Any thoughts on why the predicted slope of flow rate vs. pressure rise is shallower than experimental, yet crosses at or near the design operational point?

The Cfd model is 250,000 nodes, multiple frame of reference, total pressure specified inlet and static pressure outlet. The model is a single blade passage with tip clearance. The solutions are 'converged' and show balance across values...

Basically at higher flows I am overpredicting pressure rise and at low flows underpredicting pressure gain.

Any thoughts would be appreciated. Thanks.

(1). It is hard to know what you are doing. But one thing I can say is, the design condition flow field is normally smoother than the off-design conditions. (2). In other words, it is more difficult to predict the off-design condition flow field. (3). Because of the errors involved in prediction, the slope and the optimum design value will be different from the experimental values.

You are using multiple frames of reference. So what are you expecting?

Thanks for your input. I was hoping to see the pressure/flow slope similar to experiment, yet the curve offset either over or under... Have seen several papers where this has been the case with this particular software and model approximation. What I am getting is a shallower slope intersecting the experimental at or near design.

Mfr is a more or less strong simplification. Thats why it will work good in some cases and will completely fail in other cases. There are 2 articles about mfr on www.adapco-online.com ...

For turbomachinery calculations I always use mfr only to get an initial flow field for a sliding mesh calculation.

What are you using for a turbulence model? Many models don't work well for flows with streamline curvature, maybe this is causing you trouble.

George

Standard K epsilon turbulence model. Any suggestions?

Hi Erich,

Check out this paper, it offers a fairly simple correction.

Launder, Pridden, Sharma. "The Calculation of Turbulent Boundary Layers on Spinning and Curved Surfaces". Journal of Fluids Engineering March 1977. pg 231-239

The reason that k-epsilon may be at fault here is that it assumes isotropic turbulence, while the turbulence in your problem is likely anisotropic. The correction suggested here will not make k-epsilon anisotropic, but it is a 'fudge' that has worked well on previous problems. As a bonus it is also relatively simple to code.

Good luck,

George