Hi, Some comments (useful or not): - If it's possible, try using velocity-Verlet integrator so that the the value of velocity is computed at time step (not mid-step) - Time integration (trapezoidal rule) involves approximations. Vaclav is right in believing that this error accumulation should decrease when the time-step is decreased.
Should I suggest to perform the balance on a single contact by using a tiny time-step? Other things come in my mind: - Are the angular rotations taking into account in the kinematic energy? - Is the Cundall damping shut down? Valclav, could you tell me which file has to be looked at? ciao, Vincent Le 14 nov. 2010 à 00:26, chiara modenese a écrit : > Hi Vaclav, > > why not trying to calculate the velocities at the right time? I remember I > was doing it and energy was conserved (without damping, either local or > global). We just need to shift them by adding acc*dt/2 (btw, we need to > _subtract_ this term in order to shift in the right way, as discussed with > Bruno on the list). It is worth a try. However, reducing the time step should > help, if we do not apply the shift. > > Chiara > PS nice idea to track energies in this way > > > 2010/11/13 Václav Šmilauer <[email protected]> > > I am not sure about the cause; two possible explanations come to mind (i) > mid-step values of velocity vs. on-step accelerations etc. and (ii) without > damping, many elastic collisions take place (their number increases if > damping is small, as the system never goes to quasi-static state) and > integration error accumulates (what contradicts this is that the precision > would be increased with smaller dt, which is not the case, and also that the > imprecision effects would cancel each other in average). > > Could someone have a look at the code and suggest something? > > Cheers, Vaclav > > _______________________________________________ > Mailing list: https://launchpad.net/~yade-dev > Post to : [email protected] > Unsubscribe : https://launchpad.net/~yade-dev > More help : https://help.launchpad.net/ListHelp > > _______________________________________________ > Mailing list: https://launchpad.net/~yade-dev > Post to : [email protected] > Unsubscribe : https://launchpad.net/~yade-dev > More help : https://help.launchpad.net/ListHelp
_______________________________________________ Mailing list: https://launchpad.net/~yade-dev Post to : [email protected] Unsubscribe : https://launchpad.net/~yade-dev More help : https://help.launchpad.net/ListHelp
