yade-dev team mailing list archive

[Václav Šmilauer <eudoxos@xxxxxxxx>]

> One remark : implementing this for arbitrary deformation needs 9 
> "goals", compared to 9 components of F or stressTensor. A typical 
> example : compression along z axis with imposed dεzz/dt while σxz is 
> kept constant (i.e. velgGradxz will be assigned by the engine to keep 
> sxz constant, and velGradzz will be defined to make log(Fzz) reach goalzz ).
I think just 6 goals will do, since shear stresses will be (in
equilibirum) symmetric.

> One question : how can it be adapted to isotropic compression, where 
> goal is (sx+sy+sz)/3 and the constraint is ex=ey=ez?
Too many constraints, the system is overdetermined. You would just need
ex=ey && ex=ez && sx=sig/3. Plus one condition for strain rate, which is
the fourth, time DOF.

I actually wanted to have one big system (matrix) and then eliminate
columns corresponding to given goals (which would move to RHS) etc, but
that will need incremental formulation, and a few thoughts how to handle
absolute goals in that case...


> This is why I implemented cell inertia. Did you fix this or it was not 
> really a problem?
I didn't fix it, but inertia has nothing to do (?) with material
stiffness? It is just "dynamic" stiffness.

> If you apply a (non-isotropic) scaling on normalized Hsize vectors, it 
> will "shear" them (the angles between axis will change), right?
Yes, unless I misunderstand what you mean ;-)

v