RBD fractured plaster layer

[Cuenk89]

Hi!

I have a RBD question.

 

I have a sculpture covered with a thin plaster layer, and this layer has to fall.

 

I have a sculpture as a concave collider, and the plaster layer as a concave fractured object. The problem is, during the simulation, some plaster pieces interpenetrating the sculpture.

In the initial state, the geometries (sculpture and plaster layer) don't overlap, but it seems that some plaster pieces don't respect the collision geometry.

 

I tried to subdivide all and increase substeps until 10 without solution.

 

Any ideas to fix this problem?

 

Thanks in advance!

 

PD: I attach a simplified example.

Prova_01.hip

[Atom]

I made several tweaks and clean ups and it seems to be working. I am not sure exactly which tweak was the fix.

 

I reset the substeps to 1, you only need 1 for most cases.

I cleaned up the fracture geometry. I removed the extrude from the collider, this may be what was causing the problem. A piece might have been sneaking through the extrusion amount. I also deleted the rigid body solver and create a new default one. I was not sure how many edits you had done to that node so it was my way to reset it to default. After that, playback seemed to work.

ap_Prova_01.hipnc

[Cuenk89]

Thank you so much Mr. Atom!

I tried with a setup like this but mysteriously didn't work...

I tried with many systems and methods with bad results, (for example give a normal force to collider object). I will try it on statue model, I hope it will work.

Thanks too for tips such as 1 substep is good, and use convex hull instead concave for fractured pieces.

[Cuenk89]

Hi again, it doesn't work on my model, still have interpenetration.

I made another simplified example with a sphere and mountain deformer.

Prova_04.hip

[Atom]

There are a few problems with your setup. The main problem is that your fracture pieces are inside the collider.

 

I rebuilt the scene but instead of importing the fracture into the collider I went the other way around and brought the collider into the fracture. This made more sense to me.

 

Because the skin needs to be an exact copy of the collider, only bigger I created a second chain in the collider network made up of Reference Copy nodes. A Reference Copy node is a copy of a node with expressions linking every parameter essentially making it an instance of the original node. Then I broke the reference only on the radius of the sphere. This way you can adjust the sphere1 radius to make it smaller than the reference copy shell that will contain the fracture pieces. One thing to keep in mind is that inside the fracture object extrude you can not exceed this radius reduction of the collision geometry or your pieces will interpenetrate the collider and you will start off in an immediate collision state.

ap_Prova_04.hipnc

Edited March 10, 2016 by Atom

[Cuenk89]

Thanks a lot for your time Atom!

I have a couple of questions:

-Why you use non packed object for mountain sphere?

-Why you use convex hull for collider instead concave?

Thanks again!

[Atom]

In the second example they are both convex. Typically concave is used if your colliding shape has an open area such as a bowl or cup.

 

I choose non-packed because I couldn't get Houdini to make the packed pieces initially. The tools set gets kind of finicky at time if you don't have the correct selection. In this example packed or non-packed does not really matter. Unpacking does take time especially on large piece counts. So it took me a few times, working through using the Tool Shelf icons, to get the simulation setup.

Edited March 12, 2016 by Atom