[Chris_321]

Hello and merry christmas,

i want to simulate a rotating axial fan and meassure the velocity and pressure at these two points.



The fan is rotating with 1000 1/min and my timestepsize is 0.0005 s (~120 timestep per fan revolution).
My boundary conditions are a total pressure inlet(0Pa) and an entrainment outlet at 0Pa( the reference pressure is 0Pa aswell).





I think i have a problem about convergence, because i expected something like a sin wave for the velocity over time with the amplitude right when the blade passes.

Is my mesh maybe too corse or my timestepsize too large? Which turbulence model would you suggest? Rans or les approach?

Best Regards,
Chris

[ghorrocks]

FAQ: http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

[Chris_321]

Hi Glenn,

at the moment im working through the faq steps, thanks!

Will you think i can meet the results i expect with the rans approach (with a small timestep) or do you think i need to do it with les?

[ghorrocks]

If the fan is operating near the design point (that is, attached flow) then RANS is usually fine. You usually need to consider LES as you get into off-design point operation where there are large separations.

Your mesh looks quite coarse as well. Make sure you do a good mesh sensitivity check as well.

[monkey1]

Just one more remark...as I saw n our attached images...you just simulated 0.5 s of time. I would expect, that your solution is still stuck somewhere between your intialisation and the "real" solution.
If you started with a motionless initialisation, I would expect that it takes more than 0.5 s before reaching the point where the whole domain reaches a "steady point" (knowing that a steady point for this kind of problem is not existing, but at least a periodically returning event).

And the image of your mesh is showing the whole domain or just a dteail of it? If it is showing your whole domain, then your boundaries (especially top and bottom) seem much too close to your fan for me. Therefore a negative influence of the boundary on the solution and vice versa could not be excluded...possibly leading to the fact that you don't see a periodic sin wave

[Chris_321]

Thank you for your anwsers!

I think the first thing i should do is remeshing and changing the domain size a bit.

Do you think it will work with just thin walls as blades, or do i need the blade profile?

[monkey1]

What do you mean with thin walls instead of the blade profile?
If you don't take the real shape of your fan, then you will get a different flow field....

[Chris_321]

Im more interessted in a generic chase than in the real flow for a given velocity. Do you think it will give me the assumed results( like a sin wave for velocity) ?

[ghorrocks]

Your mesh is too coarse to resolve much at all. Also you have not run it long enough, it is still going through the initial transients.

[Chris_321]

Originally Posted by ghorrocks

Your mesh is too coarse to resolve much at all. Also you have not run it long enough, it is still going through the initial transients.

I will improve that! I promise

But do you think i can go well with thin blades in this chase?

[JuPa]

Originally Posted by Chris_321

I will improve that! I promise

But do you think i can go well with thin blades in this chase?

Have a read of my thread regarding initial transients.

http://www.cfd-online.com/Forums/cfx...date-time.html

[ghorrocks]

Thin blades will be harder to simulate because they always have a separation at the leading edge. Separations are hard to simulate. So an airfoil shape operating near its design point will have attached flow over the whole length and will be much easier and more accurate.

[Chris_321]

Is there any generic axial fan geometry available online?

[ghorrocks]

No idea, try google.

[Chris_321]

i simplified the geometry to just one small part of one blade.




The steady state simulation works really well, but now i want to change to transient because i want to simulate one blade passing like you see below.

From:



To:





I know how to transform the domain location by hand, but how can i tell the solver to do that?



Got it! : Transient Stator Rotator :-)