Hi, everyone! In CFX solver, I monitor the Boundary Flows and have a momentum flow mismatch between two sides of Stator-to-Rotor interface: e.g., U-Mom Flow = 0.29478 on its one side and â€"0.33322 on another side. (Unfortunately CFX does not quote the units.) Anybody knows if this difference is a sign of a faulty solution? And what are the units there? Thank you for you help.

Do your stator and rotor have same pitch?

Yes, they do; 1:1

Greeting to everyone!

Could anybody advise me please a practical criterion for solver run completion:

During a CFX solver run, apart from residuals, I also monitor other variables like forces or moments and boundary flows.

QUESTION:

= Should I wait until the boundary flows stabilise?

= Or once, say, forces are stable (and residuals are reasonably low), I may finish my run?

= Or just low residuals are a sufficient condition?

Thank you and Best regards

Steven

Low residuals are very sufficient. The residuals won't converge if you have big instabilities. If you want to produce academical results just lower the RMS to 1E-05. Then your solution is very fine. But depending on your boundary condition and meshing maybe you are unable to reach 1E-05.

Dear Quest, thank you for your advice. In some cases, the amplitude of my boundary flow fluctuations rises, while the residuals drop (say, from 1E-3 to 1E-4).

Right now, I am monitoring the run, where:

U-, V-, W-Mom residuals by domain are all under 5E-5 (and are going further down),

P-Mass Rotor < 7E-7 and P-Mass Stator ~1E-7,

-- but Interface Flows, Pressure- and Viscous Moments have not relaxed yet, after 536 iterations.

And again, the flows @ stator-rotor interface do not quite match:

U-Mom = +0.5895 & -0.5869 (of which units?), but

V-Mom = +0.105 & -0.0428, that is a 150% imbalance!

Frankly, I do not understand, why CFX counts the DOMAIN Imbalance, but disregards that for each virtual surface?

As far as there is no sources inside of a (zero-thickness) interface, all flows on it should be balanced, should not they?

That all makes me concerned regarding the practical criterion (criteria) for completion of CFX runs.

Very best regards

Steven

Hi Steven,

If there is a frame change across the interface, your momentum will be based on the relative frame velocities, not the absolute, so you should expect a difference in momentum on either side. If you compare the stationary frame momentum across the interface in Post, they should match to within round-off.

The solver reports domain imbalances because these are what is important. The interface formulation is strongly conservative, meaning that it is conservation that is enforced across the interface (not simply the interpolation of values), therefore there is no need to report interface imbalance as it is zero by design.

I'm curious, since you have a 1:1 connection, are you solving for a full 360? If so you can also set the "Transformation Type" to none, which will save some overhead as the solver will not search for the periodic tranformation.

Regards, Robin

Dear Robin, thank you for your explanations.

What disturbs me is after ~700 iteratrions, I have U-moms = +0.623 and - 0.624, but V-moms = +0.136 and -0.067! Is not something going wrong?

Also thank you for your advice: I really run a 360 deg model and have set the "Transformation Type" to none. Thank you.

Best regards Steven