Why Doesn't FLIP Collision Work?

[Atom]

Hi All,

I am trying to make my characters foot splash in the water. I have followed the setup steps for deforming geometry in my FLIP sim and I can see, in BLUE, that my collision mesh is represented ok. However, I get no collision. When I turn on the guide for collision on the FLIP object, I can see why, in RED. The RED mesh is what FLIP is actually using to collide with and you can see in the image that it is too low of a resolution to correctly collide.

I can't find any parameter to increase the resolution of the RED mesh.

Why doesn't FLIP just use the BLUE mesh, I mean what is it for anyway?

This work flow does not make sense to me. Why is there a secret collision mesh, that is hidden by default, that I can not even configure?

[rbowden]

The red mesh is just from your collision visualization on the flip object. The default settings are probably a little bit low res. If you change anything on that mesh, it doesn't change your simulation at all. It is just for visualization after all...If you want to take a look at the settings for it, go about 4 tabs over to the collision tab. All you need to worry about is how your collision looks on your static objects dop (your blue mesh).

Also, is the water going to be hitting the whole body? If not, why are you bringing in the whole body as a collision? You can cut off everything but just the legs if you want just a foot splashing into the water. 

[Atom]

Ok, then why doesn't my collision work as is? The BLUE mesh has a solid and strong collision representation.

I watched Jeff Wagner's FLIP webinar and he mentions that all FLIP can see is the Collision surface, which is represented by the FlipObject visualization which is RED. I can increase the RED resolution by lowering Particle Separation but if I go low enough to get a solid RED representation I am faced with impossible sim times and Houdini crashing because I only have 32GB. I was able to "Beef" it up a little by subdividing the collision source.

[blaykenadeau]

Hey,

I think you can unlink the collision separation from the particle separation.

[Sepu]

did you try "use volume based collision? as Vdb then using the proxy volume? 

[rbowden]

I can't answer that just by looking at the picture of your network. I could take wild guesses but, you should just upload your file.

[Guest tar]

Not sure why @Atom doesn't' just create test files and then compare what works and doesn't. I had a collision test file working in <3min.

[Atom]

Thanks for all the feedback.

A test file is not really feasible at this point because of the animated character. I can't supply that at this time. And yes, a simple sphere dropping in a fluid works just fine. But I am trying to push beyond the beginner level. I am trying to discover why it does not work so I can make it work.

@seup: Yes, I did construct a VDB version of my collision mesh and referenced that in the Proxy field of the Static Object but that does not change the resolution of the RED geometry. The static object node can already create a volume based representation of the source geometry so I am just using that to create the BLUE collision volume.

@Blayke: Thank you, that was what I was looking for. Activating and lowering the Collision Separation does increase the resolution of the RED collision surface. So I am getting some collision but the reaction is very tame. I have added a Trail in, compute velocity mode, to my deforming geometry to introduce velocity but it does not seem to help much. I want a stomping splash.

I have cookied my character with the emitter geometry to create gaps where the character is standing in the water. Jeff mentioned in the webinar that this helps keep particles from being instantly in collision by removing them from the interior of the collision mesh.

However, I still have a problem with the creation of the emitter. The resulting particles extend beyond the emitter boundaries. Notice I have a thin grid as my emitter. But the Fluid Source node makes this giant blob of white particles around my emitter.

How do I get the Fluid Source node to honor the shape of my emitter? I do really want a thin sheet of fluid to splash around in, think puddle not pond. I have tried lowering and raising the scale of the IsoSurface but all values produce the same oversized result.

Edited January 6, 2017 by Atom

[blaykenadeau]

The way I've fought against blobby emitters that don't match the intended shape of my emitter geo, is to use the sdf from geometry tab. You can pull down Edge Location to try to better match the shape.

You might also be able to use a Points from Volume node and bypass your fluid source. You can specify input type in the Flip Object (Particle Field, for Points from Volume) and then plug in the sop path. I don't know if there are obvious downsides to doing this.

In terms of getting more exaggerated water reactions from your character; What about multiplying the velocity or adjusting Velocity Scale under Collisions on the Flip Solver.

Edited January 6, 2017 by blaykenadeau

[Atom]

Thanks everyone for all the help. I think I finally have a starting workflow.

Instead of creating an emitter I use fluid from object to create the stomping pool. This keeps the shape of the original object but the flip solver has no sourcing in this case. If the character starts out in direct contact with the fluid use a cookie to cut a hole in the emitter to limit any initial fluid particle penetration with the collision object.

Also don't use the Deforming Object button use the Static Object button then remember to click Deforming Geometry checkbox inside the Static Node created inside the AutoDopNetwork. On the Static Object node enable Volume Base Collision and raise the divisions from 30 to 96. This gets a fair representation of my character.

If I raise the Velocity Scale on the FLIP solver it affects the entire fluid grid, instead I dropped down a Trail on my deforming geometry and increase the velocity scale on the points before they reach the solver. In this GIF the point velocity scale is set to 10.