Dr. Zsolnai shows a new method for simulating fluids and solids together, handling extreme contact like spinning propellers and draining water. It processes up to 38 million particles and runs 10 to 100 times faster than older approaches, with some simulations reaching 58× speedup.
160 RPM
Up to 38 million fluid particles
More than 30 times smaller time steps were required with the previous method
Typically from 10× to 100× larger time steps
Up to 58× faster
Dear fellow scholars, this is two minute papers with Dr. Zsolnai. In this paper you will not only see an amazing technique for two way coupled fluid solid simulations, but you will also see some of the most creative demonstrations of this new method I've seen in a while. But first things first, what is this two way coupling thing? Two way coupling in a fluid simulation means being able to process contact. You see, the armadillos can collide with the fluids, but the fluids are also allowed to move the armadillos. And this new work can compute this kind of contact. And by this I mean lots and lots of contact. But one of the most important lessons of this paper is that we don't necessarily need a scene this crazy to be dominated by two way coupling. Have a look at this experiment with these adorable duckies. And below the propellers are starting up. Please don't be one of those graphics papers. Okay, good, we dodged this one. So the propellers are not here to dismember things. They are here to spin up to 160 RPM. And since they are two way coupled, they pump water from one tank to the other, raising the water levels, allowing the duckies to pass. An excellent demonstration of a proper algorithm that can compute two way coupling really well. And simulating this scene is much, much more challenging than we might think. Why is that? Note that the speed of the propellers is quite high, which is a huge challenge to previous methods. If we wish to complete the simulation in a reasonable amount of time, it simulates the interaction incorrectly and no ducks can pass. The new technique can simulate this correctly. And not only that, but it is also four and a half times faster than the previous method. Also check out this elegant demonstration of two way coupling. We start slowly unscrewing this bolt and nothing too crazy going on here. However, look, we have tiny cutouts in the bolt allowing the water to start gushing out. The pipe was made transparent so we can track the water levels slowly decreasing. And finally, when the bolt falls out, we get some more two way coupling action with the water. Once more. Such a beautiful demonstration of a difficult assimilate phenomenon. Loving it. A traditional technique cannot simulate this properly unless we add a lot of extra computation, at which point it is still unstable.
Ouch.
And even with more extra computation, we can finally do this. But hold on to your papers because the new proposed technique can do it about 10 times faster. It also supports contact rich geometry as well. Look, we have a great deal going on here. You are seeing up to 38 million fluid particles interacting with these walls given with lots of rich geometry and there will be interactions with mud and elastic trees as well. This can really do them all. And did you notice that throughout this video we saw a lot of delta T's. What are those? Delta T is something that we call time step size. The smaller this number is, the tinier the time steps with which we can advance the simulation when computing every interaction and hence the more steps there are to compute in simpler words. Generally time step size is an important factor in the computation time, and the smaller this is, the slower but more accurate the simulation will be. This is why we needed to reduce the time steps by more than 30 times to get a stable simulation here with the previous method. And this paper proposes a technique that can get away with time steps that are typically from 10 times to a hundred times larger than previous methods and it is still stable. That is an incredible achievement. So what does that mean in a practical case? Well, hold on to your papers because this means that it is up to 58 times faster than previous methods. 58 times? Whoa. With the previous method I would need to run something for nearly two months and the new method would be able to compute the same within a day. Witchcraft, I'm telling you. What a time to be alive. Also, as usual, I couldn't resist creating a slow motion version of some of these videos. So if this is something that you wish to see, make sure to visit our Instagram page in the video description.
For more this episode has been supported by Lamda GPU Cloud. If you're looking for inexpensive cloud GPUs for AI, check out Lambda GPU Cloud. They've recently launched Quadro RTX 6000, RTX 8000 and V100 instances. And hold on to your papers because Lambda GPU Cloud can cost less than half of AWS and Azure plus they are the only cloud service with 48 gigabyte of RTX 8000. Join researchers at organizations like Apple, MIT and Caltech in using Lambda Cloud instances, workstations or servers. Make sure to go to lambdalabs.compapers to sign up for one of their amazing GPU instances today. Our thanks to Lambda for their long standing support and for helping us make.
Better videos for you. Thanks for watching and for your generous support and I'll see you next time.
