yade-dev team mailing list archive

[Vincent Richefeu <richefeu@xxxxxxxxxxxxxxx>]

> De : Vincent Richefeu <richefeu@xxxxxxxxxxxxxxx>
> Date : 9 décembre 2009 09:35:42 HNEC
> À : boon chiaweng <booncw@xxxxxxxxxxx>
> Objet : Rép : [Yade-users] Questions on equations (contact laws, velocity, libraries)
> 
> 
> Le 9 déc. 2009 à 03:06, boon chiaweng a écrit :
> 
>> Q3. The position/velocity integration  follows the 2nd order explicit method, as was upgraded by Bruno :). The velocity and angular velocity are in the (t+dt/2) timestep, while the position & acceleration (also rotation) are in the (t) time step.   But it is not adjusted back to the current time step:  v(t) = v(t-dt/2) + v(t+dt/2).   Ommitting this will make the velocities in the simulation ahead of time by dt/2?  Is this alright? Why?
> 
> Hi boon,
> 
> I cannot answer all your questions, but only give an opinion for Q3.
> I think you're right. This may be critical point only for "viscous" damping. I think that something like the Beeman's algorithm should be a better candidate than leap-frog or velocity Verlet when viscous damping (with high velocities) is "activated". With quasistatic simulations, it seems not to be critical since viscous damping is a kind of "regularizator". With Cundall Damping, no problem since velocities are not required if the force-laws.
> 
> Vincent
> 
>