aperiodic randomDensePack huge overlaps and strange behavior

Bug #1666838 reported by Jan Stránský
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Yade
New
Undecided
Unassigned

Bug Description

Based on question https://answers.launchpad.net/yade/+question/473518, I did some simple tests using Yade 1.20.0 on Ubuntu 16.04

########################
from yade import pack
pred=pack.inAlignedBox((0,0,0),(2,1,1))
spheres=pack.randomDensePack(pred, radius=0.1)
O.bodies.append(spheres)
O.step()
print max(i.geom.penetrationDepth for i in O.interactions)
O.bodies.append(box((1,.5,.5),(1,.5,.5),wire=True))
########################

I got:
- packing smaller than predicate
- huge overlaps: r=0.1, max penetrationDepth=0.09 (!)

playing with cropLayers and dim parameters did not help much (especially concerning the overlaps).

Am I missing something, or is the default behavior that bad?

cheers
Jan

Revision history for this message
Bruno Chareyre (bruno-chareyre) wrote :

You are speaking of the non-periodic case, aren't you?
It seems it is using the pre-processor "TriaxialTest", I can imagine that this old lady is not giving exactly what it was years ago, then maybe the bad results you get.
It should not be too difficult to test what the curretn code is doing, and fix or replace by a more recent one inspired by some triaxial simulations from example folders.
Bruno

Revision history for this message
Janek Kozicki (cosurgi) wrote :
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers