Titulo: Discrete Element Methods in Structural Mechanics
Conferenciante Prof. M. Bischoof
Area Tematica: Computational material failure and fracture
Fecha, lugar y evento: 26/10/2011 , Barcelona, II International Conference on particle-based Methods Fundamentals and Applications PARTICLES 2011
The presentation discusses application of discrete element models to problems in structural mechanics. The emphasis is on investigation of their potential advantages (and drawbacks) in comparison to sophisticated continuum-based approaches in modelling non-linear behaviour of frictional material as well as representation of materials with microstructure and evolving discontinuities. Typical arguments in favour of discrete, particle based approaches are the essentially discrete nature of matter (not only on the atomistic or molecular level but also in terms of individual grains, crystals, fibres or aggregate on a meso-scale), natural inclusion of intrinsic length scales and heterogeneous microstructures as well as relatively easy treatment of topology changes due to cracking and fragmentation. A discrete element model with polygonal particles, developed earlier ( ? ) is discussed in some detail. Contacting particles are allowed for a certain overlap, the amount of which is utilized to compute normal and tangential contact forces. As opposed to approaches with super-ellipsoids, here the sharp corners of the polygons may cause some trouble and need special attention. Various possibilities are investigated for modelling of cohesion, trying to find the optimal trade-off between realistic representation of the complex phenomenon of cohesion and computational efficiency: a brittle beam model, a beam with damage and a softening model. Simple, conceptual experiments with steel nuts are utilized to verify the two-dimensional model with respect to its capabilities of predicting the deformation behaviour and development of shear bands in a ?material? with a regular microstructure. Using a homogenization procedure, computational results obtained from the discrete element model may be post-processed to obtain continuous quantities like stresses and strains which facilitates comparison of the results with those obtained from continuum models. Due to the computational expense of discrete element models large scale computations of entire structures are at present prohibitive. Research in progress therefore focuses on a model-adaptive scheme combining discrete and continuous (finite element) schemes. In this context, instead of applying a domain decomposition technique and dealing with interfaces, the idea is to naturally obtain a smooth horizontal transition using elements of different levels of resolution within a modeladaptive process. Some of the technological background is borrowed from the quasi-continuum method.