incorrect results with a tetrahedral mesh
 

I am having some trouble getting the simple case of laminar flow over a sphere in a rectangular channel (3D) to give correct results on a tetrahedral mesh. This problem is only present on tet meshes, not hex. I've pasted an example of the results I get (in this case from potentialFoam, but all other apps give roughly the same result) plotted on a planar slice down the center:

http://www.bme.unc.edu/~masters/images/sphere.jpg

Uniform flow should be entering from the left, top and bottom (as well as front and back not shown) are walls, and outflow has a pressureInletVelocity BC applied to it. The results in all cases seem to converge fine. It doesn't seem to be attributable to discretization schemes or linear solvers, I've tried them all.

I am running OpenFOAM-1.5 on a 32-bit dual core intel machine loaded with Ubuntu 8.2. I also reinstalled version 1.4.1 (which I had used previously with success) but I found the same thing. I have tried all the meshing apps at my disposal (Gambit, Workbench Meshing, ICEM, Salome Platform) and have found the same thing with each. Has anyone else found this? Could this simply be a post-processing problem?

the same as ever:
- check Boundary conditions
- use pressure outlet
- create boundary layer
- don't use potential foam, use turbfoam. it's the most stable solver in my opinion

Thanks for your suggestions. I've tried several solvers (icoFoam, turbFoam, simpleFoam, ...) but the problem persists, BCs all look good (same simulation in CFX looks fine), outlet BC is pressure. I also had a more complex geometry with BL inflation and found the same thing. checkMesh gives me no errors. I've included non-ortho correctors to no avail. Using more diffusive discretization schemes (e.g. upwind) had no effect on the splotchiness of the results.

Hi Jonathan,

I am running thousands (literally!) of simulations of creeping flow around a sphere, and did not find anything like the results you show. My setup is similar to yours, although I have symmetry on front and back.

Maybe you could double-check your BC's, as suggested by Wolfgang.

Regards,
Jose Santos

Hi Jose

I agree, it's definitely possible. I've been using OpenFOAM for several year now and have had no trouble. I'm wondering whether it has something to do with my OS, although I don't know why it would. It works great with hex grids (I get very stable solutions for simulations of fluids with complex rheology) and should for any grid type.

Are you're simulations on tet grids? What OS are you running on?

Thanks

Jonathan

Hi,

I am working with meshes generated with snappyHexMesh, so I have a mix of hex, prisms and polyhedra. My OS is openSUSE 10.3 64 bits.

Regards,
Jose