test data, you may find the streamlines and veocity profiles from Re=1 to Re=25000 by FEM method on unstructured mesh from here: <A HREF="http://www.mema.ucl.ac.be/~wu/cavity2d/cavity2d.html">http://www.mema.ucl.ac.be/~wu/cavity2d/cavity2d.html </a>, although i don't have enough time to write some background of this benchmark testing. If you need the result files at different Reynolds numbers for comparison, just drop me mail with the specfied Reynolds numbers you need (only tecplot binary PLT file format available).

Nice work.

Just a thought regarding your 'unstructured' mesh. Looks like a stretched 'structured' mesh to my eye. The structure, although stretched, looks like a typical structured triangular FEM mesh.

diaw, you are right, from the topology sense, this mesh can be treated as a structured one, however, for an unstructured Navier-Stokes solver, it's not important that the topology of the triangular mesh is structured or unstructured. Actually, I once generated unstructured triangular meshes for this benchmark test by my own code mesh2d (now as a build-in unstructured mesh generator of commercial product FEMAG), however it increased the size of mesh dramatically when highly refined mesh along the boundary and near the wall is expected due to the isotropic algorithm of mesh2d. That's why I wrote of another small piece of code to generate this "structured" mesh. However, I already have very good idea/algorithm to generate ansitropic meshes and it should be ready soon.

Finally, to those who asked for the data, please specify the Reynolds numbers you want. If there are lots of requests, I will simply give a link on that site.

To Jonas: I will contribute to the your CFD-wiki validation and test cases area if I'm a little bit free, maybe after come back from vacation at the end of this month.

Cheers,