Turbine stage: Transient RANS simulation - stage interface problem
 

Dear fellow users,

I am modelling unsteady flow through small-scale radial outflow turbine. Due to computational resources I am limited to RANS, mainly 2-eq. models, such as k-omega based SST.

Number of stator blades: 12 (pitch 30 deg)
Number of rotor blades: 24 (pitch 15 deg)

I have performed mesh sensitivity and found my mesh densities with Y+ below 1.

Mixing plane gives a little unphysical behavior at the interface (reflections and too strong shock wrt single stator simulation) between domains - hence I use it only as initiation for the transient case to resolve the higher order terms.

Thus the pitch ratio is 0.5 which already rules out Time Transformation method but since 12 and 24 can match each other (1 stator & 2 rotor blades) maybe it is not even necessary and my plans to use Fourier Tranformation are too off...

I am suspecting the mistake in either the Time Step, either the periodicity definition (or other interface definition) but the same happens for the automatic timesteps, see picture:

https://s31.postimg.org/60d127hwr/mach_edit.png

or

http://https://s31.postimg.org/60d127hwr/mach_edit.png

As you can see, per one stator domain there are 2 rotor domains. The wake from the trailing edge matches on one of the rotor domains but due to periodicity it is also on the second one which is not physically true. the wake should be simply translated on one and the other one should translate the other phenomena, such as the shock...

The rotational speed is 43092 [rev/min]

I have computed my timestep as:

Period_of_stator/100

where: Period_of_stator = 1/stator_frequency =~6.96E-04

Could you please help me track the source of my error?

You are observing reflections and a strong shock in your stage. What version of the software are you using ? There have been recent improvements on the mixing plane model that deal with reflections back upstream as well as dealing with shocks.

If I recall correctly, ANSYS CFX R17.x mixing plane default behavior includes some of those improvements. However, additional settings may be needed depending of the complexity of the shock/reflections and I will advice to contact ANSYS CFX support personnel for the details.

Before attempting a transient simulation, I would advice to understand why the steady model is not good enough. Transient simulation are not cheap, and they should only be done when the benefit outweigh the cost. For your blade count, you can get away with 1 stator passage and 2 rotor passages, and there is no need for a pitch change model. The blade count is very simple, and does not include prime numbers; therefore, a very small accurate model can be setup and solved without any approximations.

Thank you for your comment, I will contact Ansys outright. Answering your question, I use Ansys 16.2.

I am also familiar with nonreflective BC - worked well with static pressure at the outlet for smaller, quasi-3D meshes (sometimes crashed for larger meshes) - unfortunately I didn't see the option to apply nonreflectivity on the interface.

If I go for a mixing plane approach, I have:
1) implicit pressure avg (beta)
2) implicit stage avg (beta)
3) as a downstream velocity constraint i used constant total pressure as it was recommended by users

I am afraid I have checked all though, with similar result. The machine is very small - the gap is 1mm - this might be a strong reason for it.

best regards

I have tried this (the pictures show it clearly I hope):

1) A stator outlet has a mixing plane with 2 rotor inlets
2) between the rotors there is a general connection but every 2 rotors there is a Rotational Periodicity condition (I am not sure about this)
3) Rotational Periodicity on the stator remained the same and it is obvious.

Please find my results on a relatively coarse mesh (test).

I am particularly hesitant about the 2)

a) Mixing Plane Model:
https://s32.postimg.org/6emmqelfp/mi...lane_model.png

b) General connection between stator (left) and rotor:
https://s31.postimg.org/o3vgpsnzf/ge...ection_1_1.png

c) General connection between the rotors ??? <-- not sure here
https://s31.postimg.org/o0wu7rqh7/ge...ection_1_2.png

d) Periodicity for rotor <- not sure neither
https://s32.postimg.org/4yzexpub9/Pe...or_section.png


My results:

e) Steady-state, coarse mesh to test the model, y+ <1
https://s32.postimg.org/dxjnfgg91/mach_steady_state.png

f) Transient:
https://s31.postimg.org/69q9qazcr/mach_transient.png

g) Transient, zoom in:
https://s32.postimg.org/5h04z7c05/ma...ent_closer.png


I know it is a of pictures but since Mach is non-dimensional, it can explain a lot here.

Best regards