[Wolfram]

Dear all,

I measured the static pressure rise of a fan in test bench. Afterwards I am trying to obtain the same results in my simulation.

Test bench results:
60m³/h = 26Pa static pressure rise
80m³/h = 0Pa static pressure (free flow condition)

Now in the simulation I am facing the following situation: The finer the mesh, the larger is the deviation between the measured and simulated results. With a finer mesh, the performance curve becomes flat - so the deviation is not consistent for the duty points mentioned above.

CFD-Results:
- coarse Mesh (8 Mio Elements)
60m³/h = 23Pa (-3Pa compared to test bench)
80m³/h = 2Pa (+2Pa compared to test bench)
- fine mesh (20 Mio Elements)
60m³/h = 18Pa (-8Pa compared to test bench)
80m³/h = 5Pa (+5Pa compared to test bench)

I reviewed the mesh a thousand times, the quality is pretty equal since all face sizing remained the same. The only difference is the element size of the "free volume" not touching any walls like housing or blades. The inflation height is also equal.



Furthermore I observed a complete different flow pattern in the rotating region between the blades. The static pressure distribution (blade loading) is also different. Please have a look to the attached pictures where I tried to visualize the different flow pattern.

cfdo.jpg
cfdo2.jpg
cfdo3.jpg

Is there any explanation of the influence of a finer mesh for the different flow pattern shown in the pictues and thus the deviation of the static pressure rise?



Ok, next Information for you. I ran the same blade geometry meshed with a perfect mesh created (block structured hex mesh) by turbogrid and a perfect matching pressure rise to the test bench results. In the following picture the static pressure in the blades pressure side is shown:

cfdo4.jpg

I appreciate any hint! Please let me know if you need any further information.

Many thanks for your help in advance
Wolfgang

[Wolfram]

Dear all,

in the following I used exactly the same setup for my fan
- Total pressure @Inlet
- Mass flow @Outlet
- same Mesh
- same rpm, and so on ...

The only difference is the timescale factor in CFX and as a result the deviation of the static pressure is huge! I measured the geometry with a test bench and I am expecting a pressure rise of approx 25Pa.

In the following screenshot the difference is shown:
aaaa.jpg


What is the reason for this behaviour and how to avoid it?

Many thanks in advance

[Wolfram]

Could anyone explain the huge difference of the static pressure rise and especially the different flow pattern as shown above depending on the timescale factor in CFX?

[naffrancois]

Hello,

Never used CFX nor expert on the kind of simulations you are doing. Are you doing unsteady analysis with a pseudo-time integrator ? If so maybe your time steps are not well converged on the finer grid

[Wolfram]

Hi,

it is a steady run and the results differ heavily depending on the time scale factor.

[naffrancois]

I am not so sure about what exactly does the time scale factor. A rapid search seems to me that it somehow allows to control convergence speed.

In theory converged is converged independantly from the path. So are you sure your simulations are converged to a reasonable threshold ?

I am under the impression that your run using timescale factor 70 is far from converged, do your residual plots indicate convergence ?

[arjun]

Originally Posted by Wolfram

Could anyone explain the huge difference of the static pressure rise and especially the different flow pattern as shown above depending on the timescale factor in CFX?

Here I think the time scale factor is time scale of turbulence. This is used in calculating turbulent viscosity

turb viscosity = rho . kinetic energy . Time Scale

Lower values would result in lower turbulent viscosity (and usually lower pressure).


With finer mesh, it seems your turbulence production has gone down and now results have changed a lot.

[Wolfram]

Dear all,

many thanks for your replies. I added the residual plots (rms) below the pressure plots from the post-processor as shown in the previous posts above.

The run with a time scale factor shows much lower residuals.

Please have in mind that I measured the geometry on a test bench and I am expecting a static pressure rise of approx 26Pa and the run with a time scale factor of 70 is much closer to this expected value despite the higher residuals.

Furthermore, the pressure plot seems to be more realistic? What do you think?

Do you think the "bluff body" and maybe occuring vortex shedding in combination with different time scale factors of a steady run could affect the results?
aaaa.jpg

[JBeilke]

I'm doing turbomachinery simulation for about 25 year now and only used steady calculations as starting point for a transient simulation with rotating mesh.



http://www.beilke-cfd.de/beispiele.html

[Wolfram]

Originally Posted by JBeilke

I'm doing turbomachinery simulation for about 25 year now and only used steady calculations as starting point for a transient simulation with rotating mesh.



http://www.beilke-cfd.de/beispiele.html

That is a long time .

Unfortunately, I did not find any helpful hints related to my questions above neither in your post nor on the shared website.

[JBeilke]

The hint was, not to trust steady turbomachinery calculations at all :-)

[Wolfram]

Originally Posted by JBeilke

The hint was, not to trust steady turbomachinery calculations at all :-)

Many thanks for this helpful hint related to my questions from post #8.

Let us switch from "trust" to "unterstand" the calculation (I am really trying!).

Is there any explanation for the complete different flow pattern as shown in the screenshots associated with:
- the time scale factor
- bluff body
- high and low residuals depending on the time scale factor ?