcing the resistance network that already opposes the lateral movement of the cross member, or adding an additional opposition network on top of the pre-existing one. This work thus offers a numerical tool to respond to the railway problem raised and proposes various avenues for reflection.