Foam detaching and flying across screen

[jesse nicodemus]

Hello,

I have posted about this before but there has been no response. This afternoon I am working on getting approval to sent a playblast which will show the glitch.

When parent domain particles die or spread out too much during a splash the attached foam flies left to right across the screen in a uniform fashion. With a large enough splash the fly away chunks start to become a swarm. It seems clear what is happening but there aren't controls to fix it. My next step will be scripting the particles to die and convert over to splash if their velocity spikes on a single direction

This bug is almost making foam unstable. Is there a current fix for this or will it be addressed in an upcoming patch?

Thanks,

Jesse

[FlorianK]

Had a similar issue, recently. I had to increase substeps to solve the problem (yes, for Dumb foam...and I was also wondering why that was so.

[jesse nicodemus]

Hey Florian,

Glad to know that substeps fixes some of it. Currently kSpeed seems to make things a little better but it is hard to nail down the right speed and I know that will not work if my objects were moving much faster. That also makes foam pop off when I would rather it stay in place in the scene. Sending them a sample playblast. Really interested to hear what is making this happen.

[LuisMiguel]

Florian are right. I think the problem is not the foam but the Hybrido simulation. There is a graph in the demo scene kill_particles_isolated which can remove isolated particles. In this way you could remove the particles so the foam won't appear for those hybrido particles which are isolated. If you have the hybrido scene already simulated, you can modify this graph changing the getParticles node by LoadParticles node, and adding a save particle node to end of the graph.

I hope it helps.

Let me know if you need to know how to do that.

luisM.

[atena]

Hello,

Luis is right. I would like to ask you if that Foam was created "From splash" or "From domain", or maybe both. It seems to me that it was created "From domain", assuming that I would like to ask you if you change the "@ neighbors threshold" parameter. Foam is only created around a HY_Domain particle only when the number of neighbors is larger than this value. This can really help to remove those artifacts.

In any case I think that the root of the problem is, of course, in the simulation of the HY_Domain, if you have foam in those areas is because the core simulation was kind of exploding in those areas. The geo of the ship seems to be very complicated and for sure is not "water-tight", this is really bad for the solver because fluid might get trapped in between the pieces of the geo and the solver will be end up computing high pressures that will send particles flying away at high speed. To get rid of this I will advice you to create a "water-tight" proxy geometry of the ship (sorry if you already done it and I and assuming too much).

One more thing, in the case that your geo is "water-tight" and you still have isolated core particles flying away you can use the "Track lonely particles" parameter in the HY_Domain01. You can change the "@ min lifetime" and "@ max lifetime" to say for instance 0.0 and 0.02 so that lonely particles will be removed automatically as they show up during the simulation.

Please let me know more about the scene and we can help you better.

[jesse nicodemus]

Good morning,

Thank you for the information. I will look into some of those things.

The ship in the playblast is the full res model. For the sim I am using a low res version which is a full closed shape and specifically modeled on a way that hybrido will like. After a few tests we realized that hybrido responded much better if I gave it a volume. The ship is water tight. The base domain sim looks great and it seems to be very stable. I needed to crank the resolution to get the sim to account for some of the thinner parts.

I am using both foam and splash simming from the cached domain but the playblast only showed the elements coming from the foam emitter. The one thing I do have really low is my neighbors threshold. I will change that.

I will continue to try your suggestions on my end but it still seems like the consistent, predictable nature where the chunk of homeless particles lose form and reshape into a tight fist then fly away at a uniform speed and direction is a bug. If I was in Houdini I would think it was a cluster of particles around a detached metaball.

Thanks,

Jesse

[atena]

Hello,

I guess that it is a bug if there is no core particles in the center of those foam particles, can you check and confirm that?

Thank you.

[jesse nicodemus]

Hello,

I guess that it is a bug if there is no core particles in the center of those foam particles, can you check and confirm that?

Thank you.

I have been working to check all the blocks by trying everything that has been recommended. Currently running a new pass with substeps and killing the isolated domain particles with a much lower value.

I can confirm that the fly away foam does not have anything from the base domain connected with them. I also noticed while trying to get the k-speed daemon to work that the rouge particles have little to no velocity (.1-.3) even though they are the fastest moving particles in the scene.
Over all the base sim looks very stable. It doesn’t have the traditional things that I look for in a breaking sim.

We have the NDA from Gus so once we have the FPT info I am sending over a packet.

[atena]

Thank you very much. I will check that here and be back to this thread with the solution or the confirmation that is actually a bug.

Again, thank you for bringing this up.

[atena]

I’ve found the problem. Actually this is something I thought it was detected automatically by RF, but when I checked the source code I saw comments there so the situation at the end it is not detected.

What happens is that there are a couple of core particles that because of their distribution are creating distance field around them, the foam is placed on the surface of that distance field. At some point the same core particles have a different configuration in such a way that they not contribute to the distance field anymore. So suddenly the foam particles don’t know where to go, the distance field just disappeared. As I said this situation it detected and controlled (by removing those foam particles) in the code but unfortunately it was commented out and not taking any effect.

The workaround is to do that removal using a simulation graph. In your scene open the “Simulation Flow” panel by pressing “Ctrl+F2”, then in the “FramePost” simulation stage right click and select “Add Graph”, then go to the menu and select “File->Open”, open the attached simulation graph.

This graph is very easy to understand, it just use the distance field of the core simulation and use the position of the foam particles to sample it, when the foam particles are farther than some distance (this distance is now 0.3, you can change that in the “Real_01” node, I recommend to be the cell size of your core simulation) they are removed.