Options
2015
Conference Paper
Titel
Bridging the gap from CT-analysis to predictive finite element modeling
Abstract
Predictive modelling of the mechanical behavior of granular solids such as fertilizer or sand is a difficult task as the heterogeneity at the mesoscale (scale resolving individual grains) strongly affects the response at the macro-scale. Therefore, multi-scale simulation approaches describing this class of materials are of particular interest. A common multi-scale approach is the bottom-up approach where a series of connected simulations are performed. For example, simulations of representative structures are performed at the meso-scale, resolving individual grains. The results obtained from meso-scale simulations are then homogenized and used for describing the material at a larger scale, the macro-scale. It is however difficult to obtain accurate descriptions of the meso-scale due to the lack of information about grain sizes and grain size distributions. Computer Tomography (CT) imaging can provide some information about the structure. Unfortunately, the resolution is often not good enough to obtain reliable data which could be used to generate finite element meshes. This paper describes a continuous process chain from high resolution CT imaging to generation of FE meshes for the simple example of sand. Using CT, a 3D-greyscale image is generated. This is the critical step since CT-images often suffer from spurious random disturbances which may prevent reconstruction. The image quality is enhanced by applying a smoothing filter, followed by a series of detection filters to extract structural information from the image. These manipulations allow the detection of borders between individual grains, and the detection of interfaces for contact zones between grains. With this information a surface grid is generated by distributing points among the surface of the detected grains. Those points are then successively connected to a surface mesh. Finally, the mesh is filled with tetrahedral elements using meshing tools such as Gmesh. The finite element mesh generated this way is an exact representation of the sample originally scanned in the computer tomograph. Having material models in place describing the individual grains, this finite element mesh can for example be used in the framework of a multiscale analysis of granular materials.