Welcome to ask.woodem.org. You may post when you login through your GitHub account.

How to improve the run times of a simulation?

0 votes
asked Aug 26, 2016 by andrewDEM (290 points)
  • the current model is using:
    • dtSafety=0.5
    • woo.dem.FrictMat(ktDivKn=.2,tanPhi=.2,density=1000,young=50000)
    • S.engines=woo.dem.DemField.minimalEngines(damping=.5)
  • the typical time step that I obtain thru most of the simulation is 0.006sec. The simulation takes about 2hrs (for D=200mm particles 400sec simulation) and 24hrs+ (for D=100mm particles 400sec simulation).
  • are there any parameters that I can change or models that I could use instead to make the simulation run much faster?

1 Answer

+1 vote
answered Aug 29, 2016 by eudoxos (43,490 points)
selected Oct 21, 2016 by eudoxos
Best answer
  1. Set dtSafety to something between 0.7 and 0.9; unless you use clumps, that will be fine.
  2. Increase density; since you are going for a quasi-static result, that is OK; but make sure you increase stiffness a little bit as well so that overlaps are not too big, or particles don't fall through the triangulated surface. Remember p-wave critical timestep $\Delta t_{\mathrm cr}=r\sqrt{\rho/E}$ (https://woodem.org/woo.utils.html#woo._utils2.pWaveDt), so play with all of those -- $\rho$ up, $E$ as low as possible, $r$ as big as reasonable.
  3. Changing radius: angle of repose does not depend on particle size (small heap will have the same angle as big heap), as long as they are much smaller than the heap size (say 15 particles accross the smallest dimension), and I think differences will be attributed to static/dynamic angle of repose, which are different; you can test with smaller particles once you have something working, but not necessary to run everything with tiny particles. In fact simulation time increases with the 4th power of inverse particle size (number of particles in constant volume increases with the 4rd power of decreasing particle diamenter, plus number of $\Delta t_{\mathrm cr}$ timesteps within contact time increases linearly with decreasing radius)

HTH, Vaclav