Random Technical Questions

[CH3NO2]

Hello all,

I've been wondering about conical diffusers and the effect of swirling gas flow on optimum diffusion angle.
If you have a swirling gas flow entering the inlet of a diffuser, can the diffusion angle be increased while maintaining a selected % pressure recovery?

Take for example a conical diffuser at the turbine exducer.  The normal optimum pressure recovery diffusion half angle for laminar flow is ~7 degrees.   But what does swirl do to the optimum diffusion angle?..

My guess is turbine outlet swirl momentum would allow for a steeper diffuser angle... but I cant find literature on the subject.

Thanks,
Tony

[racket]

Hi Tony

Yep , I think I'd agree with you on that one.

But , I wouldn't be trying to "guess" what comes out of a radial inflow turb exducer that has a scroll feeding it, you can have a "dogs breakfast" of angles and speeds all at the same time .
proceedings.asmedigitalcollection.asme.org/proceeding.aspx?articleid=1894039

Stick to the standard angle and pray that its working ;-)

Cheers
John

[CH3NO2]

Hi John,

I had a random question on compressor scrolls.

I noticed that air is fed into the scroll, from the diffuser, at the outer circumference of the scroll flow passage. This is in contrast to the turbine scroll where the incoming gasses are fed into the turbine from the center of the turbine scroll flow passage.

It is easy to see why the turbine is fed from the center of the turbine scroll. It makes for a lower average Cd and symmetrical flow.
However, with the compressor scroll, the air is fed to the outer edge of the circular flow passage and this sets up a strong swirl or tangential component to the flow at the outlet of the compressor scroll.

Its like a powerful tornado is spinning its way out of the compressor scroll.

Why are all compressor scrolls made this way? Compactness of design?
Unless maybe a wide angle diffuser was attached directly to the compressor scroll outlet, I can't see how a spin contributes anything.

Thanks,
Tony

[racket]

Hi Tony

Yep , gotta be compactness of design , it certainly can't do anything for efficiency with all that swirl increasing fricional losses .

I remember seeing those helical witness marks inside my TV84 scroll and thinking that needs to be "straightened out" especially if theres any sort of length involved between scroll and combustor , it probably influenced me to make the ducting as short as possible .

Cheers
John

[CH3NO2]

Hi John,

When you noticed the helical witness marks what was the ~approximate~ relative angle you noticed?

It seems strange that they ALL do it the same way. Dumped tangentially from a sharp edge. As if the spin is done for a purpose.
Certainly not like any other subsonic diffuser I've ever seen.

Thanks,
Tony

[racket]

Hi Tony

LOL..............it was very long time ago , but I think it was a fairly "stretched out " helix, as one expect considering the relationship between radial and whirl velocities in our diffusers .

Even the Allison 250 heli engine has a similar setup www.google.com/search?q=Allison+250+diffuser&source=lnms&tbm=isch&sa=X&ved=0ahUKEwiB_ZiK2q3fAhUJLI8KHYPODS4Q_AUIDigB&biw=1680&bih=951#imgrc=NeoyfhIM6Br8gM:

It can only be for compactness, but can't be 100% certain though

Cheers
John

[finiteparts]

Tony,

If you think about the flow that the compressor volute has to accept, there is a large flow angle tolerance that it incurs. At low flow rates, the discharge from the impeller is at a large angle from the radial, while at higher flow rates the impeller discharge has moved towards a more radial flow. This is quite a challenge to accept a flow at such a large range of inlet flow angles and then basically turn it 90 degrees so it can flow into the diffuser section, just after the tongue region of the scroll housing, with minimal losses.

The fact that the cross-section is nearly circular and being fed from one side is a means of minimizing dump losses in as compact an arrangement as possible. A circular cross-section minimizes the wetted wall area, thus the amount of wall scrubbing that would occur. If you move the diffuser discharge into the center of the volute passage, then you will generate two passage vortices as opposed to the single passage vortex that is produced from the side entry arrangement.

The compactness is sort of set by the vaneless diffuser outer diameter requirements. Once the diffuser exit diameter is large enough to reduce the discharge velocity to some level, the scroll can then accept the flow with some acceptable level of loss. The flow enters the scroll and then the tongue sets the flow range of the housing. Once the flow passes through the tongue, it enters the final stage of diffusing in the housing. The area ratio after the tongue can be significant depending on the allowable length of the discharge and how it is to be installed.

There are some good papers out there on the design of centrifugal compressor volutes if you want further info.

- Chris

[CH3NO2]

Dec 22, 2018 15:21:53 GMT -5 finiteparts said:

Tony,

.......

1)  The compactness is sort of set by the vaneless diffuser outer diameter requirements. Once the diffuser exit diameter is large enough to reduce the discharge velocity to some level, the scroll can then accept the flow with some acceptable level of loss.

2)  The flow enters the scroll and then the tongue sets the flow range of the housing.  Once the flow passes through the tongue, it enters the final stage of diffusing in the housing. The area ratio after the tongue can be significant depending on the allowable length of the discharge and how it is to be installed.

There are some good papers out there on the design of centrifugal compressor volutes if you want further info.

- Chris

Hi Chris,

1) For reducing the velocity sufficiently, the radial diffuser diameter is enlarged...  Other than the upstream pinch, I've notiiced radial diffusers seem to be predominantly parallel walled to the point where they dump into the scroll.  Why isn't there a divergent angle employed into the radial diffuser?

At least for a diffuser used at the inlet cowel of a turbojet or ramjet engine, diffusion angles up to about 12' degrees can be used without significant flow separation.... So why cant a 5' degree - 8' degree diffusion angle be employed on the radial diffuser of a turbochargers compressor scroll?    It would seem to allow for a smaller diameter, increased pressure recovery and increased compactness...

2)  I was going to ask a few questions on this but seeing that you mentioned there are good papers on volutes I did a quick search and was able to answer many questions here:  www.conceptsnrec.com/blog/important-considerations-when-designing-a-volute
While it does discuss the length and area ratio of the diffuser section after the tongue, it doesn't go into the effect of axial spin on the diffusion angle.   And axial spin has me curious.

Can axial spin significantly affect diffuser stall angle?   Anything beyond the usual 12' degrees?

Thanks,
Tony

[enginewhisperer]

due to the increase in area as the diameter increases, a divergent radial diffuser would end up with much higher than 12° effective angle.

From memory, based on conversations on this forum, some radial diffusers actually converge slightly to maintain the correct expansion.

[CH3NO2]

Yes, this is true, and I've read that it has been specifically stated to NOT use a divergent section in a radial diffuser... but I dont understand why. What is the mechanism for flow separation if divergence is less than 8 degrees? Even if it is radial. That's weird.

Leaving out boundary layer for the sake of discussion, how does subsonic flow separation occur for anything less than 8 degrees?
The conditions in a radial diffuser seem so perfect for rapid and efficient diffusion. Even better if a divergent angle could be included.  The divergence wouldn't even have to be linear. The upper and lower plates could be gently curved like an air foil to help keep flow attached.


I know its not supposed to work but I dont understand why.
Then again, I've never studied radial diffusers either.

[finiteparts]

Dec 23, 2018 16:55:34 GMT -5 CH3NO2 said:

Yes, this is true, and I've read that it has been specifically stated to NOT use a divergent section in a radial diffuser... but I dont understand why. What is the mechanism for flow separation if divergence is less than 8 degrees? Even if it is radial. That's weird.

Leaving out boundary layer for the sake of discussion, how does subsonic flow separation occur for anything less than 8 degrees?
The conditions in a radial diffuser seem so perfect for rapid and efficient diffusion. Even better if a divergent angle could be included.  The divergence wouldn't even have to be linear. The upper and lower plates could be gently curved like an air foil to help keep flow attached.


I know its not supposed to work but I dont understand why.
Then again, I've never studied radial diffusers either.

Hi Tony,

It is all an area ratio play with diffusers...and you can never leave out the boundary layer when discussing diffusers, because flow separation is the fundamental loss mechanism.

The diffuser divergence angle is only valuable in reference to the specific type of diffuser that is being referenced, because it defines the area ratio of the diffuser. So, is the 8 degree divergence angle that you are referencing valid for a channel diffuser, conical diffuser, annular diffuser or some other type? What I am trying to illustrate is that it is hard to generalize diffuser geometry across different types of diffusers. What you are really looking to define is the overall area ratios of the diffusers and their impact on pressure recovery.  If you had a parallel wall diffuser, the area increases linearly with diameter...ie. A2 = (pi*d)*height, where d is the only thing changing.  But, if you now linearly expand the walls also, you are increasing the area by a non-linear function, where d and height are both changing.  If you change the area too quickly, you can get flow separation. 

The problem with almost all diffusers is that the boundary layer does grow and at some point, it breaks away from the surface and you get separated flow. The cross sectional area is increasing as you go out in radius and even with the boundary layer growth causing an increased aerodynamic blockage, the radial component of the velocity is dropping. The tangential component of velocity is dropping slower than the radial, thus the absolute angle is turning more towards the tangential direction. At some flow conditions, the radial velocity goes to zero and the flow gets pulled back into the impeller. This backflow is one form of instability that can occur in a vaneless diffuser. There are other forms of flow breakdown in the vaneless space that can drive the diffuser into instability.

Often times vaneless diffusers use pinch at the entrance to increase the radial component of velocity and thus turn the inlet flow vector more towards the radial direction to stave off backflow or rotating separated flow cells...thus expanding the walls of the diffuser would be quite abnormal.  You may be able to get it to work somehow, but the general trend suggests against it.

Also, I have found some good information on the impact of swirling flow in diffusers that I will try to find time to post tomorrow.  It turns out that the on conical and annular diffusers, some inlet swirl (Inlet swirl up to 10 to 15 degrees ) does positively impact the pressure recovery.  The radial pressure gradient appears to help stabilize the boundary layer on the outer wall delaying separation and flow breakdown.

I hope that helps,

Chris

[monty]

For viscous flow... at a certain pressure gradient separation will occur.. Pressure gradient....it's the thing. Reynolds number....Viscosity vs Momentum. Does a fluid particle wish to stick to the wall, or its surrounding flow, or is the momentum just too great and it tears away?

The area in a radial diffuser is changing with dr as a function of pi...so any additional area increase via divergent angle can push things over the edge.

such is the perversity of nature...


Monty

[CH3NO2]

Dec 24, 2018 0:02:04 GMT -5 finiteparts said:

Dec 23, 2018 16:55:34 GMT -5 CH3NO2 said:

Hi Tony,

It is all an area ratio play with diffusers...and you can never leave out the boundary layer when discussing diffusers, because flow separation is the fundamental loss mechanism.

The diffuser divergence angle is only valuable in reference to the specific type of diffuser that is being referenced, because it defines the area ratio of the diffuser. So, is the 8 degree divergence angle that you are referencing valid for a channel diffuser, conical diffuser, annular diffuser or some other type? ...

Chris

Hi Chris,

Conical.  I suppose my misconception stemmed from thinking in the context of the flow being able to diverge from normal like it does for air foils or conical subsonic diffusers.  One thing I've noticed about radial diffusers is they follow different rules from conical diffusers.

Tony




MERRY CHRISTMAS EVERYONE!




..

[finiteparts]

Tony,

I haven't forgotten about this post...I just haven't had much free time to collect all the diffuser and swirl information that I wanted to cite. Maybe this weekend.

- Chris

[CH3NO2]

No hurry at all Chris. You pretty much answered the question already. Spin can affect diffusion angle. Thank you!

When you mentioned "There are some good papers out there" it prompted me to do some google digging on the subject. Because of the digging I also learned that axial spin can also help to equalize flow distribution immediately downstream of a tight bend radius.

If for some reason a silicone elbow is placed near the compressor inducer it can cause asymmetric flow disturbances of the air entering the compressor blades. Leading to instabilities within the compressor and diffuser. So what I found was at least part of the asymmetry and instability problem can be mitigated if an axial spin can be imparted on the flow just upstream of the bend. The spin helps the flow stay attached while traversing the bend.
This was a nice little tidbit of info to learn.


...Now its about time I need to start research on design of radial nozzle guide vanes. Starting with no prior experience on NGV it should be a heavy subject to take on ... but it looks so tasty.