I am working on a a finite difference model of a simplex atomizer. The crux of the problem I’m having is existence of a conical geometry converging at the orifice. I need to take advantage of axisymmetry to keep the computational time and memory requirements to a minimum. I was working in cylindrical coordinates and making progress until I realized that the geometry was causing considerable skewness in the nodal positions. I can handle this with Cartesian coordinates but then I have no way to approximate the angular velocity (normal to the solution plane) which was straightforward in cylindrical coordinates. Can anyone suggest a method to compute normal (tangentilal velocities?