Hi I am trying to rotate a cylinder in a pipe flow using Ansys 8.0, but so far I´ve been unsuccesful. At this point I´m having some dificulties with the assigment of the boundary conditions for this particular problem. Can anybody give me any clues?.

Thanks

Tell us more.

2D, 3D, interested in transient or steady state?

If steady state, are you solving the transient to get there?

Compressible or incompressible? What rotational speed, fluid axial velocity (laminar or turbulent)? Typical dimension?

Thanks Jim for your time, Actually we are modeling a 2D model in steady state, laminar and incompressible fluid. Low Reynols number.

Density: 1.2 kg/m^3 Viscosity: 1.5 N-m/sec in SI system, the fluid is glicerine.

Vy=0

Vx=0 �¨

____________________________________________

CYLINDER TO ROTATE Vx=0.1 �¨ �ª

O P=0

_____________________________________________

Vy=0

Vx=0 �¨

The dimensions of the system are L=0.5 m H=0.1 and the radius of the rotating cylinder is 0.025 m. How should we handle the displacement boundary conditions, are they really neccessary?

Now I started an investigation using ANSYS, but my knowlegde in CFD is still very low. I am a civil Enginner an I am interested in learning CFD, can you tell me please when can I get a good information about it.

I've 'playing' with trying to simulate a 2-d (axisymmetric) flow - incompressible, laminar myself.

I'm using a code based on the sola series originally from Los Alamos. I've had a great deal of trouble getting convergence (using the code's natural transient orientation to approach a steady state). I've had to add the angular momentum equation and that involved a lot of choices in the difference equations used.

First order of business is to find very simple rotating flows that have analytic solutions. Schlichting's boundary layer book is where I started. When the code will do these and get the same answer, that's advancement. Making the code get those answers is the learning experience. And, in my experience, it's the addition of the angular momentum equation (rotation) that makes convergence difficult.

You might try very simple (but definitely rotating) problems in ANSYS. I know, it's like using a sledge hammer to kill a fly. But you'll learn something about using the code from the experience.

As to BC's, all three components of velocity must be 'no slip' at the solid walls. If the outer cylinder (pipe) is not moving, then the velocities are zero there. On the surface of the rotating cylinder, the angular velocity of the fluid is the same as the angular velocity of the cylinder's surface. The radial and axial velocities are zero there as well.

I'd try to specify the inlet velocity components.

The outlet velocities are 'continuitive' - you assume the boundary is downstream far enough that the velocites don't change with distance.

These are pretty general suggestions - any refinement would depend on the details of your problem. Are you trying to simulate a physical experiment. If so, the BC's should match that as closely as possible.

But I'm rambling ...

Good luck.

Hi Jim,

How do you deal with the sudden variation of the wall velocity from the non rotating part to the rotating part of the cylinder ? The circumferential velocity is then discontinue (as the total temperature) inside a portion of the boundary layer.

Thx

Kern

I've been dealing with concentric cylinders, rotating cylinder in a pipe.

I haven't tried the problem of a rotating cylinder 'butting up' to a non-rotating cylinder. In the sola code differencing system, the velocities are specified midway along the cell sides, so numerically there's not a jump discontinuity where the two cylinders meet.

But there is a pretty big change between the boundary condition for neighboring cells across the junction, so it wouldn't be a shock to find numerical convergence problems occurring in that region. I'd fuss with the mesh size in the area and expect to use a smaller time than the usual stability criteria allow. It might be necessary to start with the two sections rotating at the same speed, then slow the fixed one down to zero velocity over several (hundreds?) time steps.

Sorry for a "guessy" answer, but that's all I have until I try the problem. Have you worked on that problem?

I worked a little on that problem. I used an explicit compressible time marching solver. Both laminar and turbulent flows (with simple mixing lenght model) were tested.

The major problem for me was the "discontinuity" in the solution (physically) when passing from the fixed to the rotating cylinder. I tried to relaxed this by applying a linear wall velocity variation from 0 to the rotating wall velocity on a short distance before the rotating wall. This was not fully satisfaying on the physical point of view but i experienced no particular difficulties for converging.

K