← Back to team overview

yade-users team mailing list archive

Re: Some questions about YADE


> (1) When you are doing the finite difference approach for each step,

not sure if I understand the question.

if you are talking about finite difference method for time integration,
then just look into scene.xml, and you will see that we are using
leapfrog time integration (also known as verlet leapfrog method). But
bear in mind that yade is designed to be really flexible - adding new
time integration method is as simple, as writing one small class that
contains time-integrating formulas, and then changing appropriate class
name in scene.xml - and instatnly you are using a new method.

> e.g. 
> assume I am playing the SDECSpherePlane, put 5 balls each for x,y,z direction, 
> do you perform triangulation during each time step? Or do the triangulation 

what do you mean by triangulation? perhaps it is specific way for
contact detection method that you are familiar with?

> every several steps? Or you have better contact detect schemes?

to detect contacts currently we are using Sweep and Prune collider
method. But here also - if you want to detect contact with another
method - just write this method inside a class, and start using it by
changing class name in scene.xml

> (2) For the SDECImpact test, the automatic (critical) time stepper from the 
> example is quite small that at first I thought the simulation is not 
> working:), do you really choose the fraction of minimum period for all the 
> balls? I know larger time step will soon cause numerical stability.

elastic criterion is taken from some respectable book (I'll check its
title and author later). it is selected based on only existing contacts,
or if no contacts exists at all - then based on assumption as if the
spheres were in contact with themselves. This formula contains modifier
equal to 0.1, but it is hard to say that it's "0.1 of minimum period".

> (3) Do you have any termination criterion for the simulation? e.g. when you 
> are performing a triaxial test, there should be some kind of value (for 
> example, maximimum unbalanced force ratio as Itasca did) to evaluate the 
> termination?

there are some termnation criterions, and we are using them sometimes.
But each termination criterion is specific to performed simulation. So
if you need one - just write it, and put it in the engines list. (or
actors list, I think we didn't renamed everywhere from actors to

> I know these might be very easy basic questions, hope you can provide more 
> documents! Thanks a lot!

yep, we have some documentation, and we just need some time to make it
ready for the public. sorry - if you get what we have now - you would be
rather more confused...

Janek Kozicki                                                         |
Yade-users mailing list