validation of code---Time step

raguelmoon's picture
Submitted by raguelmoon on Sat, 03/24/2012 - 16:44

Hi LIGGGHTS users,

I want to do some simulation in order to validate the CODE with my experimental data. My experimental findings are related to static granular packings in which radius of particles are varied from 0.1mm to 16mm. I used this command to check time step but i get different values for different diameters:

fix ts_check all check/timestep/gran 1000 0.1 0.1

For example this warning is printed during simulation, the value is constant for one kind of diamtere.

WARNING: time-step is 31.483188 % of rayleigh time

Is it right way or i have to make same values for all diameter packings? How to fix it? by decreasing time step or by fixe_chek/timestep/gran command?

Thanks for help,
Ram

Forums:

Philippe's picture

Philippe | Mon, 03/26/2012 - 09:06

Hello Ram,

the Rayleigh time

dt_r = PI*r*sqrt(rho/G)/(0.1631*nu+0.8766) (see documentation)

is independent of the particle velocity. So, your result with one constant value for a certain diameter is correct.

The contact models assume that the overlap between two spheres is small compared to their diameter, thus for smaller particles, you will typically need smaller timesteps to correctly resolve the collisions. 31% of the Rayleigh time seems a bit too much to me, but one thing you can usually get away with to keep the timestep in a reasonable range is decreasing Young's modulus and thereby increasing the Rayleigh time.

best,
Philippe

raguelmoon's picture

raguelmoon | Mon, 03/26/2012 - 10:18

Hi Philippe,
Thanks for help. This helps me much. But, one question: % of time should be low in warning? For example, 31% is higher value, it should be 1% or less? Am I right?
Thanks in advance,
Ram

Ram

Philippe's picture

Philippe | Mon, 03/26/2012 - 10:44

You can control the threshold value for the warning in the fix command, please check the documentation for more information. Generally, values higher than 10% of Rayleigh time might cause trouble for some simulation setups, but for others you can get away with values up to 30% and still get useful results. The only way to find this out is extended testing of your simulations.

raguelmoon's picture

raguelmoon | Mon, 04/02/2012 - 07:38

Hi Philippe,
I can control time step by fix check/timestep/gran command. I just choose higher values of fraction_r and fraction_h and dont get any warnings. Is it safe way to run simulation without changing timestep or youngs modulus for particular diameter of grain (just changing fix check/timestep/gran properties)?
Thanks,
Ram

Ram

Philippe's picture

Philippe | Mon, 04/02/2012 - 10:03

Hello Ram,

you have just removed the warning about the timesteps, and did not change anything about the actual simulation - thus your simulation is as (un?)safe as it was before. You can find out if your timestep is safe by doing two test simulations with small and large time step and see if the results agree. Another hint about stability is the thermo output - if your kinetic energy stays within reasonable values, then the timestep is at least not much too large.

cheers
Philippe