yade-users team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

kan a écrit :
> Hey, guys,
>  
> I have a problem to get the critical timestep, could you please give 
> me hand?
> if I need to calculate the critical time step for each time steps, how 
> can I calculate it?
> From papers, I found that equation is:
> critical_time_step=2*sqrt(mass/k),
>  
> my questions are:
> 1. is the mass the particle mass?
Its the generalized mass. i.e. one mass for each degree of liberty : 
actual mass for 3 translation axis, and inertia matrix for the 3 
rotation axis. Its a bit difficult to uncouple rotational degrees of 
liberty because corresponding "mass" is a matrix...
Cundall proposed to use diagonal terms as an approximation, its what I 
did in GlobalStiffnessTimestepper.

> 2. k here is kn or ks or some other value else?

Again generalized stiffness : a matrix including translational and 
rotational stiffnesses (taking into account all contacts on bodies). And 
again I used the diagonal terms to associate a scalar stiffness to each 
degree of liberty.

> 3. I need to calculate all the critical_time_step for each particle, 
> and then use the mininum one, right?
yes
>  
> and I hear that in the past, the critical_time_step need to calculated 
> at each timestep, then why and how?
Contact created/deleted = stiffness changed.

> if I do it at the beginning and get the mininum critical_time_step, 
> and then keep it as the critical time step for the overall simulation, 
> will it work?
No, unless you set big security margin, which would mean a waste of time.

Bruno

-- 
 
_______________
Chareyre Bruno
Maitre de conference

Institut National Polytechnique de Grenoble
Laboratoire 3S (Soils Solids Structures) - bureau E145
BP 53 - 38041, Grenoble cedex 9 - France
Tél : 33 4 56 52 86 21
Fax : 33 4 76 82 70 43
________________

_______________________________________________
Yade-users mailing list
Yade-users@xxxxxxxxxxxxxxxx
https://lists.berlios.de/mailman/listinfo/yade-users