j84 Gas turbine

[ripp]

Hi Beaker,

the ball bearing preload front installation




www.john-tom.com/html/Jet.html

members.tele2.nl/geraldensuzanne/turbines.htm

Cheers
Ralph

translate.google.at

[beaker]

Thanks guys.

Got another one.
This sound about right?
Some of my diffuser numbers are.



Compressor exducer 87.4mm dia
inducer 63mm dia
outlet height 6.45mm
@ 100000 rpm 457.39 m/s peripheral speed.
Mass flow 0.45 kg/sec
pressure ratio 3.8
ratio after compressor 1.95 temperature 370k
radial speed 134ms
speed plan 394ms
diffuser angle approx 19deg +1 for blade cross section.
diffuser start Diameter 98mm
Diffuser to elbow into combustion chamber 127mm
total OD 140mm

probably forgot something.
Just to be sure should I go for it!

Now for a screen shot._

Attachments:

[turbotom]

Beaker -

Your figures look quite resonable except for the pressure ratio. I would be very surprised if you could reach a PR that high at the specified tip speed. My spreadsheet program tells me this would require a reaction of 0.36 which is not very likely to result in a stable compressor... I would expect rather a PR somewhere around 3 (impeller PR 1.72), diffuser starting angle 21° and I would have a much smaller vaneless space. Experience tells that these turbocharger compressors work at best efficiency at almost no vaneless space. So I would have the diffuser vanes start at a diameter of maximum 90 mm which gives you much longer channels -- better compressor efficiency. A thrust of 220N at very moderate temperatures should be possible.

Good luck and have fun,

Thomas

[beaker]

Thanks heaps Thomas.

This is my first design so lots to learn.
Have no real compressor map so just used the AMT olympus data for PR ( I think the J84 parts are just a Chinese copy.)
Have always questioned the vaneless space? what I see with with other designs is a very small space, but books like kamps model jet enines book recommended
1.12-1.2 so just used that.
Sounds better now.

Should I keep the wedges sharp at the entrance or should I put a small radius like 0.2-0.5?

Beaker

[finiteparts]

Beaker,

I would tend to agree with Kamp's book on including the vaneless space, but the numbers cited in the design literature are smaller. Cumptsy's "Compressor Aerodynamics" suggests 1.1 and Japiske's "Centrifugal Compressor Design and Performance" suggests between 1.02 to 1.12, with both of these being defined as the ratio of the diffuser vane inlet diameter to the compressor impeller outlet diameter. If the exit flow from the impeller was above Mach 1, then Cumptsy states that a vaneless spaces of 1.25 may be required, so the 1.2 from Kamp's book would be an upper bound on a highly loaded compressor for our uses.

The reason for the vaneless space is quite well documented in the published literature as a means to allow the distorted flow from the compressor to "mixout" before entering the diffuser. The reason the compressor outlet flow is highly distorted is due to flow within the blade passages themselves. The pressure side of the vane is the one that appears to be pushing the flow , while the suction side is the side that would be pulling the flow as the impeller rotates. The suction side produces a lower flowrate at the outlet than compared to the pressure side. This leads to what is termed a "two-zone" flow or a "wake-jet" flow at the outlet of each impeller passage. Without the vaneless region, the diffuser is fed with a more distorted (unsteady) flow and since the flow velocity into the diffuser fluctuates (or more accurately is seen as pulsating), the flow incidence on the diffuser vanes fluctuates.

This pulsing flow is the main reason that it is highly discouraged to have the same number of diffuser passages as impeller passages. If the pulsations excite all the diffuser passages at the same time, then if the flow were to separate in a diffuser passage, you could get all of them to separate and thus a complete diffuser stall as opposed to the potential for a single passage to stall if the vane/diffuser passages are not the same number. Additionally, the presence of the diffuser vanes imposes an upstream pressure pulse on the impeller vanes as they pass by and can cause a vibration or if they are the same number, they can cause the blades to enter resonance and thus fatigue failure...the vibrating vanes are like a paperclip bent back and forth till they break.

With all that being said, the diffuser design is a real challenging problem. The downside to the vaneless space is that vaneless diffusers generate thicker boundary layers, which act to block the flow area of the diffuser, limiting performance. Minimum flow area in the diffuser is key to determining when the diffuser "chokes" and thus the operating range of the diffuser. In engineering there is always a trade-off...this one is between a longer distance to mix out the impeller discharge and a shorter distance to minimize the early boundary layer formation. So, my suggestion is to use a vaneless space of around 1.06 as a good compromise. Keep the diffuser vane numbers slightly higher than the compressor vane number and limit the diffuser passage divergence angle under 14 degrees (since most channel diffuser maps stop at 16 degrees or so).
As for the leading edge,  I think that using as sharp a leading edge as you can make mechanically robust, would allow for the smoothest transition to the flow entering the diffuser passage so that you don't get any weird local accelerations or the like...I'll look around and see if their is any research on that and get back to you.

[Johansson]

What is the benefit of the double vane diffuser (one set of radial vanes and one set of axial) compared to the much simpler to make single vane diffuser?

Spontaneously the double vane looks less efficient since the vane less space around the edge must be hard to predict with regards to the angle of the air entering the axial set of vanes.

[beaker]

Great info there Finiteparts, all makes sense.
Will use your advice. 

That's a good one Anders had the same thought many times.
Also with the double vane how important is the axial vane shape into the combustion chamber?
Its only purpose is to direct and stop swirling. correct? 
Have seen many versions.
Which would work best. have seen both used.

Attachments:

[finiteparts]

I think the benefit of the radial+axial diffuser arrangement is that it was somewhat easier to correlate to available diffuser maps. The radial wedge style diffuser (vane island, etc) is commonly referred to as a channel diffuser because the airflow is really just "seeing" a flow channel while the axially-vaned diffuser is quite common in axial turbomachinery. The important point is that both sections are diffuser stages, by removing the tangential velocity component of the flow, you are converting that portion of the flow kinetic energy into potential energy (pressure). This arrangement is very common in T62s, T700s, T53s, etc...The single channel seems like there may be some benefit to the fact that it doesn't have the sudden flow expansion that occurs behind the tailend of the wedges, which looks like a source of pressure loss...but maybe it suffers from thicker boundary layer growth? I really don't know on that one. Sorry...

If there is a remaining tangential velocity component to the radial diffuser outflow, it is not as efficient to try to recover that pressure in the dump region just upstream of the combustor. Additionally, a swirling streamline has a longer path than a straight streamline and thus will generate more frictional losses and potentially a thicker boundary layer growth. Since the optimum length to width ratio for a typical channel diffuser is around 16, at your given dimensions, the upper bound to what you could achieve by setting the diffuser inlet aspect ratio to one (which is 42 passages and would be awful due to the high boundary layer blockage/losses) is an L/W of around 6. If you pick a really shallow passage expansion angle, say 10 degrees, this still only has you with a pressure recovery, Cp, in the 50 percent range (a general sense from several diffuser maps in Japiske and Baines, " Diffuser Design Technology"). The peak pressure recovery shown for a channel diffuser is around 86% assuming only 2% blockage due to the boundary layer growth...and that is for a single lab tested passage. When you start putting a bunch of them together and getting boundary layer effects, flows entering the diffuser at off angles, pulsing flows, etc., things get complicated fast.

Some general statements about diffusers that might help you pick a design. Vaneless diffusers usually offer the widest flowrange capacity, easiest design, but they do it at the lowest pressure recovery and highest frictional loss. Colin Rodgers states that below PR of 2.5 a vaneless diffuser is "advantageous to employ", because their wide flow range and they have a comparable efficiency to other diffuser types below PR = 2.5 ("Influence of Impeller and Diffuser Characteristics and Matching on Radial Compressor Preformance", SAE Centrifugal Compressors, Technical Progress Series, 1962).

To get better pressure recovery in the smallest package, channel diffusers are used. But, they have much smaller operating ranges due to the amount of area blockage that the vanes impose. Using a vaned diffuser could reduce you choke margin by having a smaller throat area than the compressor inducer. Even if you design for a higher geometric area at the diffuser throat, if the boundary layer is thick enough it can reduce you effective flow area such that you choke at the diffuser throat before the compressor inducer does, loosing choke margin. On the other side, you can reduce you stall margin due to flow separation or shock formation at the diffuser inlet. There is a lot of things going on and the design work is more challenging. 

The middle ground between the vaneless and the channel diffuser is a low solidity vaned/cascade diffuser. Vane setting angle is important, but not as critical as the channel diffuser.  The pressure recovery is not as good as the channel, but it's better than the vaneless...the design is a bit easier.  I think I have a paper somewhere that has some work done by C.Rodgers on showing that airfoil/airfoil cascade data can be used effectively to design the low solidity vaned diffuser.  If I remember that correctly, it could be that you could use Xfoil or something similar to design the vane loading and vane setting angle for the diffuser...



When making your design, think of the passage area and how it changes as the flow passes through it. Sudden area changes can cause flow instabilities/losses and should be avoided. Try to make sure that you are not restricting the flow area in the diffuser throat, since this could reduce your compressor flow capacity due to the lower diffuser flow capacity.

And again, try to avoid making the number of vanes (looks like 14 on your compressor) and the number of diffuser passages the same or even multiple integers of each other. This will minimize the potential for dangerous blade resonance and/or large scale flow issues. A good method is to make the diffuser an odd number so that the 14 per revolution pulses from the compressor don't line up with it's frequency response (i.e. if it has 15 vanes, 15 per rev).

The final thought...there is nothing written in stone here...it is still primarily trial and error...the newest and hottest Large Eddie Simulation (LES) CFD can capture some of the issues, but the flow physics is highly transient and complex. That is one reason that I am trying to design my diffuser system to be able to change it out, or adjust it in some way so that I can play with the big parameters (throat area, setting angle, vaneless spacing, etc). When the Chrysler guys were working on the turbine cars, they had bolt in wedges that they could replace and test. It might be worth trying...see picture. Notice how long their L/W ratio is (length of vane, L, verses the throat width, W)...this was done to efficiently recover the most pressure possible.  Since they were striving for fuel economy, every bit counted and the larger engine diameter was not a big issue. 

By the way, here are some nice papers on Chrysler's work in case you guys haven't run across this site before...  www.turbinecar.com/misc/SAE-Turbine/

~ Chris

[Johansson]

Thanks a lot for the thorough explanation Chris.

[finiteparts]

I am glad someone read the whole thing...ha! I hope it helped. I am still looking to see if I can get a good reference that explains the single passage vs. the two vane set-up on the component level efficiency basis...I'll get back to everyone when I do.

So, I was giving it some thought and maybe the best course of action, since this is being built as a turbojet and thus thrust is the driving parameter, might be to focus on the diffuser design that gives the best flow capability. Since a loss of mass flow DIRECTLY corresponds to a loss in thrust (F = M(Vin - Vout)), but an increase in pressure can have a very marginal increase in thrust. The increase in pressure ratio will give you a better thermodynamic cycle, thus lower TSFC (thrust specific fuel consumption), but how much does it impact the thrust?

So, as I was curious about the actual breakdown of this, I plugged in some numbers. I was using GasTurb 11, and I highly recommend new "turbine designers" download the demo version and play around with the parameters to see how things effect the big picture (thrust in this case)...here is the link:

www.gasturb.de/software.html

If you haven't played around with this, it is an awesome program even with the limitations imposed on the free version and I must really commend the developers for offering such a great program for free demo. It has an awesome "Help" section that will be really valuable to those new to the gas turbine design world.

So back to the comparison...I did a sensitivity analysis and as expected, the thrust showed a 10% increase for a 10% increase in mass flow. The thrust showed only a 1.1% increase for a 10% increase in compressor pressure ratio, but a 2.6% reduction in TSFC was realized.

So what does this mean for the homebuilder? I would say that unless you are striving for peak efficiency, the best course of action might be to minimize the amount of "stuff" that you put in the diffuser flow area. By adding vanes and wedges, you are potentially reducing the flow capability of the diffuser and getting further away from what is shown in the original manufacturers compressor map, which likely used a vaneless diffuser and a scroll housing. So, maybe put 7 vanes with a large open area between each...in the style of low-solidity vaned diffusers. Similar to the one in this picture, but without the variable vanes (unless you are feeling frisky and want to have all kinds of control of the diffuser).



From the numbers calculated above, the potential reduction in flow capability of the compressor system due to mismatch of the diffuser would be more detrimental to the total available thrust than a loss of potential pressure ratio that could be recovered with a more advanced diffuser.

Just a thought...

[beaker]

Thanks Chris for, again very interesting and helpful info.
Makes sense have to put more effort into gasturb.
Had modeled a replaceable diffuser with 15 radial vanes now will also do one with 9. worth a try!
Easy enough to manufacture.
Will add a picture.

[beaker]

Gota get building. 

Attachments:

[JM 765]

hi to all members,

i m almost new member,
are the information on these two threads enough
to build this engine

thanks

[beaker]

Smashed up a diffuser plate this arvo.
15 wedges looked ok in the model but in the flesh it was just too busy.
Flow would be surely restricted.
Will give it a try when engine is up and burning kerro but for now will try an 11 wedge.

[Johansson]

You just smashed one up you say, I think I spent 100 hours on the manual lathe and rotary table to make mine. Bastard.