Air Flow Through Tube Symmetry Issue?


I’m a total noob at Simulations. I want to investigate the air speed and turbulence flowing through a tube with different flared shapes.

This is the complete part - two volumes with a flared tube between them.

I split it in quarters and put symmetry on two sides.

Looking at TKE I get much more on the one plane than the other. That doesn’t seem correct to me for a circular part?

What is wrong here? I did try increasing the mesh detail, but no change.

Project Link:

Thank you!

Hi @amyeunji,

Thanks for posting your question! Very interesting project. Maybe a stream tracer help to get the understanding of that.

For me the simulation setup looks ok, so it seems to be a matter of interpretation. If the flow at the inlet is predominantly laminar (as the image above shows), there may not be significant generation of Turbulent Kinetic Energy.

Please let me also tag @DaleKramer here who might have some valuable insights to point out.


1 Like

In my experience this is expected behavior in regions where you would expect the flow to be separated and turbulent.

If you want to study these eddies and vortices which I see may be forming unsymmetrically in your plot, I would suggest learning how to use the LIC plot feature of Paraview.

You can see how well LIC plots present the actual flow lines in a planar view here:

Have fun ! :grinning:

Thanks for the replies!

It may well be my expectation is not correct or I am not doing the right type of sim, allow me to explain what I wanted to do in more detail.

The reason I wanted to do this is because I was talking to someone about making a Helmholtz resonator. I wished to illustrate that adding a flare and rounded end to the tube can reduce unwanted noises caused by turbulence. However I seem to be disproving myself as a straight tube sim shows lower max TKE!

While I am aware a resonator is far more complex than air flowing from one volume to the other through a tube, I assumed perhaps wrongly, this would be enough to illustrate how a flare can reduce turbulence and thus noise.

I think we all know from experience that sharp edges produce more unpleasant noise than rounded ones when air flows over them, but perhaps max TKE is not a representative metric alone? I suppose area of turbulent flow could affect total SPL and in a real situation the pitch of the undesirable noises will also contribute. Any thoughts on this?

Thanks for the suggestion of using ParaView,. Truthfully as a hobbyist I don’t think I have enough invested in learning what looks like a pretty complex program, at least for now.

The logic in my brain says that there would exist a pressure differential , lets call it X mbar (you now have a differential of 5 mbar) where the flared/rounded tube would maintain laminar flow throughout the whole volume but with squared tube at that X mbar you would likely see separated flow at the square outlet of the small squared tube, and also some just after the inlet of the small squared tube.

Above X mbar, both would have separation in the volume as you have now.

I have a quick look at you sim setup and it appears on the surface to have the possibility of some reasonable results but I would likely extend the inlet and outlet faces another 5 big diameters away from the small flared tube.

I am a hobbyist too :wink:

Thank you! I’ll play a bit more later.

You might be completely right. I remember reading an experimental paper that said at the highest pressure levels a straight tube actually beat the flared one in the metric of ‘undesirable noises’, but at lower pressure levels the flared one was better. Perhaps 5 mbar is too high.

To do this with ‘better practices’, you would need add some boundary layers and start trying to that verifiably achieve acceptable surface yPlus values (30-300 for turbulent flow), at the differential pressure you choose, and also use ‘wall function’ (log-log region) for the no-slip walls.

I reduced the pressure to 0.05 mbar and this looks much more like what I expected!

I ran out of ‘credits’ so can’t see the scaling of the values but NVM…

So this was actually very interesting because it supports the research paper I read that at higher pressure levels / air speed a straight tube can perform better than a flared one. Fascinating!

Dale, as a hobbyist how do you handle the credits situation? Make multiple accounts? I asked about prices but they are not affordable for me.

I learned a lot about CFD and FEA and then was offered the status of a PowerUser to help others in the forum etc. I am not sure if they are looking for any more at this time. It looks like since they have limited the free usage to 10 simulation runs that there are lots less hobbyists in the forum nowadays. But still lots of students with free educational licences without the 10 limit.


Well thank you for your help!

1 Like

To do this with ‘better practices’, you would need add some boundary layers and start trying to get the surface yPlus values acceptable (30-300 for turbulent flow), at the differential pressure you choose, and also use ‘wall function’ (log-log region) for the no-slip walls.

Hi Dale, I wonder if you are able to link some videos or tutorials that explain the details of what you mention here? Much of it goes over my head such as wall function and no-slip! Boundary layers are automatically created in the mesh, how does one know what a ‘suitable’ size or settings are for an application?

I am sure you can find some info here : SimScale Documentation | Online Simulation Software | SimScale and the tutorials there.

I gave you keywords that can be searched for on the web and this forum. I do not have a link list handy myself, sorry…

It might help if I added that you would have to use ‘Hex-dominant Parametric’ meshes :wink:

Edit: I wasn’t satisfied with my short answer, this is not a tutorial or anything but you can see how I precisely verified my surface yPlus range in this project of mine: '01234 layer HEX mesh Test' simulation project by DaleKramer

1 Like