generating meshes with subdomains

[MariusCausemann]

Hi,
for my applications, I need to generate meshes with subdomains, e. g. a ball in a box. Is this currently possible with bichon or could it be implemented with reasonable effort? Thanks!
Marius

[jiangzhongshi]

Hi Marius,

That's a good question: while it is not difficult to extend to multiple domain, it is not implemented.

However, if you only want one object within a box: see figure 26 of our paper, then it is possible with small modifications.

[MariusCausemann]

I've seen figure 26 in your paper, it looks like a good starting point! However, I probably need a more general support for multiple domains. Is that something you are planning to look into in the near future? Or alternatively, guide me a bit if I decide to work on it myself?

[jiangzhongshi]

Hi Marius,

Unfortunately, I would not implement it in the near future.
However, extending it to multiple enclosing shapes, and each domain is bounded by manifold meshes, it will be a small modification on the export.
But if you have non-manifold interfaces (whenever more than two materials share an interface e.g. https://youtu.be/JY0kozIdIQo?t=572 on the right) they will be non-trivial

[MariusCausemann]

The meshes that I'm thinking of are similar to this one:

Except both the outer shape and the inner embedded objects are more complex. So I guess being restricted to manifold interfaces would not be a problem?

[MariusCausemann]

Hi Zhongshi,
could you maybe quickly comment on whether you think this should be possible or not? I'm wondering if the points of contact of the material interface with the outer surface are a non-manifold interface, which would make it complicated? Or could I get away with a small modification?

[jiangzhongshi]

Hi Marius,

Unfortunately, I think the non-manifold interface does complicate things, and I don't have a quick code fix for that. Would love to know better the specific application and maybe there could be workarounds?

[MariusCausemann]

Hi,
I'm interested in finite element simulations of different physiological processes in the brain on the cellular level. So for meshing, I usually start with segmented electron microscopy images, generate stls describing each individual cell and then use tools like fTetwild of CGAL to generate meshes of the cells embedded in the (usually fluid-filled) extracellular space. Usually, the cells don't touch each other, so each domain is bounded by a manifold mesh, except for the outer boundary of the whole domain.
Here is an example of a mesh, with just the cells on the right and including the fluid around it on the left.

If only the non-manifold interface with the outer boundary is a problem, it might be a workaround to extend the outer boundary, such that it is not in contact with the cells anymore?

[danielepanozzo]

Hi @MariusCausemann,

this is an interesting use case, and we are working on a similar problem for a different type of images (microCT). Can you reach out to [email protected]? Would be happy to discuss this in more detail and see if we can add this feature to our todo list for the next iterations of fTetWild/bichon.

Daniele
https://cims.nyu.edu/gcl/daniele.html

[MariusCausemann]

Hi @danielepanozzo ,
I just dropped you an email, meshes based on microCT images are probably very similar to the EM images I'm working with!