[Sasquatch]

Dear Users,

I have not found a thread that treats the above question therefore, I would like to know your opinion on that.

If the computational time is affordable for both, which solution provides more physical results? Simulation of the Radial Outflow ORC Turbine, 10 cm long, 3 stages, prismatic blades.

My understanding:

Frozen Rotor:
- modelling errors rise when the flow is highly unsteady
- rotating component is not included, losses would be bigger for transient case

Mixing Plane:
- I have read that: averaging between passages accounts for time average interaction effects (could sb please explain this)
- it is not recommended if the circumferential change on the flow is large wrt component pitch (I think my case, as transient result showed)

Both do not include transient effects at the interface.

Early Frozen Rotor results showed a good wake impingement and it gave me a hint that the results looks 'more physical' but I know it can be insidious.

The final goal is to have at least a preliminary loss estimation of the machine.

[turbo]

That depends on what your priority is.
If you look for an aero design validity only at design point, one-pitch segment mixing plane is enough for a steady-state solution in your machine (I guess you have multistages of rotor/stator rows because you mentioned "radial outflow ORC turbine"). If you want unsteady interaction between rows, you will need the transient interface once you have steady solutions from the frozen rotor.

[Sasquatch]

If by "aero design validity" you mean also loss mechanisms (to some extent at least, since it is still RANS for now) yes.

basically I would like to know the total to static eff. of the machine and from your answer I conclude that the results from the frozen rotor as such do not have a good qualitative meaning and can only serve as init. conditions for a transient case.

I did that too.

But conversation with Ansys Support gave me a hint that frozen rotor can be more accurate if confirmed by transient (a bit of chicken and an egg problem since you need both but still...)

What do you think?

[turbo]

Think about your performance testing of hardware. Are you going to measure unsteady pressure ratio and unsteady efficiency? Of course not. Everything you get from the testing will be time-averaged after all after waiting until test rig stays at steady state. That's why I said the mixing plane approach is enough and almost perfect if you want a comparison with tests. The frozen rotor is created mostly for a special case where there is a large mismatch of pitch length between rows, like a volute and one-pitch blade row.

[Sasquatch]

Thank you for your reply, I think the mixing plane will be of course, the most comparable with the tests.

My problem is that the mixing plane simply fails because the flow is strongly unsteady and the distance between the blade rows is small - averaging over the circumference simply does not handle the shocks and separation regions and results in a unphysical red spots.

I indeed have a large pitch change in the machine, such as 1.3-1.4.

Due to lack of computational resources now, I will simply stick to the mixing plane. For the 1st stage at least, I will compare with the quasi-3D, transient case (still feasible) and see how well mixing plane deals here.

Frozen rotor of the full machine (163 blades) gives indeed a special case but there is many un-periodic configurations of the blades in such simulation.

163 blades and each of them in relatively different config, therefore I thought it is interesting to perform it, especially because steady state case, even on a 16 million cell is computationally feasible for me now.

Do you think it is interesting to make such comparison? Or maybe from the other side, would you even say that it is a waste of time?

[turbo]

If you judge the flow is strongly unsteady, and if the gap between blade rows is extremely small, you do not have many choices but to go with the transient rotor-stator interface starting from steady-state fronzen rotor, because a strong unsteady flow interaction exists between rotor and stator. Huge computing resource will be required.

If I were you, I'd choose a better machine design configuration because you will not be able to achieve higher efficiency after all due to the dominant unfavorable byproducts. It would be better for you, too, if you strive to be a CFD engineer rather than CFD technician.

[Sasquatch]

Dear turbo,

Changing the machine to have a pitch ratio = 1.0 and being able to apply periodic conditions and hence transient sim. at lower cost is possible but then, it is simply not the same machine anymore.

But since we talk about this, from the stage load I conclude the following modification to unload/load some passages for now:

Blade numbers:
1stator: 12
1rotor: 24
2stator: 31 --> 24
2rotor: 37 --> 36
3stator: 30 --> 36
3rotor: 29 --> 36

I understand both mixing plane and frozen rotor assumptions. I was just interested whether the wake effect on the downstream component performance at a single relative position for the full circumference (no periodicity) would be worth analyzing.

Although you did not write it, I conclude from your message that it is a waste of time and 'unprofessional'.

[turbo]

That also depends on what your boss wants to hear. If it is just what will happen with this design, you can arbitrarily modify the machine solidity, as you said, within your computing resource. However what would be you and your boss' takeaway from a different machine simulation?

I guess the extremely small gaps would come from the radial OUTFLOW turbine due to wanting to keep the disk diameter as small as possible. Nature is so fair in terms of pros and cons. That is my favorite saying to people.