Printer-friendly versionPDF version

When my kids were younger, one of the favorite entertainment options at children’s parties was the balloon artist. These artists were able to make anything from a monkey to a pirate hat simply by twisting balloons together. Usually, but not always, you could tell what the animal or object was meant to be, but it would take the most imaginative child to see the multi-colored lumpy lion in front of them as the real thing.

Whilst lions are not the most common shape for non-spherical DEM particles, should you have the need to model such a thing, previously your only option would have been the composite DEM particle and the result would have been much the same, a bunch of spheres of various sizes stuck together.

All of that is set to change with the polyhedral DEM particle in STAR-CCM+® software version 12.04. Now you will be able to accurately model real objects, putting the corners back into your particles, by building or importing a realistic representation as a geometry part which then forms the basis of your particle. For many objects, this new polyhedral particle is less computationally expensive then a composite particle which can require many spheres to get close to a realistic shape. Polyhedral particles also provide a more efficient solution, reducing simulation time.

So, to an example. Let’s imagine we want to find the angle of repose for a particular type of gravel. We can do this by filling a cylindrical sleeve with particles and then sliding it upwards to release the rocks. First, we need to generate a part to represent a typical rock and then link it to the DEM phase. The part and the resulting polyhedral particle are shown in the left image below:


If we compare the equivalent composite particle representation of the rock, created with 10 spheres (right image), we can see that we lose the sharp edges and produce something circular in cross section. The polyhedral particle on the other hand is as accurate as the part definition we use to represent it.

If our rocks are of various sizes and shapes we can simply set a distribution of sizes and elongations on the injector to represent a population. The geometry part represents the topology, not the actual size and shape. If different topologies are required, multiple phases can be modeled.


As this example shows, the real benefit of polyhedral particles is not for smoothly curving objects, which could be represented fairly well with composite particles made of spheres, but for modeling objects with sharp corners or high aspect ratios. In these cases, composite particles would need a huge number of spheres to come close to a reasonable representation.

Consider another example, a hexahedral nut. With 20 spheres, the composite definition still does not come close to capturing the detail of the corners, whereas the polyhedral particle models the correct shape with the contact model considering the edges.



 So, polyhedral particles provide us with a more accurate representation, but at what cost? To assess this, consider the triangular prism particles below represented as both polyhedral and composite particles. The polyhedral particle in this example is comprised of 20 spheres.



For this comparison, we used a rotating drum with 4,000 particles being mixed as the drum rotates, similar to the hexagonal nut example shown above. The computational time for 100 timesteps on one core is shown below for both the polyhedral and composite definitions. As you can see, the polyhedral particles give around an eight-fold speed-up over their composite counterparts for this example.



When making this comparison we need to consider that the cost of the polyhedral particle is a function of the number of faces that the underlying part has, and the cost of the composite is a function of the number of spheres. With this in mind, when trying out polyhedral particles for the first time, use a coarse tessellation for the part in question and refine later if required.

Polyhedral DEM particles are one of those rare features that bring both speed and accuracy improvements and provide a great new addition to the DEM modeling capabilities in the upcoming release of STAR-CCM+ v12.04. To see what else is coming your way bookmark our blog page to receive all the latest updates.


Matthew Godo
STAR-CCM+ Product Manager
Stephen Ferguson
Marketing Director
James Clement
STAR-CCM+ Product Manager
Joel Davison
Lead Product Manager, STAR-CCM+
Dr Mesh
Meshing Guru
Ravindra Aglave
Director - Chemical Processing
Karin Frojd
Sabine Goodwin
Director, Product Marketing