Scanning and simulating unique particle shapes: an experiement

So you’ve identified that your Rocky DEM analysis could benefit from a more accurate particle shape. You say you have access to a sample particle and a 3D scanner? Excellent! So now what on simulating particle shapes?

The office I work in recently received our own 3D laser scanner so I took it upon myself to test it out.

After scouring the landscape around my building, I identified a rock that was “interesting” enough to suit my needs, in the sense that it accomplished all of the following feats:

  • Wasn’t easily represented by the default shapes in Rocky
  • Contained some concave faces.
  • Was flat enough on one side to sit on the scanner bed.
  • Was a size appropriate for the scanner itself.

After I appropriated my interesting rock sample (Figure 1), I acquired a second interesting object from the local grocery store: a shelled pecan half (Figure 2).

Figure 1: Photo of the rock. Figure 2: Photo of the pecan half on a clay mount

I then scanned each object and exported them to .stl files, as required by the Rocky DEM software, using two different output options to control the accuracy: Octree depth and degree of fit.

Octree Depth

The octree depth option basically determines the size of the many (smaller) individual cubes being used to measure the object relative to a single (larger) cube which encompasses the entire object. There is a lot more literature about octree depth and how it is programmed but a good place to start is this example using the Stanford Bunny.

Octree Degree of Fit

This option determines the degree of the polynomial fit of the octree points defined. The settings range from a quadratic to a quintic fit. There were minor variations in the pecan .stl file due to the degree of fit setting. This was due to the amplitude of the surface variation being high relative to the size of the pecan itself (Figure 3).

Figure 3: Close-up image of the pecan’s octree levels

I then experimented with these settings to come up with a resolution matrix showing that the higher the number of vertices and faces an object has, the better the resolution (Figure 4).

Figure 4: Sample object resolution matrix for varying amounts of vertices and faces

The particular scanner I used showed some limitations with very complex shapes, as highlighted by the lack of dips or depressions in even the highest resolution scan of the pecan (Figure 5).

Figure 5: Actual image (left); location of depressions (center); high-resolution scanned object showing no depressions (right)

Additional degrees of freedom would be required in the scanning mechanism to accurately capture the intricacies of a surface feature that has a large depth relative to the size of the object scanned. As you can see from the photo of the scanner device I used (Figure 6), this particular model allows the object to be rotated and the capturing device to be moved up and down. This provides for great capturing of relatively round objects, but if they are too convex some limitations may be observed, or additional creativity may be required to capture a realistic image.

Figure 6: The particle scanner used in this experiment

The advantage of having accurate shapes like those I scanned here is that the particles will behave more realistically and simulating particle shapes.

That is to say, the physical shape will resist rolling or will roll naturally as opposed to using a spherical representation and then conducting an experiment to determine the rolling resistance. In addition, the contact forces between both the boundary and the other particles in the simulation will be more accurate due to more realistic contact surfaces, so there is no need to artificially modify the friction factor of a sphere to match experimental results.

The disadvantage of using scanned shapes like these is that higher resolution particles with greater numbers of vertices and faces will necessarily take more time and effort to generate, so a balance must be reached between accuracy and processing time. In general, it is good practice to aim for as few triangles in your design as possible that still results in a uniform mesh with good shape representation.

Below is an example of a pile I made in Rocky using the pecan-shaped particles I scanned in (Figure 7). What are some examples you have identified for using more accurate particle shapes in your industrial applications?

Figure 7: Pecan pile simulated in Rocky DEM

Brent Nelson

Brent is Application Engineer and has over 5 years of experience working with various simulation softwares for different industry sectors. He earned his Bachelor’s degree in Mechanical Engineering from Rose-Hulman Institute of Technology and his Master’s degree in Mechanical Engineering from Texas A&M University.

Related posts

Leave a comment

Get Fresh Updates on Email