I am encountering an interesting effect when doing CGP-DEM simulations. I am trying to simulate a CGP-DEM system in which there is a central obstacle/cylinder that injects gas. Everything runs smoothly (see video). However, when analyzing the resutls, I observe that there is some particle pinning in the system. That is, particles tend to be more frequent in some vertical sections of the bed than in others.
As a reference, I attach a video of the simulation, together with the .mfx file and two snapshots of my postprocesing. In those snapshots, I have firstly created a grid that is two times smaller than the mesh in the simulation, and I have computed the number of particle centers inside those grids (figure 1), together with the standard deviation of the particle number along the simulation (figure 2). Note how in some regions particles are more frequent than in others and that also the standard deviation seems to be affected. Note the wavy behavior of the figures at the bottom of the bed. I have checked the postprocessing and it is does not seem something related to it.
What could be the cause of this particle pinning and how can I solve it?
Yes, Np is the total number of particles that visit that cell over the simulation. I am aware it is not a proper variable to monitor, as it depends on the number of exporting steps and the area of the cells, but it serves to show some results.
Sigma_p is computed sequentially. That is, for each time-step the number of particles in each cell is computed and sigma_p is updated cell by cell from the previous value in each cell. Doing otherwise would require a lot of memory.
Yes, the first 5 seconds of simulation (over 30 s of simulation time) were removed.
That is probably why we see some darker region in Np close to the walls? I will take a look at this. Maybe this effect is hindered by the size of the grid.
Yes, I did that and when one uses a grid size similar to the mesh of the simulation this effect dissapears (the regions with high number of particles are averaged with the ones with low number of particles). Further decreasing the size of the postprocessing grid does not change the behavior of Np nor sigma_p. That is, the results are not depending on the size of the grid as soon as it is sufficiently fine to caputre this pinning behavior.
I will try to take a look a the suggestion of Charles to see if this happens in DEM, but due to limited computational resources I may end up simulating a completely different case.
Sorry for the late response. I have done a test using DEM and, when operating close to Umf, the particle pinning still appears. Please see attached the snapshots for the particle number (per unit of front area) and the standard deviation of the particle number as described in other posts. Note again how there are regions in which the particles are more frequent. Why may this be happening?
Please see attached also the .mfx file for this specific case.
Yes, I am concerned about that, which happens both for DEM and CGP. As far as I understand this represents a higher probability of particles to be present in a given vertical coordinate (higher average number of particles and thus lower standar deviation). I don’t see how this can be related to the flow pattern. Shouldn’t the snapshots be very uniform in those regions instead of showing those stripes?
I am thinking the particles just above the inlet are more constrained than those that are higher up due to the proximity of the bottom plane. This makes them move almost as layers rather than being fully mixed. This could explain the stripes you see.
Thank you for your answer. What I don’t really understand is how these layers have a preferential positioning in the bed, that is, particles are more prone to appear in specific vertical coordinates. In my opinion, that is more a numerical effect than an actual physical one. I don’t know if there is any other test I can do to double check this.
I am only speculating and I could be wrong. I am suggesting the stripes are related to nearly uniform vertical motion of particles near the inlet plane. This, in combination with the postprocessing grid could produce this result. I am not sure. It it were purely numerical, you should see the stripes all over. Maybe you can try to increase the inlet velocity or try with a central jet (say between x=0.4m and x=0.06 m) to see if you still get the stripes.
