Particle (Whitewater) Collision Penetration

[gpapaioa]

Hello everyone,

 

I am doing a whitewater sim and want the partcles to collide with geometry.

 

I am using VDBFromPolygons to create a Distance Volume (SDF) -> Inside the whitewater sim I merge a static object and use the SDF as VolumeProxy.

The collision works OK but I have particle penetratng the geometry in some areas.

 

Any tips for making it more accurate or change method of collision? Is there a way to check the penetrating areas and see what is going on?

I put more substeps on the whitewater sim but with no result.

Thank you.

Edited February 8, 2014 by gpapaioa

[gpapaioa]

Found out that I didnt setup the Static Object correctly. Now it seems to work but its dead slow! Any tips on improoving performance? Cache out the SDF in bgeo?

[gpapaioa]

Still having penetration. Here is a screenshot. Particles are black due to popcollisiondetect node

 

 

[reny]

I assume that you don't have "trail" node on your collision object SOP.

Attach one before the output and have it to "compute velocity", so your SOP goes into VOP with velocity that pushes particles away more properly.

 

Oh, and for your collision obj you can actually see if your guide geo(which should already be a SDF volume) is above/below actual geo in DOP network.

Edited February 9, 2014 by reny

[gpapaioa]

Hello Reny

 

I did have a trail for computing velocity. I have tried every different way and ended up using a popcollisiondetect with SOP and Killing the particles that hit.

 

If i understand correctly about the guide geo.. yes I made a very detailed SDF and checked the collision guide. Its nice and blue and covers the actual geometry.

 

I dont know, maybe the pipe is two thin to calculate the collision with too much velocity.

[reny]

ahhh....high velocity could be the problem. well, the default auto substep can only do so much about detecting changes.

If everything fails, you just need to increase the max substeps(lazy but slow), or globally more substeps on dop network.

 

Or if you can do some wiring and caching, you can cache the SOP animation out, and read in the "future frame",

 and preemptively push the particles with POPVOP with velocity enough to avoid being "inside" the dropping pipe(use SDF from the future frame geo), the draw back of this it might not look as accurate.

 

Also, you can have high enough SDF as collision source and just a tiny bit of offset so it's not so slow compare to just really high detailed SDF.

[gpapaioa]

Thank you Reny for the tips. I also found that it helped to cache the whitewater source (the main FLIP was already cached) and then let whitewater solver do its thing.