yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

>  if the time step is ok without viscous damping, it will necessary be ok with the damping. 
Actually not, or you are thinking of stiffness and viscosity in series?
If they are in parallel, dt will have to be smaller.

>  Anyway it's million times better than the so called Cundall damping which must be proscribed (in my own opinion) when the inertial number is to high (say > 10e-4 ?).
If realistic dynamic is required in one's problem, then Cundall damping
doesn't apply since it doesn't reflect true dynamics.
If quasistatic problems are simulated, or if transient dynamics is not
important (pluviation), then Cundall is better, since it is numerically
more efficient and it doesn't change the timestep.
> I don't know your topic, but if you want to dissipate a certain rate of energy (defined as the ratio of energy = e_n^2) you could simply multiply the force (even non-linear with respect to the overlap) by this rate during unloading...
> Think about it...
If (1) viscosity is just a numerical trick to damp the equations of
motion with elastic contacts, then it results in wrong dynamics exactly
as with Cundall's damping (then better use Cundall). The same remark
applies to multiplication of contact forces by an artificial dissipation
rate.

If (2) visco-elastic contacts are supposed to reflect the realistic
behaviour, then there is a contact _viscosity_ resulting from material
behaviour, and this viscosity will not be tweaked just to give a certain
dissipation or over-relaxation.

I was assuming (2) in this discussion, because if the question of
time-step is solved for any user-defined viscosity, then it will also
work for the special cases in (1) (e.g. viscosity=critical viscosity).
It's better if we have dt determination in the most general case.

Bruno