← Back to team overview

yade-users team mailing list archive

Re: [Question #695226]: RungeKuttaCashKarp54Integrator with gridConnections is causing errors


Question #695226 on Yade changed:

Rohit John posted a new comment:
Dear Bruno,

> what happens with super-small O.dt? (defining what is "super-small" needs to think about mass-viscosity-stiffness parameters)
- It depends on the damping fraction and the material stiffness. I wrote the code such that the damping coefficient is a fraction of the stiffness. (Same fraction for normal, shear, twist and bending stiffness in Law2_ScGeom6D_CohFrictPhys_CohesionMoment). At small time steps it works, but if I want higher damping I must reduce the time step. For one simulation it had to be 1e-10.  Even days were not enough to get some useful result.

> are the directions of forces, moment, velocities and spin consistent, with just 2-3 nodes after 2 iterations?
I checked the values and when the damping and dt were badly set, the forces and velocities grow exponentially, but when I reduce the dt or damping they "calm down". 

>From my understanding the integrator in YADE is similar to forward Euler
which is unstable when there is damping. So I though changing the
integrator might help. Janek Kozicki solution works but the
RungeKuttaCashKarp54Integrator is extremely slow.

If possible could you check could the code found here:
https://github.com/rohitkjohnedu/yade_laws .

Kind regards,
Rohit K. John

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