[wcj1n18]

Hello,

I'm intending on modelling sediment transport. As such, I have a rectangular channel with 1000s of hexagonally-packed spheres constituting the bed- these will all be stationary for now. Any advice on what the best means of meshing such a set up would be? So far I have tried trimmed mesh. This works well for the bulk of the channel but near the bed it becomes 'messy' (see attached images). I don't mind too much how difficult it is to do, but just want to ensure I get solution convergence to ensure the simulations are less computationally expensive. As an example, would people advise Immersed Boundary Method (though I don't think this is possible with Star-CCM+)? Is it possible to set up a trimmed mesh for the bulk of the channel but a polyhedral mesher for the part where the spheres are? I don't want to do DEM simulations as I will only be modelling on mobile sphere, the rest of them are stationary. For the mobile sphere I intend on using DFBI, overset mesh.

Any comments/ discussion are welcome

Thanks

Will

[pi120]

Hi Will,

Do you really need so many spheres?

You might need a better prism layer mesh surrounding your sediment spheres. I would at least impose 10 layers, with low prism layer stretching (1.1 or 1.2). Ideal would be to have the last prism layer cell half the size of the first cell outside the prism layer.

I would probably set it up all with the polyhedral or the trimmed mesher. But it is possible to have a polyhedral mesh for the sphere and trimmed mesher for the rest as well.
You just need to create two different regions, one for the low region close to the spheres and the other for the bulk.
To do so you need to separate the two bodies when modelling your CAD geometry. After that you need to generate the parts, and click on "assign parts to region", "create one region for each part".
Finally create two meshes, assign the bulk region to the trimmed mesher and the sphere region to the polyhedral.

hope this helps
Pietro

[wcj1n18]

Hi Pietro,

Thanks for the suggestions. I'll respond to each part separately:

Do you really need so many spheres?

I may well reduce this number later on.

You might need a better prism layer mesh surrounding your sediment spheres. I would at least impose 10 layers, with low prism layer stretching (1.1 or 1.2). Ideal would be to have the last prism layer cell half the size of the first cell outside the prism layer.

Thanks for the suggestion. I may try something like this.

I would probably set it up all with the polyhedral or the trimmed mesher. But it is possible to have a polyhedral mesh for the sphere and trimmed mesher for the rest as well.
You just need to create two different regions, one for the low region close to the spheres and the other for the bulk.

To do so you need to separate the two bodies when modelling your CAD geometry.

How do I separate the bodies? Do you mean when I create the geometry (which I am using Star-CCM+ for) I create two separate bodies that can be separate regions but combined as one domain? I shall look into this anyway

After that you need to generate the parts, and click on "assign parts to region", "create one region for each part".
Finally create two meshes, assign the bulk region to the trimmed mesher and the sphere region to the polyhedral.

Thanks for the help

Will

[pi120]

Yes you need to do it when creating the geometry. I use the split feature in SolidWorks usually so not sure how to do it in star but shouldn't be too hard.

I suggest you to generate your mesh on a simplified geometry (just 1 or 2 spheres), manage to get the right sizes for the boundary layer, and then apply them to the real thing.