Criterion on the generation

[adrienluyckx]

Hi,

I was wondering if a different criterion could be implemented to determine when the clump generation stops—perhaps based on reaching a specified relative error threshold for volume or moment of inertia?

Best regards,
Adrien Luyckx

[vsangelidakis]

Hi @adrienluyckx,

This is feasible and you can adjust the criterion to your liking. We have already tried this in a few studies, e.g. see Fig. 2 here: https://doi.org/10.1680/jgeot.23.00162, where we monitored morphological characteristics for increasing numbers of spheres.

Since CLUMP is a generic tool, we do not plan to hardcode a very specific criterion such as this one within the clump generation functions, but this is something you can either (1) tweak in the clump generation functions yourself or (2) perform in your example script.

For instance, you can generate clumps with increasing numbers of spheres, use the ExtractSurface function to extract the surface of each clump in the format of a surface mesh, and then use the RigidBodyParameters function to get the volume/moments of inertia or you can use SHAPE to calculate a shape parameter of interest. Then you can decide which number of spheres is appropriate depending on your criterion.

Which clump generation technique are you employing? Are you using the MATLAB or the Python implementation of CLUMP?

Hope this helps,
Vasileios

[adrienluyckx]

Hi,

Thanks for your work on this! I have implemented my own version and am currently experimenting with different techniques in Python.

While testing the code, I noticed some inconsistencies. Specifically:

• In the implementation of ExtractSurface, the parameter clump is described as a "clump" object, but it is actually checked as a dictionary immediately after, e.g.:

if isinstance(clump, dict): #This behavior seems inconsistent with the description.

• In the utils module, instead of explicitly passing Vertices (P) and Faces (F), it might be smoother to allow the function to accept meshes directly, making it more user-friendly.

Thanks

[vsangelidakis]

Hi Adrien,

In the implementation of ExtractSurface, the parameter clump is described as a "clump" object, but it is actually checked as a dictionary immediately after

You are correct about this. A Clump object is not exposed to the user level at the moment, i.e. it is fully contained within the GenerateClump_* functions, as exporting a clump as a dictionary was considered more versatile when we developed this. Feel free to send a pull request if you want to change this and offer the additional option to have a Clump as input for the ExtractSurface function (I would still keep the option of providing a dictionary as input, and would amend the docs).

In the utils module, instead of explicitly passing Vertices (P) and Faces (F), it might be smoother to allow the function to accept meshes directly, making it more user-friendly.

This is a good idea, thanks. We can add another option to load meshes directly. Do let me know if you would like to send a pull request for this yourself, that would accelerate things.

Feel free to reach out with any other ideas or pull requests, happy to discuss.

All the best, Vasileios