Hi all,

I'm carrying out some simulations on a turbine blade.

During the results analysis I have found a value of total pressure slightly higher than inlet total pressure in some zones of the computational domain, probably due to numerical errors. These zones are placed into the wake where, I think, the turbulence model is more likely to fail.

The problem is that taking into account those zones leads to estimate a wrong trend of loss coefficient downstream the trailing edge of the blade (when I make the software calculate the average variables in correspondance to these sections).

The mesh quality is good, the turbulence model used is K-omega and the value of y+ < 20 (maximum placed in a region of hub).

I would like to know if:

- It's possible minimize this computational error - The problem could be the turbulence model and if, in that case, it's worth trying to use either SST-turbulence model(considering that I should refine the mesh until y+ =1 about, in order to make it effective) or other turbulence model - There are some other reasons for this simulation behaviour.

Thanks in advance.

Ermanno

Try refining the mesh in the regions where you are having problems with total pressure. It is likely a spacial discretization error if your case is well converged.

Hi,

Also, the SST turbulence model does not require y+=1. It runs fine with wall functions when y+>10. Try SST on your existing mesh before generating the finer mesh. As Johnny says there is a good chance your simulation is not yet mesh independent.

Glenn Horrocks

Thanks for advice, I'm going to try

Just one more question .

Generally speaking I used to use SST as regards K-omega when I don't want use wall functions but I want resolve boundary layer, in the other cases that I simulate I usually use k-omega.

As far as your experience concerned can using SST-model give better results than k-omega when it has to use wall function ?.

I used to think that in that case the model works like k-epsilon and k-omega model because near the wall is like a k-omega and in the other zones is like k-epsilon.

Am I wrong ?

Thanks

Ermanno

Hi,

Which turbulence model is best is problem dependant. The SST is a good general purpose model which can handle most flows from full boundary layer through to wall functions when used with the automatic wall functions. I recommend you test both models to determine the most accurate one for you.

Glenn Horrocks

SST is a good choice. It blends between k-omega and k-epsilon. If you have the right mesh then you will get integration to the wall using k-omega in the boundary layer.

If you don't have enough mesh then you get wall functions and something like k-epsilon where the mesh resolution is not high enough.

On your total pressure question. Total pressure can locally rise above an inlet value, both physically, and more often in CFD simulations. This usually happens when the eddy viscosity is overpredicted by the turbulence model, leading to artifical shearing/pushing of adjacent fluid layers.

Overpredicted eddy viscosity usually happens when you are not yet resolving the flow, so more mesh is probably required in the regions where you see the total pressure rise.