Which factors cause the fluctuation/uncertainty in DEM simulation

Asked by Leonard

Hi,
I'd like to ask that which factors cause the fluctuation/uncertainty in DEM simulation.

For example, let's say the stress-strain curve from a triaxial compression test, it can be found that (1) the curve is not smooth, why? (2) It seems that different results can be got when run the same simulations for several times. which factors cause this uncertainty?

Thanks

Leonard

Question information

Language:
English Edit question
Status:
Solved
For:
Yade Edit question
Assignee:
No assignee Edit question
Solved by:
Jan Stránský
Solved:
Last query:
Last reply:
Revision history for this message
Karol Brzezinski (kbrzezinski) said :
#1

Hi Leonard,

any code to the example would be helpful. Otherwise, there are unlimited possibilities.

What would you call uncertainty, because two issue are mentioned:non-smoothness, and repeatability.

I can only guess that non-smoothness can be caused by relatively big grains and high angularity of particles (or rolling friction). Different results of 'the same' simulation, may be related to changing initial configuration of the particles.

It is just guess. It is hard to discuss it without more concrete example.

Cheers,
Karol

Revision history for this message
Best Jan Stránský (honzik) said (last edit ):
#2

Hello,

> (1) the curve is not smooth, why?

DEM, discrete/distinct element method, is **discrete**.
You are dealing with discrete points in space and discrete instances of time.
The results are values at discrete time steps.
The resulting "curve" is not smooth by definition..

If you mean non-linear, non-monothonic or something similar, there are a few reasons.
- triaxial engines for prescribed stress does not directly prescribe stress, but rather "guess" strain (rate) such that stress matches the desired value. This "guessing" may cause some fluctuation.
- inertial and dynamic effects cause fluctuations naturally

> (2) It seems that different results can be got when run the same simulations for several times. which factors cause this uncertainty?

There are a few main factors:
- uncertainty by design, e.g. using python random package, for some stochastic analysis
- uncertainty by "accident", e.g. using makeCloud without seed parameter
- multi-core running [1]
- using PyRunner with realPeriod, which can be executed at different simulation times each run
- nondeterminism in 3rd party software

Cheers
Jan

[1] https://yade-dem.org/doc/formulation.html#result-indeterminism

Revision history for this message
Leonard (z2521899293) said :
#3

Hi,

Thanks Karol and Jan for your reply.

Currently I don't have an example at hand, I had this question when I read DEM papers. I know this is the feature of DEM, but I don't know why :(

Good to know the above details!

Cheers
Leonard

Revision history for this message
Leonard (z2521899293) said :
#4

Thanks Jan Stránský, that solved my question.