Stochastic 3D modeling of fiber-based materials

Abstract A stochastic multi-layer model is developed describing the microstructure of materials which are built up of strongly curved, but almost horizontally oriented fibers. This fully parametrized model is based on ideas from stochastic geometry and multivariate time series analysis. It consists of independent layers which are stacked together, where each single layer is described by a 2D germ-grain model dilated in 3D. The germs form a Poisson point process and the grains are given by random polygonal tracks describing single fibers in terms of multivariate time series. Exemplarily, on the basis of 2D data from SEM images, the parameters of the multi-layer model are fitted to the microstructure of a non-woven material which is used for gas-diffusion layers in PEM fuel cells. Therefore, an algorithm is presented which automatically extracts typical fiber courses from SEM images. Finally, the multi-layer model is validated by comparing structural characteristics computed for 3D data gained by synchrotron tomography from the same material, and for realizations drawn from the multi-layer model.