ESyS-Particle

Discrepancies between mesh and wall interactions

Asked by Quinn Reynolds

Hi,

I'm running a simple simulation in which a collection of particles is dropped into a square box, and settles under gravity. The particle interaction group is NRotFriction, and the four walls of the box are standard infinite elastic walls.

I'm getting significant differences in behaviour depending on whether I model the floor of the box using an infinite wall, or a flat mesh consisting of two triangles. When the infinite-wall floor is used, the particles fall and settle quickly to a steady state configuration. When the mesh floor (in the same position, with the same elastic spring constant) is used, oscillations and instability persist and actually seem to get worse the longer the simulation runs for.

Is there any reason that mesh surfaces and simple infinite walls should be subject to different numerical stability constraints, given how they are treated in the ESyS-Particle code?

Kind regards,
Quinn

Question information

Language:
English Edit question
Status:
Solved
For:
ESyS-Particle Edit question
Assignee:
No assignee Edit question
Solved by:
Anton Gladky
Solved:
Last query:
Last reply:
Revision history for this message
Best Anton Gladky (gladky-anton) said : 		#1

Hi,

it can be related to this bug
https://bugs.launchpad.net/esys-particle/+bug/442881

_____________________________
[ENG] Best Regards
[GER] Mit freundlichen Grüßen
[RUS] С наилучшими пожеланиями
[UKR] З найкращими побажаннями

Anton Gladkyy

2010/1/27 QuinnReynolds <email address hidden>

> New question #98840 on ESyS-Particle:
> https://answers.launchpad.net/esys-particle/+question/98840
>
> Hi,
>
> I'm running a simple simulation in which a collection of particles is
> dropped into a square box, and settles under gravity. The particle
> interaction group is NRotFriction, and the four walls of the box are
> standard infinite elastic walls.
>
> I'm getting significant differences in behaviour depending on whether I
> model the floor of the box using an infinite wall, or a flat mesh consisting
> of two triangles. When the infinite-wall floor is used, the particles fall
> and settle quickly to a steady state configuration. When the mesh floor (in
> the same position, with the same elastic spring constant) is used,
> oscillations and instability persist and actually seem to get worse the
> longer the simulation runs for.
>
> Is there any reason that mesh surfaces and simple infinite walls should be
> subject to different numerical stability constraints, given how they are
> treated in the ESyS-Particle code?
>
> Kind regards,
> Quinn
>
> --
> You received this question notification because you are an answer
> contact for ESyS-Particle.
>

Revision history for this message
Quinn Reynolds (quinn-reynolds) said : 		#2

Thank you Anton, it is very likely the two are related. I'll wait and see if the bugfix for your problem also solves mine.

Kind regards,
Quinn

Revision history for this message
Quinn Reynolds (quinn-reynolds) said : 		#3

Thanks Anton Gladky, that solved my question.

Revision history for this message
Dion Weatherley (d-weatherley) said : 		#4

Hi Quinn and Anton,

I discussed this with Steffen and we agree that it is probably the "Hanging Spheres" bug causing the problem. I consider this bug highly critical to resolve before the next stable version release (hopefully early this year). Unfortunately other work commitments are higher on the list of priorities right now.

Steffen suspects the bug is due to spurious interactions between particles and the edges of triangles. Perhaps you might be able to find evidence for that in your model by examining the velocity field of particles near the floor as the instability grows?

Quinn, if you are willing, could I please ask for a copy of your script(s) and meshfile? Your setup may be a good test-case for hunting the bug. You can email me directly if you prefer not to post the script on launchpad. My email is <email address hidden>.

Cheers,

Dion.

Revision history for this message
Quinn Reynolds (quinn-reynolds) said : 		#5

Hi Dion, thank you for the response, I'll email you the scripts with pleasure. I appreciate the effort given your other work commitments - our research project using ESyS-Particle only officially starts around April this year, so there is no rush from my side.

I have found some additional evidence which may or may not help you in your troubleshooting:

1. The time step size used in the simulations is quite large given the parameters of the particle and wall interactions, and seems to be quite a critical variable in exaggerating the numerical instabilities on meshes; if the time step is reduced considerably, the differences between the mesh-floor and wall-floor cases mostly disappear.

2. I also set up a third case where the floor mesh consists of one oversized triangle (i.e. the floor of the box is completely contained within the interior of the triangle, no edges or vertices). Interestingly this case shows similar unstable behaviour to the case using the mesh floor with two triangles, suggesting that the problem may not be confined to the mesh edges.

Kind regards,
Quinn

Revision history for this message
SteffenAbe (s-abe) said : 		#6

Hi Quinn,

turns out that your problem wasn't related to the "hanging particle" bug after all but that it was due to a wrongly scaled stiffness parameter in the elastic mesh interactions. Fixed in the current svn trunk version (r1278) - at least your test scripts do work as expected now.

Steffen

Revision history for this message
Quinn Reynolds (quinn-reynolds) said : 		#7

Thank you very much for the troubleshooting Steffen. I'll download the latest svn version to continue my testing with.

Kind regards,
Quinn