CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   Separation (

An Modh Coinniolach March 27, 2002 11:33

I am getting an area of separation on the pressure side of this axial turbine I a modelling (3-D). I am using a mixing plane model and the k-e turbulence model with wall functions. I am using a hybrid grid; hex cells near the blade, tet cells everywhere else. The separation is confined to an area at the tip of the blade.

The CFD results for torque compare poorly with experimental values.

My question is this: how well does Fluent cope with separated flow? Is is possible that I will always get questionable results for torque with an appreciable area of separation on the blade?

A different version of this blade (Different hub-to-tip ratio) has been modelled previously in CFD in my group. Torque values were spot on (in comparison with experiment) and there was no significant region of separation seen in that blade. It is my intention to model this geometry aswell and see if the results of that simulation are any better at predicting torque. But in the meantime, I wonder does anyone have any views?

kim March 28, 2002 03:33

Re: Separation
Prediction of separation strongly depends on turbulence model used. For really correct determination of separation point you should use k-e with two-layer model or k-omega model (Jonas mentioned good experience in one of his previous contributions). As for torque: I also have problems with estimation of moments on blades. I think it is the weakest point of turbomachinery CFD simulation. What are your experiences with torque prediction, CFD guys?

An Modh Coinniolach March 28, 2002 07:09

Re: Separation
I have previously modelled this using both k-e and k-omega using the two layer approach and I got similar results.

The turbine is quite crudely designed and this area of separation begins immediately at the leading edge due to a significant mismatch in the blade inlet angle and the velocity vector of the approaching flow.

I would have expected a properly articulated model to predict torque very well and losses poorly and I have been tending to look at torque prediction as a barometer of the health of my simulaton.

david March 30, 2002 20:56

Re: Separation
Use DNS !!! DC

An Modh Coinniolach April 1, 2002 11:31

Re: Separation
I'll take a walk over to the Dell factory and see about hardware.

Thanks bud!

All times are GMT -4. The time now is 10:03.