when running a case with cohesive particles I observe an anomalous behavior of particle clusters.
The case consists in a periodic cube with a random positioning of spheres and a random velocity in the vertical direction (between -0.5 and 0.5 m/s). Only solids and no gravity is considered. To simplify things, I have imposed a constant cohesive force with a value such that particles are sticking to each other when they contact each other (see calc_force_dem.f).
I am observing 2 anomalous behaviors:
Some clusters suddenly accelerate to a large velocity, even if they don’t collide with any other particles (see the video below looking from the top). I have changed the mesh and this behavior also manifests. In fact, it being an only granular test, how can the mesh influence the simulation apart from the particle neighbor seach algorithm?
When clusters cross the system boundaries they also accelerate. I guess this is attributed to the cyclic boundaries, but I do not understand why.
I have checked this case against other DEM solvers and the movement of particles is mainly vertical and this large horizontal velocity does not appear.
Could you please check the case or guide me on what could be happening in here? This has been a showstopper of my modelling work for some weeks…
For completeness, please see below the video corresponding to the exact case I have upladed as a .zip file. Please note the abnormal changes in velocity of the clusters.
Thank you for your answer. Yes, it also happens when using VdW default model (see attached the case). I used a Hamaker constant similar to what I would expect in the constant force case of my previous post, so particles are sticking but overlap is not excessive.
I also tested this case in series and the anomalous behavior persists. I don’t know if it could be somehow related with the cyclic boundary, it looks that is mainly manifesting at the limits of the system.
I would really apreciate it if you could take a look at it.
Do you have any idea on what could be happening? Probably some strange behavior of clusters when crossing cyclic boundaries or boundaries between processors? Could this be related to this other recently opened case?
We have been investigating this issue and we still do not find an explanation. We have tried to modify the contact parameters (max steps between neighbor search, factor defining particle neighborhood and distance/diameter triggering search) and the problem persists.
What we observed is that, when a group of particles crosses the cyclic boundary (or the limit between processors), it sometimes kind of collides with itself, as if the particles do not carry the same inertia and they do not belong to the cluster. The cluster then changes of velocity and direction when it shouldn’t. This generally occurs for sufficiently large sizes of the cluster, but we have seen also clusters of two particles moving vertically and then having some horizontal component when crossing the limits of the system.
In the video attached (looking from two different planes) you can see an example in which a group of particles moves vertically until a point in which it crosses the boundary of the system and then it abruptly changes of velocity and direction.
We are a bit concerned about this, we want to carry out simulations considering cohesion in a fluidized bed, but now we are not certain that the results would be realistic.
Do you think this issue could be fixed or do you have an explanation for it?
We are more than ready to run any benchmark or code modification that may be required!
Thank you for your help,
Eduardo
P.S: Sorry we opened another post the other day with the same topic, one of my students was using our server and did not notice that I left my account logged in.
Is it possible for you to generate a simple test case with say a single cluster that moves across the cyclic plane and is repeatable? Maybe use a particle_input.csv file as a starting point. This would help in troubleshooting the issue.
By looking closely at the motion of the particles, I have filtered the particles in the system to a simple case in which only 2 particles interact. When they join, they have a nearly vertical displacement, but they start displacing in a different direction when they cross the boundaries of the system. I would say this is what is also affecting the big clusters of particles. Please see below the bug report with the case and the video.
Also something interesting that I observed is that, if instead of the full .csv file, the particle_input.csv is limited to only the 2 particles in the system (particle_input_filter.csv), this does not happen. I don’t know if this is could be then somehow related with the particle mapping.