I'm having trouble with a basic rigid body sim (with only gravity force applied).

The “stone” pieces fall inside an invisible tank, so that they stay put and don't fall apart.
Some small narrow pieces, as shown in the images, tend to intersect.
How to avoid this?


I've tried to add more substeps in the ‘rigid body solver’ and to add some more collision padding in the ‘RBD packed Object’, but it didn't work out…

Hi there.
I had same problem before but for destructing a narrow wall.
Try to reduce collision padding and recheck the simulation.

thanks a lot Nima it worked!
Cheers

Dunno why I was trying to increase the padding, instead of reducing it, to fix the problem

ps.

there are still very few pieces that intersect (even with collision padding set to 0.0002, 1/1000th of the default value).

Is there any way to detect those intersecting pieces and delete them (all done during the simulation)?

What method are you using for your geometry representation?
That alone will clean up a lot of your intersecting geometry.

Uncheck the ‘show geometry’ and check ‘show guide geometry’ to view what your geometry looks like to Bullet.

You technically have concave hulls there as they are not square.
If you select Concave for the geometry representation, that can end up causing you to have intersecting geo.
Plus, it's going to slow things down too.

If you perform a for-each-loop through each piece of your geo and voronoi fracture it, then pack it you can select “Create Convex Hulls per Set of Connected Primitives” and that should help a lot.

This Lynda tutorial shows you how to do just that: https://www.lynda.com/Houdini-tutorials/Welcome/503880/573421-4.htm [www.lynda.com]

Jim

Penetration in regular simulation with bullet is normal because it designs to be as fast as possible however for your simulation you can increase substeps until you get desirable result (increasing too much causes unstable and long time calculation).
Also many pieces may penetrate because they really close to each other and bullet hasn't really chance to detect them at the beginning of the simulation.

forgot to say thanks for the tips you guys provided!
cheers