Options
2015
Conference Paper
Title
Modeling of interacting fibers based on 2D images for fibrous filter media simulation
Abstract
Macroscopic properties of fibrous filter media are highly influenced by the geometry of filter media used. Stochastic geometry models for fiber systems in the literatures possess some limits when modeling a fiber system with fiber bundles. In this work, a stochastic model for fiber systems with interacting fibers is developed. The starting configuration of this model is generated using random walks. During this random walk, for the first fiber in a bundle, a starting coordinate is drawn from a spatial Poisson point process and its main fiber orientation is randomly chosen from a probability density function on the sphere. All further fibers from that bundle start at coordinates generated by a small displacement of the initial fiber's starting point and use the same main fiber orientation. Additionally, shortest distance between current coordinate and coordinates of previous fibers are also considered during a random walk. Finally, force biased fiber packing is used to make these fibers non overlapping. A stochastic geometry model close to real data is generated by our new approach. Each fiber radius for this model is chosen from a fiber diameter distribution which is empirically measured from a real 2D SEM image of a real filter medium. Area density of our simulated SEM image is relatively close to the real data. In conclusion, our new stochastic model can be used to model a densely packed non-overlapping fiber system with bending fibers and fiber bundles, suitable for subsequent numerical simulation of filtration behavior.