An idealized fiber network model is developed to study particle sieving by
percolation. The model is based on a Poisson line field in two dimensions,
whereby the fibers are modeled as infinite length strips of finite width, w
hile particles are modeled as circular disks. The probabilistic sieving pro
blem is equivalent to percolation of disks through convex polygons of the P
oisson line field. The area-based probability of retention (equivalently, p
ercolation) is described by an equation which depends on a single non-dimen
sional number that is proportional to the number of fiber per unit area and
the particle size. The physical meaning and the applicable range of this n
umber are discussed. Finally, using a Monte Carlo simulation, the model is
generalized to anisotropic Poisson line fields. (C) 2000 Elsevier Science I
nc. All rights reserved.