Collision Volume

[Hans Peter]

Hi there, 

I've got some strange problems with the collision volumes in flip. The static object itself shows a clean, closed volume; the flip object on the other hand shows me a mess. Nothing is closed, the collision volume seems to be made of stripes. How does this discrepancy between static object and flip object happen regarding collisions? Actually the flip particles are floating inside the object. Got a new ship model and obviously some diffuculties making it work

I also tried to use for each loops to convert each single object the shi is made of into volumes in order to merge them together insteat of turning the whole thing into a volume at once. 

[Atom]

I have run into this as well. You can de-couple FLIP collision resolution from the particle separation. On the FLIP Object node activate Override Collision Separation. This allows you to use a lower value for collision and still keep your particle separation for the fluid.

Also, if your collision object is in direct contact with the fluid, at the start, you may want to subtract your collision shape from the fluid emitter before you begin simming. You can do this using a cookie to subtract the collision object from the emitter source. Basically you don't want any fluid particle inside the collision volume before the simulation begins.

[Hans Peter]

Unfortunately that doesn't help. The collision volumes the flip object is creating are a mess. Second image shows the collision guide for the static object, the first one for the flip object. 

[Hans Peter]

Thats the vdb collision cvolume I created

[Guest tar]

Try this- use the deforming object shelf, write out and load the file cache, turn off Use Volume Based Collision Detection on the Static Object. Ignore the FlipFluid collision field. 

Edited January 15, 2017 by tar

[Atom]

Try supplying your nice VDB conversion as the Proxy Volume under the Collision TAB of the Static Object inside the AutoDopNetwork.

 

You may be facing a scale issue. If your fluid volume is too large you may never be able to lower the particle separation low enough. Tighten up the FluidSolvers bounding area as close as possible to force more particles into a smaller area.

 

Can you post an example scene?

[Hans Peter]

I actually did this proxy thing. Its so strange, because I used the same setup for an other equal sized ship model. I don't know why flip isn't willing to cooperate this time. I'll give it a try with the deforming object.

Edited January 15, 2017 by Heraklit

[Guest tar]

whatever works IMO - just test different setups for optimal speed and efficiency. Not sure why the collision guide has come back into vogue - old tech really.

Edited January 15, 2017 by tar

[Hans Peter]

Well, its quite fast and I get very detailed collision guides without tons of polygons. Also the guide itself only needs to get calculated once not every frame. 

[Guest tar]

so whats the issue then? 

[Hans Peter]

Its not working this time

I really don't know why the flip object is producing such shitty results when the volume itself is pretty good. 

 

[Guest tar]

yeah - the steps I wrote above *should* use the vdb perfectly. It has on all my tests whilst trying to use el-crappo volume guide only causes slower processing and lesser quality.

[Hans Peter]

AAh, its not working. No matter wheter I use the surface polygons for collisions or any kind of volume. 

 

 

Edited January 15, 2017 by Heraklit

[Hans Peter]

So here is the collision volume. Maybe you guys are able to make something usefull out of it. 

coll.rar

Edited January 15, 2017 by Heraklit

[Guest tar]

Not sure how your setup is but I'm seeing flip collisions straight away.

Boat.hiplc

[Hans Peter]

Strange, I tried to load the saved version too and it seems to work better. It actually should be the exact same volume. 

[rbowden]

No reason why the saved version of the bgeo and the live version vdb creation shouldn't work. Whenever you run into stuff like this and you know it should be working, why not go back to the basics and create a simple scene to test it out? If the simple scene doesn't work for you then upload that here. It is really hard to debug FLIP simulation issues by just looking at a screenshot.

Out of all the times I have done FLIP simulations, I have never used collision guide either. I am guessing the resurgence of this is from the recent Jeff Wagner webinar. Its not awful if you want to go that route, I just don't see the point. If I want to know what my collision is going to look like, I take a look at it on my static object. That gives me the real representation that I am looking for.

[tamagochy]

It is better to use simple hull model of the ship without tiny edges. It save memory and time of the simulation.

[Hans Peter]

I'm using this technique for more than a year now and it actually works really good. I don't want a simple hull representation of the ship. That might work for a speed boat or other simples shaped objects but when you got a lot of detail in front of a ship and you make a close up shot it just looks better when the water leaks through fine structures. Or do you know a better way of acieving this? I think to use polygon collision is even worse. 

"No reason why the saved version of the bgeo and the live version vdb creation shouldn't work." Well, I know there is no logical reason but thats how it is atm When I use the live vdb version the flip collision gets mad. The saved one miracilly works totally fine. The whole sim is already done without any problems.

[Guest tar]

keen to see the end results!