[stud-many]

Hey. I'm simulating a small transonic diagonal compressor (1 stage). Rotating speed is up to 60.000 rpm.

When we were calculating the compressor with mixing-plane interfaces, the compressor efficiency was way higher then in our experiments. Therefore I switched to the frozen rotor interface (and we are not calculating periodical anymore). Now the results are closer to our experiments BUT we have severe convergence-problems in the simulations, especially if we are calculating OP's near surge.

Now the interessting thing: usually I'm using a pseudo-timestep between 0.1-10 / Omega. But at least some OP's are converging if I'm using somewhat about 50/omega. Other OP's near surge are still oscillating, regardless of the timestep. There is not even a change in the oscillations when I'm changing the timestep von 10/omega to 4000/omega! Only when using a really small timestep, the oscilating behavior is changing - but the simulation is still not converging (probably the oscillation-period is way bigger then before?)

I'm asking myself what I can learn from this. I think i was reading in the ANSYS Help, that if oscillating behavior is not affected by the timestep, then the flow itself is unstable and there is probably no steady state solution. Is that right?

And also: why are some simulations converging if I'm using a huge timestep? I'ts somehow nice that this is working but I'm not feeling smarter after all.

There might be some valuable information for those questions. I was trying to examine an unstable simulation and rendered a video of some values for the duration of one oscilation. In this compressor there are "cells of low impulse" downstream the rotor-blades. The fluid with the low impulse seems to interact with the stator and cause instabilities in the steady state simulation. Up to 3 passages of the stage are stalling for some iterations before this behaviour is repeating itself. I was able to get a stable result when i was using 50/omega instead of 10/omega as a timestep...