T04 Gas Turbine - RESURRECTED!

[ashpowers]

OK Guys, here it is. A bit of a messy start - I didn't get her ramped up quite enough and was wanting to choke around 60K. Needed a little more starting air to get over the hump.

Right now one NGV runner is choked off and this is with the full bladed turbine. Unfortunately this performance is virtually identical to what it was before with the heavily clipped turbine wheel. Maybe just 1psi more P2 or so. There is a difference in thrust output as it wasn't pushing the turbine across the table anymore. Expected though given the exit angle the gases are at with this turbine wheel.

It is interesting to see how she is running though given that I figured a full bladed turbine would at least make a notable change in her operation but she is not really behaving any differently than with the clipped turbine, RPM and P2 wise.

I also plotted out the numbers on a compressor map and finding interesting results. I would have expected that the plot would have pretty much a straight line but there is a notable parabola in it. Is this due to a difference in how this diffuser functions compared to when this wheel is in its OEM housing? Or is this a result of the diffuser throats starting to choke? It looks pretty linear up to around 1#/min and then no more massflow after that - just RPM and P2 increasing.. Probably the diffuser throats - that's a LOT of air - a good bit more than the design point.

www.youtube.com/watch?v=AumX7hQ9eWs&feature=youtu.be



I'll be pulling her apart here shortly and blocking off another NGV port to retest her tomorrow morning and see how she fares. Just surprised as I was expecting to see a more notable difference in RPM and P2 with this fully bladed wheel... Starting to look a lot like its the NGV and inducer that are affecting the flow of this engine and not so much the exducer.

[racket]

Hi Ash

You're running a 60Trim wheel , at 120,000 rpm your inducer tip is spinning at >1200 ft/sec , add on a vector for the ingoing air and you have seriously supersonic relative velocities between blade and air , the inlet simply chokes from shock waves , preventing more air inflow , it happens with 56 Trim wheels as well , we need sub 50 Trim comp wheels to prevent "problems" with inlet choking , 40Trim wheels for high PRs .

You'll notice that your red line sorta mirrors the efficiency island in shape , they start to go "vertical" once past a 2.2PR .

You might need to block off a couple more NGV passageways to get flow back into more efficient flows , start with one more and see if theres a trend :-)

One thing , you should have been running lower TOTs ........yes , no ??

Cheers
John

[ashpowers]

Hi John,

The large trim wheel unfortunately was a pick a long time ago. There is a 50 trim and the next step down is a 46 trim wheel. Depending on how things go with choking the NGV a bit I would entertain the idea of getting a new comp wheel - would be a direct replacement as far as exducer diameter and shaft diameter but I'm not sure on the tip height. If it is the same or less I would only be looking at a couple of hours of machining.

Got her all ready for another testrun tomorrow morning with an additional NGV duct blocked off.

-Ash

[ashpowers]

OK, had a chance to fire her up this morning and collect some more data.

With two NGVs blocked off there is a slight change in performance which was expected. Just given a crude figure, 18 NGV ducts = 5.5% flow per duct. Adding this 5.5% percent to previous pressures per RPM only gives 1psi blocking off the first duct and 5.8% for the second which is again just a little over 1psi (based on a 20psi P2). She is making slightly higher pressure per RPM as compared to just the single duct being blocked. I took the time to put together the comp map plot and added in the TOTs as well. The lowest figure was a bit odd at 80K she was pushing 1.8bar and that was a little more difficult to see where that point lay in the map but getting higher up in the revs was a good bit clearer given the shape of the RPM curves being more vertical than horizontal.

Its looking like I may need to block off more than just one more duct actually, probably more like two more need to be blocked just looking at how things are trending here.



The TOTs are higher than what I wanted to see especially considering there is no jetpipe on her yet but I know this should improve as I pull the flow back into higher comp efficiency. Even though being high, they are lower than in the previous run with only one blocked duct.

She is better behaved in the lower revs in this configuration than with a single port blocked off. She would idle down to lower revs and would spool back up quicker without wanting to stall but when getting the revs up higher (>110KRPM) the combustion was wanting to get a bit "fluttery". This puzzles me a little bit as I Would have expected combustion stability/efficiency to improve a little. I did inspect the fuel injector ring last night just as a precaution and I am getting good equal flow out of the needles so that's not the issue. It may just be a design issue with the combustor.

I'm thinking that at this moment lets get the flow down into the higher efficiency range of the comp wheel and see how the combustion and TOTs want to behave. It is a good chance that combustion will improve and TOTs will lower but if there is still high TOTs once flows are in check then I'll be looking to go ahead and build a new combustor with 8 vaporizers and a little more volume to it by making the inner flametube liner a little smaller in diameter as the one now is a bit bigger than it needs to be.

Given the change it made by blocking off two ducts vs just one, I'm thinking of actually jumping to 4 blocked ducts and skipping the step of 3 blocked vanes. This will also help to maintain symmetry of flow into the turbine wheel since right now the two that are blocked off are 180 degrees apart from each other and cutting the blocking plates out isn't easy given the geometry of where they are situated. So I'll do that a little later today after getting my business stuff taken care of and should have another comp map plot to look at.

BTW, the bearings are still running very smoothly, however, bearing tunnel temps seem to be rising a bit - in this run after getting the test data I let her sit and burn for another 7-8 minutes or so at around 15-16psi. Bearing tunnel temps were right around 180C in this steady state. This is under the 250C maximum that the races are rated for but on the same token, the bearings aren't rated for these speeds, LOL. I may end up slightly opening the two small jets that supply lube to the bearings to get a little more cooling to them.

It is also looking like I should make the bits needed to mount a Nissan mass airflow sensor to the inlet of this engine. The more I get the flows towards the best efficiency, the RPM curves are a lot more level and it wont be as easy to see as clearly where I am on the map. The sensor does have some pressure drop to it which will work against the engine a bit. Perhaps I could install a pitot tube into the inlet duct and attach a pressure gauge so I can calculate massflow?

[ashpowers]

So I've blocked off 4 NGV ducts now and went to do another testrun. Unfortunately the laser for the tachometer decided to stop working right at the beginning of the run. =(

Instead of bailing on the run I went ahead to at least see what the TOT's were doing at given pressures and compare to previous results. Here's where the TOTs are at:

P2 2NGVB 4NGVB
12 625 625
15 780 700
20 850 775

There is quite a notable difference in the temperatures at these different pressures.

I'm going to get the tachometer fixed up and make another test run in the morning but by the looks of it, this engine is really liking to have an excessively choked NGV, at least as per the maths...

One thing I notice between the two compressor map plots is that there is a general trend in them through the 80K-100K range in both of them. The slope of the line is the same for both configurations. The second run with 2 blocked NGVs has shifted that line to the left which is expected and it actually moved it considerably. The slope of the line up to the choke of the compressor wheel is the same though in both of these configurations. I wanted to discuss that for a moment.

It seems to me that the flow of the system is primarily controlled by the NGV throat area and secondly by the turbine itself. Tightening up the ngv throat caused a shift to lower massflow per PSI which makes perfect sense. But looking at the compressor map, if you were to draw a line right up through the maximum efficiency, the slope of that line is not the same as the slope of the plots through that lower RPM range (before choke).

It would seem to me that the only way I could really get this engine to work ideally would be to have a variable NGV throat so I could keep the flows/pressures right through the compressor's maximum efficiency. While this would work, there may be a simpler solution to this though. I guess this is where the trim of the compressor would come into play? Seems to me that this is what the whole idea of compressor/turbine matching for a gas turbine comes from - matching the flows/pressure of the turbine section to the compressor's performance, or vice versa, however you want to look at it.

From the looks of it it appears that my ngv/turbine section has a large increase in flow per psi of pressure increase compared to the compressor wheel's characteristics. It almost looks like I need a bigger compressor wheel for this turbine wheel to try and match the slope of the flow. I would have expected the slope to be different between these two runs though. Perhaps once I have RPM numbers to go with pressure I will see a change in the slope that more closely matches the compressor wheel's performance..

Getting closer!

[ashpowers]

Just an update here as I have been rather busy lately.

I've been noticing some small imbalance in the engine during all of these tests and while they aren't terribly alarming, they are of concern.

The balancer I built has been helpful and I can get things pretty close - at least, to the point where within the measuring capability of the machine, I cannot make out any imbalance over the existing "noise" produced by the bearings.

But something I've always been aware of is the simple fact that I've never seen a behavior in the signals and variations in rotational speed of the group that accurately fit the soft bearing balancing machine "model".

In soft bearing balancing, the rotational speed of the part needs to be great enough to exceed the harmonic resonance of the part and the cradle structure it is perched upon and by an amount that results in the observed imbalance being 180-degrees out of phase with the actual "heavy spot" of the rotating part.

My machine was constructed with the v-blocks of each cradle tower as part of a horizontal cantilever and with the piezo "buzzer" transducers located directly below the center of the v-block. The transducers were fixed by way of two small pieces of dense foam adhesive tape placed 180-degrees apart around the periphery of the transducer. The cantilever had a small screw in it that contacted the top center of the transducer's face. This arrangement produces the "soft bearing support" by way of the flexibility of the transducer and the flexibility of the foam tape that holds them in place.

In this arrangement, the harmonic resonance of the assembly with the rotorgroup sitting atop was around 2700RPM and was easily observed by watching the sensor's amplitude vary over speed. As speed is increased past resonance, amplitude drops pretty quickly and more increase in speed results in no change in amplitude. This is the point where the phase lag should be at 180 degrees - meaning that where the sensors are seeing the heavy spot is actually 180 degrees from where the heavy spot is.

This makes perfect sense why this occurs.. At low speeds, below resonance, the heavy spot's inertia is pulling the assembly in its direction. The rotating part and the carriages it is suspended upon move WITH the imbalance in the same direction the heavy spot is moving. But an interesting thing occurs as the rotational speed increases... As the speed increases, the inertia of the mass imbalance also increases. This additional inertia can be viewed also as an increase in mass - the effect is the same. So as the inertia increases as speed increases, what begins to happen is instead of the assembly moving with the heavy spot, the heavy spot begins to make the assembly move around IT. At a speed about 2x above the resonance of the system, the mass imbalance has obtained enough inertia that the force it carries is so much greater than the inertia of the part/carriages it is perched on that the measured motion of the carriage is in opposite phase to the location of the heavy spot. Again, the easiest way to envision this is that the part's centerline and the carriages it is mounted on are moving AROUND the heavy spot rather than the heavy spot moving around them.

Although I've known the basic "regurgitated" knowledge of the phase lag phenomenon, I didn't really understand really what it meant until two nights ago.... Still learning I guess.

So, after that thought it occurred to me that my balancer's design is quite flawed. It is flawed in one very simple way - because the vector of motion is vertical, the effect of gravity will play a role in the inertial effects and alter the phase lag effect. I've always noted in the attempts to balance the part that where it showed the heavy side was not 180 degrees from where the heavy side really was at... and it would change as the imbalance of the part changed, which now makes perfect sense - gravity was playing a role in the inertial effects of the weight imbalance.

I say flawed in a very simple way simply because all that needed to be different is rather than the vector of motion being vertical, it merely needed to be horizontal.... LOL

So, with this epiphany, new designs of the machine boiled out of my mind and I've been working since to redesign the balancer. =)

I've picked up several pieces of metal stock to build the new layout and have already started in on the new build. I'll post up some pictures of the new setup shortly.

-Ash

[ashpowers]

As promised: Have made some decent progress on the new balancer - basically have the main frame and both carriages completed. I have some sheet metal parts I cut out (not shown) that will be used to construct the sensor towers and support rod enclosure.

[smithy1]

Or.....you could just send it to me and I'll balance it for you...I have a Schenck CAB690 to play with.

Cheers,
Smithy.

[ashpowers]

Smithy, where are you located? Do you have a users manual for your Schenck?

[smithy1]

Hi Ash,
I'm in Sydney, Australia. Our old Schenck CAB690H has a nice big bed and can accommodate rotors up to 100kgs. I don't have a manual for it but there are places which you can download them, try here:

pdfcrop.biz/ebook/title/schenck-manual-cab-690.html

Cheers,
Smithy.

[ashpowers]

OK, some thoughts here regarding the balancing machine. Something occurred to me during a long drive today regarding the sensors to use on equipment like this.

The sensors I've been using are nothing more than piezoelectric buzzers. Instead of using them to produce sound, I'm using them to measure motion; ergo, I'll call them PBS (piezo buzzer sensor) for short from here on out.

The electrical response of these PBS units is proportional to the rate at which the plate is forced to flex. The actual amount of flex it experiences has no direct bearing on the voltage output of the PBS, but rather, its output is dependent on the rate in which it is being flexed - so it is a velocity sensor.

This presents a particular problem, I *think* when trying to use a sensor of this nature to detect the effect of imbalance on a rotating body. In a soft bearing balancing machine (SBB), the rotating body's degrees of motion are constrained to a single vector. The effect of imbalance in this arrangement will induce a displacement of the part and its supporting members along this mechanically constrained vector as the part is rotated. In an SBB, the part is spun up to a speed THROUGH the resonant RPM and further up to a speed where the amplitude of the displacement drops, levels off, and the phase lag angle of the heavy spot is right around 180-degrees. This is the measuring "window" RPM in a SBB.

Now the effect of the imbalance on displacement within this measuring window is such that as the part rotates and heavy spot approaches the vector "plane" of allowed motion, the inertial effect of the imbalance is causing the rotating body and its supporting members to move opposite of the heavy spot. The actual "center" of rotation the part will exhibit is a point that lay in a position somewhere BETWEEN the principle inertial center of the part and the central principle axis of the part (the geometrical "center", and in this case it would be an axis running between the center-drills at each end of the turbine/shaft). The principle inertial "center" of the part in an unbalanced rotating body is not the central principle axis of its geometry. I refer to it as an inertial center because the mass of the imbalance and the rotating speed of the part combine into an inertial unit (M * V). The vector of the imbalance is the line in which the inertial force is acting but this effect interacts with the total mass of the part and its supporting structure as it rotates. Depending on the relationship between the inertial force of the imbalance and the total mass of the part, supporting members, and in the case of a SBB, the spring response within the supporting structure, the response curve of the phase lag and amplitude over a range of RPM will be also be affected. This is not necessarily a problem as careful observation of the amplitude curve through a range of RPM allows us to identify the measuring window and proceed from there.

However, back to the sensor. As the part rotates and induces a displacement of the part and its supporting members, the relationship of the displacement to the velocity are not the same. As the part approaches maximum displacement, the velocity approaches zero and the acceleration approaches its maximum value. Because of the nature of motion within this system and the response of a PBS, there is going to be an issue. The PBS will show a peak output voltage when the carriage motion is at its greatest velocity which actually occurs at the centerpoint of displacement, not at maximum displacement. Remember, maximum displacement occurs at the point where the heavy spot is 180-degrees out of phase of the direction of displacement and this is the exact condition we are looking for to identify the heavy spot.

I could write the program in such a way that the signals from the PBSs are inverted or have the code look for that zero-value and use this as the heavy spot indicator within the targets. Probably would make more sense to do the latter as the original signal amplitude can be retained within the target display. The signal amplitude indicates maximum velocity which occurs through the mid-swing of displacement and is proportional to the amount of displacement, so it is still as useful metric. But in this new arrangement I should expect to see two of these crossing points per revolution - the carriage's velocity will experience two points of zero velocity per revolution. This doesn't present an issue though since the polarity of the sensor as it approaches this zero velocity point can be used to indicate which direction the carriage is moving.

Or another alternative would be to integrate the velocity component relative to a time base to solve for displacement. This way the two targets would be able to directly report the displacement value related to the angle position sensor. This would be exactly what is needed to determine where the imbalance is and how much it is!

At this point it looks like I'll be needing to revisit the code for my program once the machine has been completed to take proper account of the differences. Fortunately the bulk of the program is built and making these changes wont take long at all do to.... then using a "calibration shaft" I built that has a known mass imbalance built into it I will be able to test what I've come up with.

The science behind rotational "balance" is something I have obviously underestimated and taken for granted. It is now showing me just how little I actually knew about this subject! LOL It's quite fascinating, really. Unfortunately it also appears that I am only just beginning to scratch the surface in this study. The examples I have given above are illustrations of only the most basic of rotational balance problems that exist. When you get into more complex rotating bodies that require multiple-plane dynamic balancing, things get complicated really quickly. I know I'm in for a long haul before I fully wrap my mind around it all. =)

A great document to read through - not completely comprehensive but it does illustrate all of the key principles behind this science.
www.balancetechnology.com/pdf/balancing_basics202.pdf

[ashpowers]

LOL, and just came across this page... In this document they specifically speak to exactly what I was thinking in regards to integrating the velocity into a displacement figure! =)

www.reliabilityweb.com/art07/field_balancing.htm

This also goes to show you just how entirely simplified the word "vibration" really is.

Another solution that may be useful is to use a proximity sensor. This would eliminate any potential errors introduced through measurement limitations of the electronics. Integrating displacement from velocity requires very accurate measurements of the velocity signal as well as a very large number of samples per second. Using a proximity sensor seems to be the best solution to this.....

[ashpowers]

Made more progress on the balancer over the past few days. =) I've built box-style towers for the sensors that are integral to the carriage supports and constructed the mechanical bits to mount the sensors and attach them to the carriages.

For now, the sensor control rods are tack welded to the carriages as I wanted to assemble the bits far enough so I could be able to spin up a part and see what the signals look like. The final image shows an exceptionally smooth sensor curve where virtually ALL of the bearing noise has been eliminated! =) Additionally, what you see here is rotation of the part at only about 500RPM so the sensitivity of this arrangement has significantly improved as well. I still need to make mounting provisions for the phase angle sensor so the image below does not have phase information (green) but it shows the turbine (red) and compressor (cyan) signal.

I have spun the T04 group up on the balancer through a pretty wide range of RPMs and just off the cuff it appears that resonance is somewhere around 1000RPM or so, which is great - plenty low enough that the DC motor drive will have no problems getting her up high enough within the measuring window. Even through a wide range of speeds, the bearing noise is still virtually undetectable. In the previous arrangement it was analogous to having put a microphone on the bearings themselves, the signal was THAT noisy.

This part is balanced well enough right now that spinning it to 10KRPM with compressed air produces no detectable vibration by the touch. I know that is not any real sort of metric of balance but comparatively speaking, this new setup is detecting the really small imbalance with a lot greater signal strength than the previous setup and with the same transducers. I suspect this is because of how the displacement is being transferred to the sensor: in this arrangement the total motion is passing through the full diameter of the piezo disc in one direction whereas the previous setup was putting the displacement load at the center of the disc. The "stiffer" arrangement before was limiting displacement & velocity and thus the signal amplitude as well. So far, really happy with the results! Its looking like this new setup is going to be very much better than before and will also allow me to balance larger items both lengthwise and even more so, diameter-wise.

I'm going to give this route with the piezo transducers a shot rather than proximity sensors to detect displacement. Having such smooth signals now from the sensors will vastly improve post processing accuracy when the velocity signals are integrated to solve for displacement. Proximity sensors are very inexpensive (only a few dollars) so if this route starts showing complications I have at least that other option. I've also got some brand new piezo "knock sensors" from the Nissan 300ZX engines that I think may even be an option.