What steps will reproduce the problem?
1. Tension test with Courant factor are 0.5
2. Apply large strain (>50%)
3. The object will separate from the boundary conditions or from moving rigid
particles on the other end.
What is the expected output? What do you see instead?
The base boundary conditions should have zero velocity at the base of the
specimen and hold it in place. Instead, the object and boundary separate. In
some problems, the quality of these boundary conditions becomes the limiting
factor.
The results are improved by smaller Courant factors (e.g., 0.025), but then the
calculations are much slower. In other words, it seems like these rigid
boundaries are acting like a stiff material
Particles can also separate from moving rigid particles on the other end rather
then track them well. This problem may be caused by the boundary conditions
jumping for one grid line to the next as the rigid particle passes through the
cells. The jumping might induce kinetic energy that eventually causes the
separation. Running the same problem with traction BCs does not have this
separation problem.
Possible Solution
Try ghost BCs one cell back from the basic boundary line, as Unitah does at
`Symmetry` planes, to see if that changes the results. These ghost or mirror
BCs could be at a zero velocity plane or at the moving plane. Uintah has them
only for zero velocity planes. The process is to set velocity at nodes one cell
away from the moving boundary to be equal to line interpolation from node
within the object, through the boundary line, and to the mirrored node. The
basic code for this change is in NairnMPM, but needs more work before it can be
tried.
Original issue reported on code.google.com by johnanairn@gmail.com on 3 Jan 2013 at 1:01
Original issue reported on code.google.com by
johnanairn@gmail.comon 3 Jan 2013 at 1:01