
G.
William Baxter  Cellular Automata Models of Flowing Granular Media


I am studying cellular automata models
of gravity driven flows of asymmetric grains through a hopper. I am particularly
interested in orientation and density effects in these flows.
Since the correct continuum equations
describing the flow of a granular material are the subject of considerable
debate, it is not possible to directly model them on a computer. Simulations
using molecular dynamics are possible and very successful, but MD simulations
are best for small numbers of grains with simple shapes and simple interactions.
[Although computer speed and algorithmic sophistication has increased dramatically
in recent years, it is still true that computational complexity can quickly
become a limiting factor.]
Cellular automata are a natural choice
for this problem. The characteristics of an automata are as follows.

Automata operate on a lattice (here, hexagonal.)

Automata operate in discrete time steps.

Automata define a discrete value on each
lattive site (here, a site may be a wall or grain or a hole. Grains are
described by their orientation and motion direction.)

Automata use a rule to determine how the
values at each lattice site change at each time step.
In the present case, grains are described
by their orientation and direction of motion. They may have one of three
orientations, along the basis vectors of the lattive, and one of seven
velocity or motion directions, toward one of the six nearest neighbor sites
or at rest. The most challenging component is the rule. This model uses
an energy minimization rule. We define an energy function which is a function
of the orientation and motion of both a grain and its nearest neighbors.
The new orientation and motion of each grain is chosen to most lower its
energy.
In this model, typical lattices have
1/4 million sites. We begin with an empty lattice with wall sites forming
a hopper. With the outlet plugged, grains are allowed to fall in. After
filling, the output is unplugged and the grains are allowed to flow out.
The resulting flows have much in common with real flows in a similar geometry.
Top 2000 iterations after the
flow begins, the flow looks like this. Colors denote grain orientations.
White is walls; black is holes.
Bottom A density picture of
the aboveleft image. Here the gray level gives the local grain density.
Mail suggestions and complaints regarding
subject material to "gwb@shahrazad.bd.psu.edu". 





The Pennsylvania State University ©199699
Copyright
Statement
This page was created and is maintained by Isaac
Hagenbuch
Updated February 10, 1999

