Hello, hello. One and all…

I have a shape (vase-like) that i'm attempting to fill with a fluid.

Using volume-based collision detection (including the laser scan), the sim works fine but i can't actually get the collision guide geometry to accurately represent the volume of my shape.

I turned off ‘Use Volume Based Collision Detection’ in the rbd object hoping it was would try to calculate the collision of the particle with my actual geo. No such luck. For the most part, my particles fall straight through.


So in summary I have two questions. The answers of either could help my sorrow:

1) Is it possible to manually tell an rbd object what you want the collision guide geo to be?

2) Upon turning off ‘Use Volume Based Collision Detection’ in the rbd object, is there any other way to increase collision accuracy?


Thank you, you kings.

-Paperfish

Hi,

Have you gone through the dynamics tutorials? If not, please go to the Learning menu at the top of the page & click on Tutorials. Download the “Dynamics Quickstart” pdf & files to learn how to create proper volume representation of your vase.

There are also Houdini 9 Dynamics tutorials available as well.

Good luck & have fun!

Cheers!
steven

Hi Steven,

I had a peek at that dynamics pdf. I'm already aware of a lot of stuff that the paper mentions.

My main problem revolves around it not creating volumes that are accurate enough for my custom geo, despite tweaking.

And, like I said, the non-volume based collisions appear not to work.

So I'm afraid my two questions still stand.

-Paperfish

Please post a hip file with the vase.

This hip file was sent to kleer001 privately, and no solution was obtained.

Non volume-based collisions appear to not work, so the only place to keep experimenting was with the volume collision guide.

I found I had to take the external surface off my object, and boolean the inside with a cube. This produced a reasonably accurate collision guide (see image01).

The settings for this volume can be seen in image02.

As soon as i raise any of the divisions to get more detail, the volume seemed to explode (see image03). In this instance the y element of the division was just raised to 35, and boom… unusable volume.

Remaining questions:

1) Can I create a better (custom?) volume?
2) Is there anything I need to do to get collisions working when I untick “Use Volume Based Collision Detection”

are the normals on your vase pointing in the right direction?

Yes, the normals are facing in the right direction. You'll see in image01 that the volume it generates is actually resonably correct, i just need better detail in the more detailed areas of the model.

The SDFs generated by Houdini very often “leak” into big blocks hanging outside your geometry, unfortunately. Often deactivating Laser Scan makes it worse too and your only solution is to +1 to your SDF resolution.

I'd say this happens to me about 1 time in 5, making it essential to always check your Collision Geometry. Invariably you have to bump up the SDF resolution to 30x30x30 (which seems like a better default to me) and then perhaps knock it up to 31x31x31 should you see leaks.

This is using 9.0.761. I'd hope someone could bulletproof the SDF building algorithm a little in the future and bump up the default resolution to at least 30s. Perhaps if the SDF builder detects flooding all the way to domain boundary, it should be a warning sign and it should go back and jitter or bias its samples?


Cheers,
Jason

PS. SDF's are the principal geometry type for many tools at DD and R+H (not just simulation these days) and I've seen reliable SDFs built in custom tools.

Detecting blocks hitting the outside is what the Force Bounds option helps avoid.

I think a solution to the hanging-blocks problem was added in 9.0.754. However, it does not take effect by default nor is it available as an option in SIM_SDF.

On the IsoOffset SOP there are some otions to control “sign sweeps” that try and clean up those big blocks of bad geometry.