Hi Fei,

I am one of the lead developers of ESyS-Particle. Here are my thoughts on the various advantages of the 3 DEM packages you mentioned. I will try to be unbiased! ;)

ESyS-Particle has 2 main advantages. The first is that its MPI parallel engine has been demonstrated to scale (weakly) to in excess of 30,000 CPU cores and hence is definitely well suited for HPC DEM simulations. The second advantage is the BrittleBeamPrms particle-pair interactions that has been specifically designed for rock fracture and fragmentation simulations. This interaction correctly simulates Griffith crack propagation and, in particular, the formation of wing cracks. Credit to Dr. Yucang Wang for the mathematical formulation.

Aside from the BrittleBeam interactions, ESyS-Particle provides a range of other interaction types, as well as support for planar and triangle mesh walls. ESyS-Particle is perhaps not as fully featured as the other two packages in terms of the range of interaction types already implemented, and documentation for programmers is a bit lacking at present. The Tutorial has been very well received by new users and it would appear that working through the Tutorial usually provides the knowledge one needs to design one's own simulation models.

AFAIK, Yade is also relatively easy to learn how to use and has a good base of support from current users and developers. To the best of my knowledge, Yade is currently only shared-memory parallel (openMP?) and as such, will probably not scale to extremely large problem sizes on HPC clusters. You should check with the Yade developers however, as I have not tested recent versions of the code.

LIGGHTS is another good choice for HPC DEM simulations from what I can tell. However I have not trialed LIGGHTS (or LAMMPS) so am not really well placed to provide a fair comparison with ESyS-Particle in terms of parallel performance, interaction types or usability. Once again, I suggest you contact the developers for more information.

Regarding simulating carbon nanotubes, I really am not well placed to answer this as I am not familiar with the topic or your research requirements. I suspect that any one of the 3 packages would be equally capable on the basis of the interaction types each provides. You may need to implement your own interaction type(s) to meet your needs, in which case your choice will depend on which package you find easier to modify for your purposes.

I hope this answer helps you. Best wishes for your research.

Cheers,

Dion

Dear Dion, thank you for your clear answer.

I have another question: does ESyS-Particle have the "bonded particles" (an interaction type), I think this type means two particles connected with each other in the whole simulation process.

As you known, I want to study the interactions of amount of carbon nanotubes. Here, I attached a paper where the authors work on this aim but the number of nanotubes is not enough in my opinion. They used the commercial software PFC, I want to know if the ESyS-Particle can do the same thing (see Fig 15A), if you could see this paper.

Many thanks!

Hi Fei,

I have used LAMMPS quite a lot. For nano-tubes you may need to actually to do an MD simulation rather than DEM, LAMMPS can do both. I am sure that LAMMPS provides several options to include "bonds" both chemical bonds and also rigid bonds (rigid object created with spherical particles using fix rigid). 

LIGGGHTS is a redesign of LAMMPS software mainly for DEM applications. It can also simulate complicated geometries and also heat transfer and more complicated models are available (rolling friction e.g) and CFD-DEM capabilities are available but don't think the options are available for MD simulations

Both are capable of performing very large scale simulations for spherical particles but they don't scale quite well with rigid objects. 

I've never used the others and can't comment on that.

Best regards

sina

Thanks Sina, but I'm worrying about the MD model is too small scale to be used for large scale CNT networks. Actually, I'm considering the kinematics of whole CNTs rather than every atoms in a CNT. Hope Liggghts can do this job?

Dear Fei,

I am a University professor leading a research group (www.xdem.de) for multi-physics applications. We have tried LAMMPS e.g. LIGGGTHS and YADE, with commercial software we have no experience. Without going into comparisons of different tools, all these software tools were developed with a particular purpose in mind. As such all of them are excellent tools that serve the very well in their field of applicability. However, there is no universal tools that could be used for everything. Having that in mind we develop the Extended Discrete Element Method (XDEM) that includes an interface to Computational Fluid Dynamics (CFD) and Finite Element Method (FEM) combined with an extensive functionality in DEM. Among these functionality is also the application you have in mind. We investigate into material science by representing a wide range of materials with particles that are connected by functional bonds very similar to CNT. Therefore, I could offer to use our software within an academic framework that additionally provides excellent expertise and support within this area.

Best

Bernhard

While on this topic, I am interested in knowing, which of the open source DEM codes is most suitable to modeling viscous-elastic waveform propagation in granular media (i.e. sandy, unsaturated soils),  under a variety of strains.... Could DEM be made to hold all the particles fast under low strains (elastic propagation) and then allow gradually more non-linear behavior...

Thanks,

Juan

Hi Fei,

you might want to have a look at one of our latest publications. Yade supports various types of elements such as cylinders or tubes which might be relevant for your application. As pointed out by Dion, Yade uses OpenMP so it doesn't scale as good as the other packages but in my opinion it is probably the most versatile one.

HTH,

Klaus

Article Modelling of deformable structures in the general framework ...

 hi,

I wish to use discrete element method to study the wave propagation in soil subjected to impact loading. Which of the open source packages can model the problem? 

Dear Monica,

we could do this with XDEM (xdem.de). Pleas elet me know if you are interested to discuss further.

Hi Monica,

any of the above mentioned packages can be used for this purpose.

Can I use open source DEM softwares to model fatigue in specimens of asphalt beams in the laboratory? I intend to subject some samples to 4 point bending test and simulate fatigue using DEM

Is it possible to simulate liquid phase using DEM ? For example, if I want to simulate heat transfer through nanofluids (fluid containing nanoparticles), Will DEM be effective ?

I was using LAMMPS based MD simulation for this purpose but size restriction of MD is an issue. I was thinking if DEM can be used for this purpose.

Any suggestion in this regard will be helpful.

Thanks

hi,

is it possible to simulate the trajectory of a fish pellet from a distribution channel of fish feeder machine using DEM software. the concept of my design is same as vertical impact rock crusher but with four distribution channel on the wall of the machine.

LAMMPS works well for both DEM and MD. I know the parallelization of LAMMPS is very good. The only disadvantage of LAMMPS is that there are not many contact models but you can develop your owns on LAMMPS.