## Fluids

#### Convergence issues solver/mesh/physical Problem

• Shuttle
Subscriber

Hello there, I am currently simulating a thin gap b/w rotor and stator (Fig. attached) with less than 1 mm in dimension (mesh size is quite large). Initially, the rotor was rotating at around ~10 krpm and I was getting reverse flow and convergence issues, which I solved by extending my domain. Now for my further investigation, I increase the speed of rotor to ~100 krpm and the same issue occurs again, no convergence at all. Is extending the domain the only way to have convergence? Mesh quality looks ok: Skewness (max: 0.78; avg: 0.062), Min. Orthogonal: 0.22, Max. Aspect ratio: 93.87. I also tried to simulate with smaller time steps & relaxation factors, but nothing works.
Boundary conditions: Pressure inlet and outlet
Sol methods: Coupled, Green-gauss, all second ord
Solution controls: all default except energy: 0.6

• Rob
Ansys Employee

How much mesh have you got in the gaps (cell count) and are the back flow conditions sensible relative to the domain size, scale and conditions?  At 100k rpm what is the tip speed?

• Shuttle
Subscriber

Mesh size is approx. 35 million cells and tip speed based on TS=piDRPM/60 = ~665 m/s
I have not much idea how reverse flow behaves, but it looks like the higher the speed, the more reverse flow I am getting at boundaries inlet & outlet, which is also physical. Worth to mention: I am using an exp function for viscosity dependent on temperature

• Rob
Ansys Employee

As the rotation speed increases so does the axial effect on the generated vortex: look up toroidal vorticies.  Is the tip speed really Mach 2? Or do the operating conditions keep it subsonic?

Cell count is just that, how many cells between the rotor tip and the casing? I can model something that's well refined with 10k cells, or excessively coarse with 60M: it's all a question of scale and how I place my sizing & elements.

Reverse flow is common, but you need to be careful if it alters the result. Ideally, the domain is extended to prevent this happening but otherwise the domain is extended far enough that the reverse flow can't alter the flow in area we're interested in.

• Shuttle
Subscriber

ya, the opt conditions are maintaining the subsonic conditions. I just have the fluid gap as a mesh where down part is rotating and the above is stationary, so size is 35M. But it didn’t answer why after some iterations my residuals especially energy and continuity start diverging or jumping so high?

• Rob
Ansys Employee

It depends on what the evolving flow field is doing. If you have a rotor-stator is it sliding mesh or moving reference frame.

• Shuttle
Subscriber

moving reference frame

• Rob
Ansys Employee

Look at the flow field a bit before it fails. You're looking for odd velocity spikes but also the velocity field in the gaps, plot with node values off to see how well resolved the gradients are.