[Maerluz]

Hi, I'm new

First of all thank you for all your effort and dedication to this forum; it is a great resource!

I'm facing a steady simulation involving a full centrifugal pump (domains: inlet, impeller and a volute with an extended ending) in CFX. The mesh is structured for all the domains, y+ is about 5 for the impeller and about 10-20 for the volute and inlet; turbulence model is SST.
I started a simulation with frozen rotor interface between the rotating domain (impeller) and the stationary ones and it converged (RMS attached just to have an idea).

Troubles come when stage interface is used instead of frozen rotor interface .
I tried to follow some tips including initializing stage interface simulation starting from a previous simulation and changing initial timescale factor but without any results since I got fatal overflow in linear solver error (RMS and p-mass imbalance for impeller and volute domain attached). I don't know what could be meanigful to try now in order to obtain a converged solution with the stage interface.
I really appreciate any suggestions .

Thank you for your time

Mario

[Opaque]

What option are you using for the pseudo-timestep? Auto Timescale or Physical Timescale?

For Auto Timescale, you should try reducing the computed timescale, say Timescale Factor = 0.1 for a start.

[Maerluz]

Thank you for your reply.

I'm using Auto-timescale with Timescale Factor = 0.1.
I also tried 0.01, 1 and 10.

[Opaque]

If you already tried those values, I think there is something wrong/off in the setup from Frozen Rotor -> Mixing Plane

The solution for Frozen Rotor is already well behaved, so I would not have expected such a different behavior.

Perhaps more information is needed to help you.

If you compare the output files between the two setup, say FrozenR_001.out vs MixingP_001.out, what would the setup/CCL differences be?

[Maerluz]

don't seem to see any diferences

(Timescale factor is equal to one in MixingP_001 but I did the simulation with 0.1 too and it's even worst).

[Opaque]

It seems you are restarting the mixing plane solution from the CFX_002.res file of the Frozen Rotor solution, not CFX_004.res (which corresponds to the output file you provided)

If you compare the Global scales at the end of the Frozen Rotor solution with the Global scales at the start of the Mixing Plane simulation, you will see the rotor information is quite different in Velocity/Advection Time and Reynolds Number. That is, something seems off.

[Maerluz]

Sorry I didn't point it out.
The files attached previously belong to 2 different set of simulations. MixingP001 does not restart from frozen rotor file but it restarts from a simulation (with exactly the same setup of that attached) which is initialized with automatic values

[Maerluz]

If it can help, I also attach the first part of the mixing plane (stage) simulation: MixingP_000.

I'm wondering if this result could be typical of a transient behaviour. But I have no experience about such a diverging mixing plane phenomenon

[Opaque]

If you look at the global scales on the last output file (_000.txt), you can see the initial values are all over the place. It will not converge easily

Could you post the output file, for the mixing plane calculation using the frozen rotor well-converged solution as the initial values file?

[Maerluz]

Yes; the file attached is the output file you asked me

[Opaque]

Good. If restarting from the previous Frozen rotor, the global scales are the same as expected. Next is to compare the residual changes when using the mixing plane,

However, your frozen rotor case was run using a Timescale Factor = 5 while the last mixing plane case you posted is for Timescale Factor = 0.1 (perhaps what I suggested previously).

Could you run the Mixing Plane setup using the same settings as you did for Frozen Rotor?, Timescale Factor = 5

When several parameters are changed, it is difficult to isolate where the problem is coming from. If you can also post the run for Timescale Factor = 1 (the default timescale), it would help differentiate the residual behavior.

[Maerluz]

ok, files are attached.

To sum up:
MixingP_002 is .out file for the simulation with timescale factor=0.1;
MixingP_003 is .out file for the simulation with timescale factor=5;
MixingP_004 is .out file for the simulation with timescale factor=1;

All those simulations are restarting from the previous weel-converged frozen rotor.

[Maerluz]

I tried also to shift the stage interface position. Anyway it is very close to the volute entrance (2 mm) as you can see in the picture attached (the first circumference 2 mm downstream the trailing edge). The meshes at the stage interface are exactly the same since I generated the impeller in turbogrid and the stationary domain downstream is the 'outlet' automatically created in turbogrid too.

The same situation arises for all different attempts I've done: after few iterations ,from the well-converged frozen rotor, the simulation diverges. Typical pressure and velocity fields captured after 3 iterations from the beginning of mixing plane run are attached.

The maxima residuals are located at the stage interface.