yade-users team mailing list archive

[Jérôme Duriez <question672525@xxxxxxxxxxxxxxxxxxxxx>]

Question #672525 on Yade changed:
https://answers.launchpad.net/yade/+question/672525

    Status: Open => Answered

Jérôme Duriez proposed the following answer:
Hello,

The good thing is that these results seem completely consistent with the
doc at https://yade-dem.org/doc/formulation.html#motion-integration.

Motion integration in DEM in general and YADE in particular involves a
finite difference scheme, which is necessarily a discrete (in time)
approximation of true phenomena being continuous in time.

Thus, like with any other finite difference scheme, the precision of our
(classical ?) scheme with respect to theoretical equations directly
depends on the size of the time increment (compared with some time scale
for the study).

Here, your time scale is pretty much the same as your time increment,
hence you can not expect to be satisfied with your results.

On the other hand, you would get a 1% precision if you would do the
comparison at t = 100 = 100 * O.dt (or at t = 1, but using O.dt = 0.01)

Note that during "true" YADE simulations (with dynamic bodies), all this
discussion would be transparent since the timestep would be drastically
reduced for stability reasons.


PS : thank you for your minimal code, note that utils. prefix is unnecessary: you can just type sphere(..) thanks to the way the utils YADE module is imported at startup. 
Note also that the L3Geom description of contact is said to be experimental in the doc [*]. I think most users use ScGeom and related classes, and do not think there remains anyone around that could maintain/help with L3Geom.

[*] https://yade-dem.org/doc/yade.wrapper.html#yade.wrapper.L3Geom

-- 
You received this question notification because your team yade-users is
an answer contact for Yade.