CFD Online Discussion Forums - 2-D solutions in cfx???

In CFX axisymmetric calculation is performed by simulating a wedge containing a single layer of cell. The thickness of the wedge increases as you go away from the axis. This makes the area of the cell faces and the volume of the cells increase as you move away from the axis. This increase in area is crucial! It is the effect of this area change that makes the solution behave as axi-symmetric, instead of just 2D. (If you've done a university maths course you might remember all the r and 1/r terms in the Laplacian in cylindrical polar coordinates. The increase in area of the wedge with radius plays the same role in CFX's numerical method as the r and 1/r terms play in the Laplacian)

So, why a small angle?

Well, imagine a 90 degree wedge with one layer of mesh on it (Note only one layer as you want a single layer of cells in a 2-D mesh). It would look like a big triangle from 'above' (there is only one layer of mesh, so the curved surface would become a single straight edge). The areas of the mesh would still increase with radius, and so would the volumes, but not in quite the right way - the ratio of volume/area for each cell face would be wrong. The ratio would be different by a factor of sqrt(2) from the ratio in a real curved wedge. This would cause errors in the CFD results.

Instead, now imagine a 5 degree wedge with one layer of mesh. Now the difference between the perfect curved wedge and the 1-layer mesh is much smaller. It's not 100% perfect, but it is close enough to get good CFD results.

If the angle is <5% the error is even smaller.

So, you want to be working with the 5 degree wedge instead of the 90 degree wedge to get good results! Infact many people recommend using less than 5 degrees to keep accuracy. I don't know what the best angle is, but I use 1 degree often myself.

Regards, andy