[pace]

Hi

I am trying to determine which program would be best to use for simulating a centrifugal compressor-diffuser assembly.
(if a discussion on this has already been covered – I failed to find it – please post a link to it)
I want to determine volume flow and exhaust temperature at given pressure ratio (and given rotational frequency). Blade tip speed is less than Mach 2.

So far I am familiar with SolidWorks flow simulation and Ansys CFX, but am unsure of how to asses accuracy of the results. Is there a good practice for this – without having to have any physical tests for comparison? Indeed my main question is: how do I know whether to trust the simulation results or not?
Of course cross comparison of simulations results with hand calculations (or semi-numerical approximations) is a good reference point, but hand calculations are inaccurate when the compressors geometry becomes overly complex.
If possible I would at least like to obtain volume flow with a max of 20% error.

Any advice you can give me is greatly appreciated.

Thank you for your time.

[fluid23]

Without experimental data there is no waw to know how accurate your model is. There are a number of things you can do to make sure the model is appropriate. Such as keeping your mesh quality metrics in-line (cell quality, skewness, volume change, etc..). You should also make sure that your y+ values are appropriate for the turbulence model and wall functions you are using. If you keep these thing where they need to be and your physical repsresntation is appropriate (donmain is large enough, boundary conditions are appropriate) then you should be fine.

If you are worried about accuracy, Solidworks isn't the tool to use. It's nice or easy analysis or quick conceptual trade studies but lacks some very fundamental stuff. I am not even sure if you can do rotating machiner with Solidworks. I have installed and hardley ever use it because it can be so finiky and slow.

That being said, an error of 20% in flow rate is fairly generous. You shouldn't have much trouble getting there. I use Star-CCM+ regularly and typically see errors less than 10% for flow rates and pressures that interest me, usually much less than 10%.

[fluid23]

This of course assumes that your enginering quantities and residuals all converge.

[pace]

Thank you for your reply.

I will compare current simulation results and my semi-numerical results with Star-CCM+ (as soon as I learn it sufficiently).

SolidWorks is simple and intuitive – to a point – and it gives very good results in subsonic cases. In those results I do not doubt as I can verify them, but for turbo-machinery (it is doable by using a rotating region) I have – so far – no comparison studies. Thus I don’t know whether to trust it.
Well, as I see it, solidworks just requires a bit more attention to the mesh around problematic areas, but it can resolve many issues by itself. It might be a bit slow at times – it depends on the conditions in the simulation – but it’s very stabile (a few years ago I had a lot of stability issues with Ansys, though that might have been due to my lack of knowledge). Also I find it ideal for modeling (but not for 2-D drawing).
But, as I said, regarding simulations accuracy and adequacy I can only speak for subsonic simulations.

Having an error of less than 10% would be most beneficial. I was going with 20% as a safety factor, because I only have a lower limit and no upper limit, thus I figured if I surpass the lower limit for 20% I’m on the safe side.

How accurate is, by your experience, Star-CCM+ with torque estimation?

I am much obliged for your information.

[fluid23]

Solidworks, compared to Ansys, star-ccm and others, is very rudamentary. Their mesh options aren't the best and they don't have any options for turbulence models. It's a good preliminary design code, but I wouldn't try to quantify anything with it. The reason Ansys was unstable is probably due to mesh problems, 9 times out 10 anyway. Solidworks isn't smart enough to have those problems, not sure that makes it better.

Star-ccm is generally recongnized as one of the most accurate industry level codes available. Although, for the most part your body forces and moments are going to be really driven by model setup, approach, mesh, physics more than some magic quality of the software.