CFD on CT-scanned domains
 

Dear CFD Online community,
I'm investating the usability of OpenFOAM for calculations on CT-scanned samples, and I'm new to this software package. The samples are usually voxel images, of a size that can reach 10^9 cells (about 1000^3). The resulting amount of data is pretty big already (about a GB) and I'm afraid turning this into an OpenFOAM mesh might be a bit of a data inflation nightmare, even if I use very simple, cube-based structured meshes. What is your opinion about this?
VL

Hi Vincent,

My suggestion would be to create an STL surface file of your CT geometry using a software such as 3D Slicer.
Then you could use snappyHexMesh (OpenFOAM mesh utility) to generate your mesh from this STL.
The final size of your mesh (i.e. number of cells) will depend on the mesh resolution you specify with snappyHexMesh and also the geometrical feature size you wish to capture.

Best regards,
Philip

Dear Philip,
Thank you for this quick reply. I have a subsequent question: given the large amount of cells (hundreds of millions to a billion), wouldn't RAM shortage be an issue while generating the mesh? A user reported issues with meshes of this size, and since I am not trying to refine a mesh, I cannot use the workarounds proposed in that thread.
VL

What is it specifically that you are trying to mesh? Some kind of arterial system? I don't think snappyHexMesh would be the best opportunity here. You have now control on how your control volumes would be aligned with the flow, which you want if possible. I know there are some dedicated meshing tools that carefully mesh branches etc from STL. I would recommend you to look into that.

Thank you for the advice. The aimed geometry is a porous medium (some kind of composite for thermal protection). I actually believe that even with a suitable meshing program, the amount of data is going to be quite large. Like several GB big. I'm afraid it'd be too big for my computer to handle when running the simulations (it's got 256 GB RAM, 80 cores).