[Noel_91]

Hello,

I am new to the forum and I'm only now approaching to CFX.
I performed a proliminary sizing of a centrifugal compressor that works with supercritical CO2. After that, to simulate its behavior, I used Ansys's CFX. In particular, I first used Bladegen, generated the geometry and then with turbogrid I created a mesh of about 500,000 elements and made sure that the quality of the mesh was good. After that with CFX pre I made the setup, creating three domains, namely S1(inlet), R1(main passage) and S2(outlet). I set the speed to -35450 rpm and set a physical timescale of 10 ^ -5 (otherwise with 10^-4 the simulation would have stopped). The problem is that, following the simulation, the contours acknowledge inconsistent results. On the positive side, I don't think I get droplets of liquid in the meridian plane and I don't think that any vortex are created near the elbow or strange trails at the trailing edge. On the other hand, I get high values ​​of the relative Mach in correspondence of the interpalar plane, up to reach supersonic values. What do you advise me to change in order to cope with this problem?
Thank you for your attention!

[ghorrocks]

Supercritical CO2 is a very complex material model and you should expect difficulties when you use complex material models. Have you checked that your simulation runs properly using a simple material model like an ideal gas or even incompressible?

Once you have checked it runs OK with a simple material model, my main piece of advice for your model is that you need to do sensitivity checks on all tunable parameters to ensure accuracy. The main issue is mesh size (you cannot assume your 500k elements is adequate without justification), but convergence tolerance, inlet and outlet boundary proximity are other issues to check. Only once you have checked you have set all these parameters correctly can you say your simulation is accurate, and looking at the details of the flow before you have done this is pointless. Who cares what the flow details of an inaccurate simulation are?

[jmenendez]

Originally Posted by Noel_91

Hello,

I am new to the forum and I'm only now approaching to CFX.
I performed a proliminary sizing of a centrifugal compressor that works with supercritical CO2. After that, to simulate its behavior, I used Ansys's CFX. In particular, I first used Bladegen, generated the geometry and then with turbogrid I created a mesh of about 500,000 elements and made sure that the quality of the mesh was good. After that with CFX pre I made the setup, creating three domains, namely S1(inlet), R1(main passage) and S2(outlet). I set the speed to -35450 rpm and set a physical timescale of 10 ^ -5 (otherwise with 10^-4 the simulation would have stopped). The problem is that, following the simulation, the contours acknowledge inconsistent results. On the positive side, I don't think I get droplets of liquid in the meridian plane and I don't think that any vortex are created near the elbow or strange trails at the trailing edge. On the other hand, I get high values ​​of the relative Mach in correspondence of the interpalar plane, up to reach supersonic values. What do you advise me to change in order to cope with this problem?
Thank you for your attention!

How do you evaluate the Mach number?, you must take into account the relative and non-relative frame.

I see you are caculating with an academic license that is limited to less than 0.5 million elements. What turbulence model are you using and what average yplus values are in your model? (you can find the second with ave(Yplus)@yoursurfaces. The models (compressor, one passage) that I use for SST have 2M of elements for yplus < 2