yade-dev team mailing list archive
-
yade-dev team
-
Mailing list archive
-
Message #03682
Re: π=2 ?! (Ip2_FrictMat_FrictMat_FrictPhys)
> BTW, Bruno do you know if this method to get the timestep that varies with
> the global stiffness can be well applied also when we have a non linear
> relationship between forces and displacements at contact? As in the
> Hertz-Mindlin formulation, for instance. Do you think that in such a case
> the critical time step should change or should be taken differently than
> having a linear relationship?
>
>
If you assume small variations per step (i.e. no 2-steps bouncings with
Hert-Mindlin), you can compute dt exactly the same way. All you need is to
define dt with tangent stiffness : dfn/dun and dfs/dus for the current state
of the contact.
I don't remember, did you write your law with tangent or secant stiffness (I
hope french maths translate well...)?
If you use tangent stifness, you have stritcly nothing to do to get
Hert-Mindlin handled by GSTimestepper.
If you use secant stiffness, you can still exploit the old useless
intitialKn/Ks that remains in interactionPhysics, and which is currently
redundant with kn. Let kn be the tangent and be used by timestepper, and use
intialKn in you law (and rename initialKn to say what it is in your law, no
engine is using that anyway).
How is HM working for you a.t.m.?
Bruno
References