[ameya.vjti]

Hello All,

I am trying to correlate cfx results with experimental test results for radial turbine.

I have got a fair no of sample size.

But what I observed in every case is,
CFX UNDERPREDICTS TORQUE ON TURBINE WHEEL AS COMPARED TO TEST.

What needs to be done to match the test results?
Is it inherent that CFX doesnt predict forces on wheel correctly?

Need your take on this....

[-Maxim-]

I have observed the same thing with propellers. The torque on the propeller is always underpredicted by CFX.
The pressure on the propeller blades look fine though.
So far, I haven't found a solution to that yet. I would be interested in thoughts from other CFX users as well.
Maybe there's a proper work around using the pressure on the surface?

[turbo]

The biggest drawback of any CFD is like this : you cannot adjust predictions as you wish in comparisons with test data. Very limited and not many options to play around.

Different turbulence models could give you some flexibility, because turbulence closure would be the last one that is open theoretically in CFD.

[ameya.vjti]

Originally Posted by turbo

The biggest drawback of any CFD is like this : you cannot adjust predictions as you wish in comparisons with test data. Very limited and not many options to play around.

Different turbulence models could give you some flexibility, because turbulence closure would be the last one that is open theoretically in CFD.

Yes I tried using different turbulent models.
Previously I used SST & later I changed it to K-epsilon
K-epsilon is predicting slightly higher torque than SST,

Actually I am intersted to match the output power in CFX & Test,
but observed that the power is underpredicted in CFX because torque on wheel is been underpredicted.

Do you think fluent can be more effective to my cause?

or any other suggestions in CFX modeling I can try to improve torque and power?

[urosgrivc]

Did you try diferent mesh sizes?
I had problems as results were changing with diferent mesh sizes near wals quite a lot.
I think that, what is hapening with the flow near wals is wery important for this kind of simulation, (it is most posible flow sepatation is ocuring and the mesh just cant resolve it) and are you including the total energy model?

I dont think k-epsilon is the wright way to go.-(if you want to get results for forces acting on the wals => moment)
I think k-omega should be better.
I would try SST model with inflation layers with Y+<1.
And if speeds of the flow are high I would use air ideal gas, and if they are high
you are gona be suprised at how thin your first mesh layer must be to obtain y+<1 and dont forget to stay with aspect ratio of the first element<200 also, that multiplies the mesh quite quickly, so the viscous forces are aded to the pressure forces.
A picture of the problem would help as I am just guessing.
Please corect me if Im wrong.

[turbo]

Originally Posted by ameya.vjti

Yes I tried using different turbulent models.
Previously I used SST & later I changed it to K-epsilon
K-epsilon is predicting slightly higher torque than SST,

Actually I am intersted to match the output power in CFX & Test,
but observed that the power is underpredicted in CFX because torque on wheel is been underpredicted.

Do you think fluent can be more effective to my cause?

or any other suggestions in CFX modeling I can try to improve torque and power?

First of all, throw away your thought that CFD can always predict test absolute performance. Think about all possible gaps between your CFD modeling and the real test situation. CFD does not consider any outside of your internal flowpath assuming adiabatic flow. Even in the limited internal flowpath, still we can see some other gaps.