Dr. Parris Egbert knows sand—really knows sand.
Egbert, as director of the
Advanced 3D Graphics Lab, uses observation and mathematical techniques to digitally simulate real-life phenomena. One of Egbert’s current projects is creating computer graphics software that simulates how sand behaves.
“Sand is a little different [than fluids] because it acts differently depending on what it is doing,” Egbert said. “When it is just sitting there, it acts like a solid. If you have a pile of it on a piece of wood and then you tip the wood up, then it starts flowing, so it acts like a fluid. If you tip it up really fast, then it scatters all over and acts like a gas.”
Because “particulate fluids” like sand and snow behave in different ways according to the stimulation they receive, creating lifelike 3D simulations is notoriously difficult. As the sand particles transition from flowing as a liquid to shooting into the air as a gas, the computations become increasingly complex, expensive, and time-consuming.
“What we’ve been working on is creating a system that can determine when we need to switch from one state to the next,” Egbert said. “How do you optimize so that you only transition the [particles] you need to transition and not more?”
Realistic models of particulate fluid flow have a number of practical applications. The “big ones” are movies and games, according to Egbert, but applications also include medicine and manufacturing. Egbert and his lab also work on creating intuitive models for ferrofluids.
“Ferrofluids are fluids that have nanometallic particles in them, and when you apply a magnet near them, they do crazy things,” Egbert said. “There are actually some very nice real world applications for ferrofluids, like targeting drugs or targeting antibodies in the human body. They’re also used for lubricant applications in rockets and other things.”
When ferrofluids come into contact with magnets, the magnetic forces cause the metallic particles to stack up, making the fluid form hedgehog-like spikes.
“To date nobody has really been able to simulate them very well,” Egbert said.
Mathematicians and physicists worked out the theory behind fluid flow years ago. The Navier-Stokes equations, for example, describe fluid flow and viscosity and were first derived in the early nineteenth century. Egbert’s problem, however, is how to visually model these theoretical mathematical equations with low computational cost.
“We’re looking at some machine learning, some deep-learning algorithms for doing fluid simulation so you can get the computer to learn how to do the fluid simulation,” Egbert said. “This way you don’t have to do the full-blown Navier-Stokes equations simulations, which tend to be very time consuming.”
Egbert’s end goal is to make his computer simulations precisely mirror reality.
“The ideal is to take a video of sand flowing and then have our simulation side-by-side with it and have it look the same.”