Clump modelling

Hi!

Not sure, what you mean "clumps", but I think you can use "bonds". It is special "connections" between particles,
what can be destructed for some reasons. There is an example in tutorial on this topic.

Hi Jose,

Thanks for the question. ESyS-Particle does not implement "clumps" in the sense of PFC (and possibly some other DEM codes). In PFC, a clump is a rigid group of particles in which all forces and moments are computed relative to the centre of mass of the clump. In a parallel DEM code, this type of force calculation requires a large amount of communication between worker processes when a clump crosses a boundary between two subdomains. Due to the way ESyS-Particle is implemented, it is not feasible to implement this kind of "clump logic".

As Anton mentioned, it is possible to bond particles together into "clusters" that may or may not fracture/fragment during a simulation. The clusters are not rigid though: they can deform due to applied loads. If you set the elastic stiffness of the bonds high enough, this is usually not a major problem. When setting up the particle geometry, there are various ways to create clusters depending on your particular needs. Probably using LSMGenGeo is the easiest way. Unfortunately I haven't yet written a good introductory tutorial on LSMGenGeo but it does come with a few examples.

If you would like a copy of LSMGenGeo, it is available for anonymous checkout from the ESyS-Particle SVN repository:
svn checkout https://svn.esscc.uq.edu.au/svn/esys3/lsm/contrib/LSMGenGeo LSMGenGeo

Cheers,

Dion.

Thanks for the answers. Indeed, by clump I was referring to a group of particles bonded together where their relative position never changes and two particles within the same clump do not have repulsive forces, and so they can overlap, but they have repulsive forces to any other particle outstide their clump.
The big advantage is that by not needing to compute bond forces and interactions between particles of the same clump, it becomes less computationally expensive than modelling clusters, and it also allows more irregular shapes since the particles can overlap.

Is this feasible through creating new interactions or would it require going more deeply into the core code?

> Is this feasible through creating new interactions or would it require going more deeply into the core code?

It would require going deep into the code, for 2 reasons:
- to calculate the rigid body motion of a group of particles you need to know the external forces on _all_ the particles involved and the moment of inertia (linear + rotational) of the whole "clump" - so you basically need to transfer the information about forces from the "particle level" to the "clump level", calculate the motion of the clump and than transfer the information about that motion back down from the "clump level" to the "particle level" in order to update the particle positions
- you'd need to deal with the parallelization issues arising it the clump is located on a subdomain boundary

It is not impossible but it would be a fair chunk of work.

Steffen

Thanks for the quick responses. Perhaps using a cluster without overlapping particles, and very stiff bonds does not differ very much from clumps. I'll look into it.