yade-dev team mailing list archive

[Václav Šmilauer <eudoxos@xxxxxxxx>]

> With _exact_ stiffness it would work only if all bodies in the cell were 
> not dynamic, i.e. almost never.
Why? Depends on your material law. If it is linear (or approximable by
linear function, i.e. differentiable), it can work very well, since
stiffness doesn't depend on strain change. Of course it neglects
appearing/disappearing contacts. But I don't see reason why it shouldn't
work to some extent in general case and very well in quasi-static
packing.
> When we use a "stiffness", we implicitly assume that the boundaries will 
> move while the solid inside is fixed (in other words, we say the 
> relative (boundary-body) displacement is equal to the displacement of 
> the boundary, i.e. bodies are fixed).
You compute stiffness from boundary stiffness in
TriaxialStressController, but PeriTriaxController computes it from all
interactions. The computed value will actually be even closer to real
(in terms of stress change as result of cell resize) stiffness with
homothetic displacement of bodies.
> This assumption is wrong even on one single step, but the algorithm works.
I'd be more pragmatic here.
> On more than one step, the chance that it will work tends to 0 I'm 
> afraid...
I used 5 and it was just fine.