Dear colleagues of CFD community!

While dealing with rotaing flows in cavities I found the following problem: dependence of the velocity radial profile between two rotating concentric cylinders on viscosity. That was very syrprising for me, because there is well known analytical solution of such problem (see for example H.Schlichting "Boundary layer theory" chapter "Exact solution of Navier-Stokes equations"). Model problem setup: 1)Geometry: inner cylinder diameter=0.05m, outer cylinder diameter=0.1m; inner cylinder angular velocity=2radian/s, outer cylinder at rest. I consider laminar incompressible air flow in CFX (obtained velocity radial profiles for different viscosity values see in attached picture). 2)CFX implementation: stationary reference frame with rotating inner wall (r=0.05m) and stationary outer wall (r=0.1m). Laminar flow, air at constant density 1.225, isothermal solution at T=300k, refernce pressure = 1e5Pa.

Could anyone help me to find out what is wrong?

www.ImageShack.us" /></a>

Have you tested this effect for grid independance?

Are you getting this effect on counter-moving flat plates?

I would suggest artifical/numerical viscosity as a possible cause. A grid independance study and discretisation sensitivty study would help to establish this.

Also test sensitivty to a double precision solve.