Hi i'm gonna to simulate the flux around a rotor turbine blade using multiple reference frame. Grid has been created using Gturbo, so I think i could set the casing wall as fixed in the stationary frame; hub wall, bladetip, pressure and suction side walls of the blade fixed in the rotating frame. The doubt is on the fluid zone: should i set it fixed in the moving or in the stationary reference frame?

Thank you so much.

Don't use MRF technique, it is not good. Transient moving mesh is correct way to go. Read some papers on the subject. MRF is only good to use as an intial field for moving mesh.

Dear He,

would you mind listing references to those papers you are referring to? Is it for any general rotating case, or for particular configuration like axial machines? I can see why is not good for axial machines, where the main flow is along the axis..

Moving mesh discretization requires an exact geometric conservation law, or you will be creating spurious numerical sources througout the mesh.. Then, higher order transient discretization is another problem..

In your opinion, which one is the lesser evil, MRF or TMM?

Thank you in advance, Opaque..

Why shouldn't i use MRF? In Fluent manual i read that this technique is reliable in steady flow situations, which i want to simulate on the rotor flux. Thank you for answers.

It is true, moving reference frame (MRF) analysis yields to wrong analysis conclusions all the time.

To compare with experiments the data from RANS-MRF are time-averaged but many young cfd engineers do not add the 'location-averaging' of the rotating part bfore comparing and it is not comparable because the geometry is moving. This is one main mistake.

Then when you compare, RANS+location-averaged with time-average experiment, it is off 90% of the time.

The reason: the moving mesh simulations also capture secondary flow fields that are transient, that MRF misses totally and even propose totally different results.

as He said, MRF is effectively only good for intial field for a moving mesh run.

Hi Opaque and all,

there are 2 fundamental issues here:

1. What are the consequences on the solution accuracy given the discrete approximations necessary in the method.

These are sometimes subtle.

Opaque points to the issues with using a moving mesh method ie GCL satisfaction and transient discretiztion.

There also are approximations introduced by solving in a non-intertial frame: these are the discrete approximation of the spatially varying "non-inertial body forces". Most codes use a single point volume integral approximation which can be shown to be second order accuracy for regular grids.

Which is better....it would make an interesting study? ;-)

2. How well does a given approach capture the relevant physics for a given problem ie. the significance of transient rotor-stator interaction.

Here there are many issues which include:

- what is of interest to the analysis? Overall performance which is just going to be a time average of the transient results anyhow, comparison between configurations, transient blade loading, can any of the transient effects on the mean flow be "modeled in a steady way"

- what are the scales of the unsteady structures ie what is the blade count, thickness of wakes, shocks, vortices

- how close is the coupling between components ie does the downstream component see a mixed out state

Of note, all of the above issues can be significant for any machine whether the flow path is axial, radial, or mixed.

As for literature on these issues, I have seen some good stuff come out of the DLR, NASA, etc. look through some of the recent articles in ASME Journal of Turbomachinery

Good Luck,

Bak_Flow

Hi Bak_Flow,

Thanks for the information.. I will look to throught the references you pointed out..

However, my major concern with the statement implied that moving mesh is better that MRF for all cases (I understood that).. For transient rotor-stator cases, I know the limitations of a moving frame approach.. In particular when the setup does include the whole rotor-stator configuration, or when the main flow is parallel to the axis of rotation.

I think that moving mesh has similar problems if you do not model the whole geometry..

thanks again, Opaque