Issue with PERI/EPS

LAMMPS LAMMPS Mailing List Mirror
1

Dear all,

I’ve been performing some simulations with the peridynamics package. I’ve run into an issue I cannot solve by myself, and I hope you can help me.

I’m modeling the fracture of a “dog bone” sample. I made some initial tests using the standard PERI/LPS model, and I got some “good” results. Meaning, I put in realistic parameters, ended up with believable results for a series of parameters I am monitoring (applied forces/stresses, energies, fracture patterns, etc).

I then wanted to test the elasto-plastic model (EPS), which (I understand) should work in the same way as the LPS model. I modified my simulation script adding in the required yield stress, but my samples keeps on blowing up in a thousand pieces as soon as any tiny amount of stress is applied to it.

I couldn’t find any example script for the EPS model, so I wonder if I’m doing something wrong.

This is the (working) code for the LPS:
pair_style peri/lps
pair_coeff * * 2.4e9 1.0e9 0.0015001 0.0005 0.25

while this is (non working) what I was using for the EPS:
pair_style peri/eps
pair_coeff * * 2.4e9 1.0e9 0.0015001 0.0005 0.25 15.0e6

By the way, I am aware that there was an issue with the kernel stack in the eps code (fixed by Axel Kohlmeyer some weeks ago), but I don’t think that’s related. I had that issue too, and the patch fixed it for me.

Thank you very much,

Alessandro Sellerio

2

Maybe Rezwan or Mike can comment - this sounds like a Q for a PD expert.

Steve

3

using smaller timestep, larger s00 etc, could help.

4

Right now I’m running the simulations with timestep = 1.0e-7 s , and add no issue with lps in the same simulated conditions.

I’ll try with smaller timesteps. Are there rules of thumb that I should take into account?

Thanks,

Alessandro

5

there is no specific thumb rule. the horizon=3*lattice_size. you can look at the sandia lab report "A Nonlocal, Ordinary, State-Based
Plasticity Model for Peridynamics" by John A. Mitchell on plasticity model.

6

Dear all,

I tried running the same simulation with a timestep 1/10 longer.

Unfortunately I run into the same issue, the sample “explodes” as soon as a little stress is applied to it.

Please let me know if you have other suggestions I could try. If you find it useful, I could share the input script.

Thanks a lot,

Alessandro