Debugging FEM self-collision

   3030   5   0
User Avatar
Member
18 posts
Joined: Jan. 2009
Offline
Created a simple squab, and used the shelf tools to make an Organic Solid. Turned on Collide within connected components on the Solid Object in addition to the default collision options, and added a ground plane. When I simulate, the middle tentacles settle intersected starting at about frame 40. See the attached screenshot, in case your mileage varies.

I'm expecting the tentacles to stay separated. What am I missing here to make the mesh self-collide correctly?

Houdini FX Version 16.0.621, Mac Pro 2013, Sierra 10.12.5.

Attachments:
Screen Shot 2017-05-30 at 4.07.57 PM.png (1.7 MB)
SelfIntersectingSquab.hipnc (289.0 KB)

User Avatar
Member
46 posts
Joined: Oct. 2015
Offline
Hey glensandness, got the same issue with the pig head
the snout not colliding with the lower jaw but passing through.
Used it last in one of the H15 builds and it worked off the shelf,
if I remember correctly.
The reason is beyond me, but if You are looking for a fast workaround (and aren´t already aware of it)
You may consider using Granular Solid on the Grains tab - there are some
differences to FEM according to the help docs but the results look pretty similar.
(will try to attach my example)
Good Luck +
cheers!
Edited by illusionistics - May 30, 2017 17:22:09

Attachments:
jollyPigsPBD.gif (1.4 MB)

…just
…evolving
User Avatar
Staff
429 posts
Joined: June 2007
Offline
To get collisions resolved properly, you may want to increase the Collision Passes from its default of 2 to something higher, say 8, in this case.

I do find it undesirable and a bit suspect that the collision passes needs to be increased to get this example to work. I will look into this. I've added bug 83037 for this. There will be a version log entry as soon as this get fixed or improved.
User Avatar
Member
18 posts
Joined: Jan. 2009
Offline
Thanks for the replies.

Illusionistics, thanks for the tip. I'll find a use for the Granular Solid at some point.

Michiel, I was able to get the squab example working with more collision passes, as you suggested. I have a more complex model that started all this, too unwieldy to post, in which bumping the passes as high as 16 isn't helping. No sign the self-intersection is even being detected by the solver. Is there a practical upper limit on the collision passes, beyond which we know that's not the problem?
User Avatar
Staff
429 posts
Joined: June 2007
Offline
Thank you for trying this out.

Houdini 16.0.626 has a fix for the large number of collision passes needed.
The squab example should have self collisions working fine with just 2 collision passes on the solver.

Would it be possible to try your more complex example with Houdini 16.0.626?
It would be great to know if this issue is now resolved, or if there is still something else going on.
User Avatar
Member
18 posts
Joined: Jan. 2009
Offline
16.0.626 does work well for me for the squab example. It also made a big difference in my more complex example, although I have a lot of tweaking to do before I'm happy with the FEM behavior there. Self-collision is back to working off-the-shelf, as illusionistics put it.

Thanks for your help, Michiel.
  • Quick Links