Convergence problem on steady Frozen Rotor simulation

[JonasA]

Hi everyone!

I am using the turbomachinary module to study the flow in gas quenching cell with CS 5.0.7 on debian 9. The gas is nitogen at 20 bar, which is driven by a centrifugal pump. So far, I start to model it with Frozen Rotor model to get a first analysis. In the future, we will try the transient simulation latter.
I ran steady simulation with SIMPLEC algorithm, K-epsilon Linear Production turbulence model, and the rotation frequency set at 314rad/s (50 Hz). I started the simulation by slowly increasing my turbine rotary frequency (for each 1000-steps simulation, the frequency is increased by 25 rad/s each time).

I encountered some problem to have the solution converged.

I tried to change the pseudo-time step parameters. If the time step is allowed to go too low or too high, the simulation diverge. Even if its converge, the convergence is not significant and the pressure and velocity variables still vary a lot spanning a few thousands steps in the simulation. The cubic root of element's volume is between 2e-4 and 2e-2 m and the velocity in 10-150 m/s so in theory the time step should be between 2e-6 and 2e-3 for a Courant Number around 1. The lowest time step that I get without solution divergence is 1e-4 second for the reference time step with an minimal time step factor at 0.1.

Why is there these oscillations in a steady simulation? Could I improve the convergence by changing the parameters of the simulation or is it a mesh related issue and I would have to have a finer mesh near the casing/housing?

The xml, mesh, listing and monitoring are in the attachment. Using different time step and going some thousands steps further does not converge any further as shown in the figure attached.

Any suggestion to improve the convergence? Thank you.

Best Regards
Jonas

[Luciano Garelli]

Hello JonasA,

Did you check the result of the joining process at the rotor/stator interface, because I think that some face are not being joined. In the next fig is the face group(JOINRS) that you use in the selection criteria.

mesh2.png

(40.64 KiB) Not downloaded yet

Also, only the rotor volume mesh is in the listing.

Additionally, some mesh quality issue are reported in the listing file.

When I work with the turbomachinery module, I do a mesh for the rotor and another the stator. Then, I join the common faces.

If you want upload the geometries I will try to help you with the mesh.

Best regards,

Luciano

[JonasA]

Hi Luciano.

Thank you for your help, I have redo the mesh with 4 times more elements and stricter respect of the surface groups. Now the joining (JOINING.case) is correct, For the mesh quality, I have never had less than 1% distortion. I would appreciate any insight for doing a better mesh. I have made a hexa-mesh on the volute with 5mm element and refinement where my simulation had the more extreme values.
But the Velocity is still not converging (attached first coordinate of the Velocity on some monitoring point located in the different parts of my geometry from1000 to 1600 steps) after more than a pseudo second. It should be as the average velocity is 10m/s and my geometry has an almax of about 1m.

I really appreciate your offer to help me with the mesh, which software and which format do you use for meshing? I have trouble defining faces groups in Salome. Is there a way to embed faces groups in the step from the CAO and get them on Salome? The explode command would only give me the volumetric bodies.

Best regards,
Jonas

[Luciano Garelli]

Hello Jonas,

I have redo the mesh with 4 times more elements and stricter respect of the surface groups. Now the joining (JOINING.case) is correct
Jonas

This refinement should help in the resolutions, your previous mesh was too coarse.

I really appreciate your offer to help me with the mesh, which software and which format do you use for meshing? I have trouble defining faces groups in Salome.
Jonas

For meshing I use Salome or SnappyHexmesh for hexa-dominant meshes (then you have to convert to MED), but I never have problem with the definition of faces in Salome

If you can share the case and mesh I will try to help you.

Best Regards,

Luciano

[JonasA]

Thank for the proposition. How can I upload my mesh, it makes 40mb when 7ziped and the forum limit is 10mb.

By the way, is SnappyHexmesh suitable for turbine meshing?

Best regards,
Jonas

[Luciano Garelli]

Hello,

You can share a link with the files using Google drive or dropbox.

Regards,

Luciano

[JonasA]

Hi,

Thank you for your help proposition, here is my mesh and my case setting.
https://drive.google.com/open?id=1P72yZ ... MUa3Pls9KU

Best regards,
Jonas