Károly Zsolnai-Fehér explains how to simulate wet sand by tracking water saturation and cohesion between grains. Coarse grids cause the sand to expand over time, a problem the researchers fix either by increasing resolution or using a faster volume-correction method.
Dear fellow scholars, this is two minute papers with after around 160 episodes into two minute papers. I think it is no secret to anyone that I am helplessly addicted to fluid simulations. So you can already guess what this episode is going to be about. I bet you will be as spellbound by this beautiful footage of wet sand simulations as I was when I first seen it. Before you ask, yes, I have attempted to prepare some slow motion action too. As you remember, simulating the motion of fluids involves solving equations that tell us how the velocity and the pressure evolves in time. Now, the 3D world we live in is a continuum, and we cannot solve these quantities everywhere because that would take an infinite amount of time. To alleviate this, we can put a grid in our virtual world and obtain these quantities only in these grid points. The higher the resolution the grid is, the more realistic the animations are. But the computation time also scales quite poorly. It is really not a surprise that we have barely seen any wet sand simulations in the visual effects industry so far. Here we have an efficient algorithm to handle these cases. And as you will see, this is not only extremely expensive to compute, but nasty stability issues also arise. Have a look at this example here. These are sand simulations with different cohesion values. Cohesion means the strength of intermolecular forces that hold the material together. The higher cohesion is, the harder it is to break the sand up, the bigger the clumps are. This is an important quantity for our simulation because the higher the water saturation of this block of sand, the more cohesive it is. Now, if we try to simulate this effect with traditional techniques on a coarse grid, we'll encounter a weird phenomenon. Namely, the longer our simulation runs, the larger the volume of the sand becomes. An excellent way to demonstrate this phenomenon is using these hourglasses, where you can clearly see that after only a good couple turns, the amount of sand within is significantly increased. This is particularly interesting because normally in classical fluid simulations, if our grid resolution is insufficient, we typically encounter water volume dissipation, which means that the total amount of mass in the simulation decreases over time. Here we have the exact opposite. Like in a magic trick, after every turn, the volume gets inflated. That's a really peculiar and no less challenging problem. This issue can be alleviated by using a finer grid, which is, as we know, extremely costly to compute. Or the authors proposed a volume fixing method to take care of this without significantly increasing the execution time of the algorithm. Make sure to have a look at the paper, which is certainly my kind of paper. Lots of beautiful physics and a study on how to solve these equations so that we can obtain an efficient wet sand simulator. And also, don't forget a fluid paper a day keeps the obsessions away. In the meantime, a word about the two minute paper shirts. I am always delighted to see you fellow scholars sending over photos of yourselves proudly posing with your newly obtained shirts for the series. Thanks so much and please keep them coming. They are available through twominutepapers.com for the US and the EU and worldwide. Link is also available in the video description. Thanks for watching and for your generous support and I'll see you next time.
Sam.
