Master presentation: Anita Ullrich

Examinator: Mohsen Mirkhalaf, Department of Physics
Opponent: David Rinman, Complex Adaptive Systems

Handledare:

• Elin Solberg, Fraunhofer-Chalmers Centre
• Klas Jareteg, Fraunhofer-Chalmers Centre
• Christoffer Cromvik, Fraunhofer-Chalmers Centre

Development of a DEM-FEM framework for infrastructure simulations

Abstract

This thesis presents a coupling algorithm of the discrete element method (DEM) and finite element method (FEM). The algorithm formulates an explicit coupling of transient simulations of particle systems interacting with elastic bodies. To lay a foundation for the requirements in terms of stability and temporal and spatial resolution, the DEM and FEM methods are introduced. The coupling algorithm is implemented in a Python framework, using the FCC in-house solvers Demify® and LaStFEM. The combined tool is applied to three different main scenarios. As a first case, the solver exchange of forces between the DEM and FEM solver is verified using a fixed elastic beam simulation with uniform load, comparing the deflection under the load of particles to an analytical condition. Second, the dynamic accuracy and stability of the coupling method is proven on a simulation of a steel sheet deflection under the load of particles flowing on the elastic object. The simulations are compared to experimental results and show good agreement with a measured sheet deflection. Finally, the coupled solver is used to simulate the interaction between a timber sleeper and a rock particle ballast bed. The particles are in the third case represented by a polyhedron particle model. The system is studied for variations of both material properties as well as different simulation parameters. The coupled solver is shown to capture dynamic effects in the ballast bed under a dynamic load cycle. The simulation results are compared to experimental results of the pressure distribution in the bed from the open literature and demonstrate good qualitative and quantitative agreement with the experiments. The overall performance of the different parts of the solver is presented and it is shown that the developed tool is capable of simulating large scenarios with very good performance on desktop computers with a single GPU.