[andreasp]

Hi everyone!

At our institute, we work on packed bed chromatographic separation processes of large molecules (e.g. proteins) solute in liquids. From the modelling point of view, one can think of a cylindrical column which is filled with randomly placed spheres. While the mixture moves through the column and diffuses into the spherical particles, proteins are adsorbed.

For detailed 3D simulations of such a process, we usually use a standard FEM approach, which handles the spheres and the interstitial volume separately. That is, these two domains have to be meshed independently. The solution process then considers different types of equations (convection+diffusion vs. diffusion+reaction) in the domains and accounts for a boundary condition between them (film transfer).

However, the main problem with this domain decomposition approach is that meshing of the interstitial volume can become extremely difficult. Very narrow gaps between the spheres either enforce extremely fine meshing or they can prohibit simulation altogether due to badly shaped elements. In any case, a lot of unpleasant meshing/testing/remeshing is required.

Now my question is if there is a smarter approach to solve these problems. Could the Extended FEM (XFEM) be a suitable alternative? Maybe, is it possible to mesh the complete column volume regardless of the different domains, solely accounting for them by discontinuity conditions? Or could finite volumes easier handle the problems (although I'm pretty sure that the meshing problem would exist just as well)?

I am completely new to this, so I would appreciate any hints and suggestions very much!

Thanks in advance
Andreas

[Hamidzoka]

you can try cartesian meshing technique.
it may increase cell numbers considerably compared with tetra mesh.
but mesh distribution is more controllable and high quality meshing will be achieved.
you can refer to "Dawes" artciles for this case.

regards

[jchawner]

Andreas:

Hamid's suggestion of Cartesian grids is a good idea.

Depending on how much flexibility you have on the solver side, use of overset (aka Chimera) meshes might be interesting.

Now for the shameless plug. We've helped two customers recently solve meshing problems similar to the one you describe: one pertaining to voids in a foam the other to densely packed spheres in a fluid. Feel free to email sales@pointwise.com to talk through these experiences.

Best Regards