[Sidefx-houdini-list] FLIP'pin out

Nick Zutphen nvanzutphen at live.com
Fri Aug 5 05:25:15 EDT 2011


Hi Craig,

To refer to your question about ending up with less fluid than expected, which you can also define as 'volume loss'. Since I'm not familiar with your knowledge about the inner workings of the FLIP solver, a short intro / reminder: To get around the instability of SPH the particle's velocity get projected on to a volume field to handle the non divergence which is projected back to the particles for advection. In order to get the speed advantages the resolution of the volume field is drastically lower that the particle count or resolution. Think of an average of 8 particles per voxel. So each voxel adjusts multiple particles at once. 

What happens with 'volume loss' is that individual particles get to close together, which results in the problem where the volume non divergence is not able to push them apart any more and the fluid 'collapses' a bit every step, since it controls multiple particles at once instead of individual particles. To get around this, particle separation is used in an attempt to separate individual particles. This works until a certain extend, the fact that this hasn't been resolved lead into the advice to use FLIP for rivers, waterfalls and that sort of effects where the volume loss is less of an issue.

At a certain point I've attempted to create my own particle separator, which turned out to be rather useless since I was heading towards a similar solution as the existing one but  slower since I wasn't using C++. (a typical case of: 'you're young and you think you can do better that a very experienced developer' ;-)

A hack, which also has it's own pitfalls, but which seemed to work in some cases is to calculate your own divergence attribute. By doing a point cloud lookup on the particles you can determ the distance to it's closest neighbor, if you store this on the particles in a -1 to 1 range and toggle 'diverge by attribute' on the solver you can force the non-divergence to actively push fluid apart which gets to close together. The pitfall is that you add energy in to the mix, which can result in a restless fluid and you'll hit the issue of streaming fluid along collision edges.

I guess one can conclude that fluids aren't that easy, my advice is either live with it, use SPH or attempt to get the divergence volume separation to work. Alternatively you can try to increase the resolution of the volume fields by adjusting the 'division size' expression on the FLIP Configure object, but this will give you a speed penalty and I'm not sure if it will work in the first place since I haven't tested it my self.

Cheers,
Nick van Zutphen

----
http://www.nickvanzutphen.com









On Aug 5, 2011, at 3:21 AM, Craig Hoffman wrote:

> 
> 
> 
> Hi all,
> 
> I am creating an animation of a mixed alcoholic drink with multiple different colored liquids pouring into a glass (one at a time) and am using FLIP fluids and everything is working okay.  I am getting pretty good results fairly quickly but can't quite get that last bit to make it feel really realistic, so I have a couple questions:
> 
> 
> The amount of liquid that ends up in the glass seems like a lot less than what is poured in- like around half to my eyes.  Maybe it is just an illusion, but I am just going off experience with liquids in real life and how much I expect to be in a glass after a nice thick strong stream pouring in for a couple seconds.  Is there a way to fix/cheat this so that I get a result more like what I expect?  (Didn't think about animating 'particle separation' until now- and I am away from my PC, so haven't tried it yet- though I am not sure if that would work or be stable, etc.)
> 
> The liquid is a little too "active" with individual particles going off on their own, etc. and it takes a little too long to settle down.  I started with "0" for both bounce and friction for the liquid and played around with friction settings for a while, but haven't found a setting that makes it really "feel" like real liquid.  Should I venture into viscosity to settle it down quicker- or is that more for thicker fluids than water or alcohol?  
> 
> Is there any reference for good FLIP fluid settings for real-world fluids (as well as 'collision object' settings for glass/ice/etc.)?
> Lastly, can anyone drop me a hint if I should just table this project until after Siggraph (assuming H12 releases then..) because all of this is so drastically improved/sped up/revolutionized/etc.?   :)
> 
> Any help or suggestions are greatly appreciated.
> 
> Thanks,
> Craig Hoffman
> 
> 
> 
> _______________________________________________
> Sidefx-houdini-list mailing list
> Sidefx-houdini-list at sidefx.com
> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
> 




More information about the Sidefx-houdini-list mailing list