DEM Particle wall-link

In this week’s blog post we will look at the Particle Wall-link model that can be used with DEM particles. DEM, as we know by now, is the Discrete Element Method, that allows us in Simcenter STAR-CCM+ to model the behavior of particles. And moreover, in addition to what can be done in Lagrangian multiphase (LMP), we can also model particle-particle interactions and particle-wall interactions. We can specify how a particle of a certain material behaves when colliding, either with the same type of particle, with a different type of particle, or a wall boundary. There are several types of simulations that can be made using DEM particles. But here we will emphasize on using the particle-wall link model.

The Particle-wall link model

DEM particles use the Particle-wall link model to stick to walls they come into contact with, and then move with the wall if necessary. Once a particle links to a wall boundary, its translational velocity is always the same as that of the point on the boundary it is linked to. The particle keeps a fixed orientation relative to the local surface normal. This local surface normal is a smoothed, continuous normal interpolated at the link point if the tangential velocity is non-zero. If the tangential velocity is zero, the local surface normal corresponds to the face normal. If a particle moves from a linked to an unlinked boundary, that particle is removed from the simulation.

The above description comes from the documentation of Simcenter STAR-CCM+ 2310. So, we can create a wall link model between a DEM phase and any wall, making the particle stick to the wall. If that wall is moving, either using DFBI, or by relative velocity, the particle will follow that motion. This allows us to rigidly move bodies and get an interaction between particles easier than if we need to set the simulation up with rigid body motion. The animation below shows an example of this. Version 2021.3 of Simcenter STAR-CCM+ gave us the possibility to use polyhedral particles, meaning that the particles can take any shape, and you can with advantage use a CAD model or surface mesh to describe your DEM phase. In the example below both the grains and the grain elevators are simulated as different DEM-phases. The difference between the phases is that in the multiphase interaction with the wall boundary, the grain elevator has a particle wall-link defined with the wall. That together with an injection of the DEM phase that instantly make sure the DEM phase attaches to the conveyor belt, and we can easily simulate the “scooping” of the grain particles.

 

This type of simulation is possible to set up with any shapes. However, it is important to note that the interaction between the continuous phase and the DEM-phase(s) is represented by a source term. If we consider the above example again, any air flow that is passing through the particles will only consider the DEM-phase as a source. No boundary layer or so will form on the walls of the grain elevators. This is the case for all DEM-simulations. This also makes it so that this methodology can be used with simulations using the mesh-free approach, where the only important forces acting on the particles are the forces imposed by the other DEM-phase or gravity.

An example of this

We will look at a simplified setup here, where the particle-wall link method is used. In this setup we use a self-created “paddle” or moving mixer to help mix 200 spherical DEM particles that are to be cooled by a continuous phase flow passing through the domain. A simple set of shapes are copied and rotated and finally combined using a Boolean unite operation. The part “Unite”, from the unite operation, in the part-tree will then represent the input for our DEM phase.

2 Pic1 geometry

The domain is simple, from the picture below you can see that we have a cylindrical drum, on the front side we have an inlet for the continuous phase, and on the back side we have an outlet. Both the inlet and outlet are specified in reference to the DEM-phases when it comes to interactions. You can see that under the boundary for the outlet in the tree we have a folder called “phase conditions”. That allows us to specify what type of boundary the continuous flow outlet has in reference to the separate DEM-phases. We can see that the symbol for the spheres is that associated with a wall boundary, and that associated with the phase we call paddle (our polyhedral phase) has the same setting as the “outlet”. We say that the outlet is phase impermeable when it comes to the spheres. What we do to the paddle here has no actual impact to the outcome, as that has the Particle Wall-Link model defined to the wall boundary (under Phase interaction number 4). Meaning, that as soon as the paddle DEM particle hits the drum wall, it sticks to said wall. The size of the paddle DEM particle is so that it is only slightly smaller than the radius of the drum, meaning that within one or two timesteps gravity will attach the paddle to the wall. The wall relative rotation then starts moving the paddle.

From the picture below you can also see the model selection for the DEM phases, a generic setup for the spheres, but a very much simplified setup for the paddle. The reason for this is that it has no merit to add forces like drag and such to the paddle, since its trajectory will be completely set by the wall relative rotation.

Also note the number of multiphase interactions specified. In a simulation with only one DEM-phase, we would generally only need to specify the DEM/DEM and DEM/wall interactions, but as we add an additional phase, the number of interactions quickly increased. As every DEM phase can interact with itself, we have two “self-interactions” specified (even though the paddle/paddle one is redundant since we only allow for one of those “particles” to be injected). We have the paddle/sphere interaction, and each DEM phase interaction with the wall (and again it is in the paddle/wall interaction, number 4, that the particle-wall Link is set up). This brings us up to five interactions. But since we also specify phase impermeability for at least one of the DEM-phases, we are allowed to create a separate set of interactions for the DEM and that “wall”. This allows for different “bouncing” against the inlet and outlet if we so desire. This brings us to a total of 6 multiphase interactions. It should be noted that it is actually possible to create a seventh interaction here, where the paddle interaction with phase impermeable is defined. But that is unnecessary for two reasons, the first being that the paddle never touches the inlet or outlet, and the second being that we never specified phase impermeability for the paddle-phase!

Pic2 DEM model interactions

The picture below shows to some extent how the injections are set up for the DEM-phases. The paddle is injected using a point injector and is limited in time to only allow for one particle to be injected. For the spheres the flow rate specification is set to dynamic, and the number of particles is limited to 200. Further a “box” is created and the type of injector for the spheres is a random injector. Meaning that the particles are randomly generated within the confines of the input part for the speres injector.

Pic2 DEM injections

This covers the basic setup made for the DEM phases. Further the simulation was run with a simulation history file that store the DEM data, and using the different states in a screenplay animation where we also include advanced rendering, we can see the particles mix and cool as they are moving around in the drum. Note that a limited simulation like this can be run in a fast manner on a limited setup. This simulation was run on a laptop (a good laptop) in under 30 minutes. The creation of the video using advanced rendering on the same hardware was around 6 hours.

 

I hope this has been useful to you and has shown the capabilities of using DEM and in particular the particle-wall link model. Keep an eye out for the template file for DEM simulations, it will be based on the same case as was used here, it will be found in the future together with its brethren under Templates – VOLUPE Software. As usual, reach out with any question to support@volupe.com.

Author

Robin Viktor

Robin Victor
+46731473121
support@volupe.com

Scroll to Top