yade-users team mailing list archive

[Raphaël Maurin <question406523@xxxxxxxxxxxxxxxxxxxxx>]

Question #406523 on Yade changed:
https://answers.launchpad.net/yade/+question/406523

    Status: Answered => Solved

Raphaël Maurin confirmed that the question is solved:
Hi Bruno,

Thank you for the explanation. Then it should in principle work as the
cylinder I am using is always fixed (so all DOFs blocked, from the very
beginning).

I checked, and it is rather surprising but for the first interaction
computed at the creation of the cylinder (so between the element of the
cylinder), the mass of the two interacting cylinder components are both
zero. This is why there is this warning at the beginning.

There is however no link with the following. I found out that if I use smaller particles  (d = 6mm), everything is working fine. 
The origin of this dependence in size lies in the stability of the resolution of the equation to infer the value of cn from the value of en (find_cn_from_en in ViscoelasticPM.cpp). There was an initial perturbation of the solution which was rather small, and leading to a division by zero when considering bigger particles. 
I placed a limitor to avoid the division by zero, and continue the loop. This leads to more iteration, and when it does not work, it ends up with the warning message. I also tuned the value of the initial perturbation for it to work on a larger range of cases. The range of resolution depends on the mass of the particles, kn and en, and for the case with kn in the rigid grain limit, and en = 0.5, it works over a range from d  = 1e-6 to d = 10 for two particles colliding. At en = 0.9, it works on the range 1e-5 to 10, and at en = 0.1, it works on the range 1e-6 to 1. This should cover most of the applications, and in case it does not work there will be the warning message. 

I just made a commit with all that included.

Thanks for the help !

Raphael

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