Conference Trifecta
Posted June 17, 2005
If you had an interest in 3D simulation and visualization, this was a good week to be around MIT, since there were three closely-related conferences happening on campus: the Symposium on Solid and Physical Modeling, the Shape Modeling International Conference, and the huge Third M.I.T. Conference on Computational Fluid and Solid Mechanics. Whew! Naturally I had to attend some interesting sessions, and fortunately nobody cared that I wasn't really registered.
A detailed write-up of all the sessions would be gigantic, and not that interesting, so I'll just mention a few things that struck me as somewhat insightful or worth noting for my own future reference.
- For visualizing fluids, simulating the bulk motion is important, but a good treatment of the surface is critical, since that's what you see.
- European cars measure trunk capacity by how many one-liter (20x10x5 cm) boxes can be packed into it, not by continuous volume. It takes engineers sometimes days to figure out a near-optimal packing to squeeze in an extra liter for the car's spec sheet.
- One group came up with a technique that, among other things, computes the total acceleration magnitude of a tree of articulated joints without having to calculate each one.
- The ornicopter is a proposal for a helicopter that also flaps its rotor blades slightly as they rotate. This generates thrust as well as lift, to exactly cancel the torque from the motor and thus remove the need for a tail fan.
- The HYPS project is a weird-looking multi-bladed fan that promises to be able to generate energy well from flows that come from many directions. (Sorry I can't find a good link for it.)
- Fabrics tend to have two regimes of linear bending response to forces: a small-angle behavior from individual bending fibers, and a large-angle behavior from fibers slipping past each other.
- Simulating flame is very nasty, because to do it right you need convection, conduction, and radiation modeling, plus lots of chemistry: soot formation impacts the situation a lot, for instance.
- In a type Ia supernova (fusion explosion of a white dwarf), the fusion flame front is only a millimeter thick but travels at hundreds of kilometers per second. This is tough to model well, especially since it tends to evolve into a fractal structure.
- Use a disagreement between a linear and a cubic interpolation to decide where you need to insert extra points for higher resolution.
- Venus's atmosphere rotates faster than the planet, driven by solar illumination. Since the total angular momentum of the planet is conserved, people have observed seasonal variations in the rotation of the planet itself as the atmosphere gets spun up and spun down.
- People are STILL doing code development in Fortran 77. Argh.
- You can modify the way air flows over a wing by ionizing it with short sparks and exerting forces on it electrostatically.
- The formation of thin threads during droplet separation may be due to surfactant contamination in the liquid. This is odd since I thought prior results had simulated it with pure fluid properties.
- There are lots of interesting simulation techniques that I really should
have looked into over the years:
- Control Volume method
- Lattice Boltzmann Equation method
- Arbitrary Lagrangian-Eulerian method
- Front-tracking versus level sets
- (Dual Reciprocity) Boundary Element method
- Floating Random Walk method
All in all, it was a very full week of lectures, and greatly valuable to expand my picture of what possibilities are out there. I wish I could have taken the time to do this sort of unfocused survey when I was still at MIT.