electronic turbine controler

[djs500]

I have a Garret GTP-70 gas turbine I would like to convert to electronic control.
My original idea was to use an Allen Bradley PlC with 4-20 ma output to control the fuel using temperature and pressure inputs from the turbine.
(I have a couple of PLC 5's and the software.) 
I have been searching for a PWM injector control when I found the Megasquirt control.
Has anyone here, tried using the Megasquirt controller for a turbine? www.msextra.com/
I did a Google search and found a few YouTube videos where people used the Megasquirt to control the fuel without any feedback from the engine.
I was wondering if I could use the injectors as a flow control for metering the fuel into the existing the existing fuel nozzle.
My reasoning for using the existing fuel nozzles are,
The arrangement of the nozzles in the combustion chambers would make it difficult to mount an injector nozzle in its place and I don’t think the injectors would withstand the heat.
Any thoughts on this?

[cursorkeys]

I wrote a huge response and the forum ate it Sorry, this is a little terser than the original.

Car fuel injectors are designed for typically up to 120 PSI operating. Gas turbine HP fuel pump pressures are typically 1000-2000 PSI unless centrifugal injectors are used.
Biggest 'normal' car fuel injector is 500cc (0.5litres/min). Fuel flow on even a small'ish gas turbine like, for e.g, a RR Nimbus is 4 litres/min.

The very talented Roger Marmion used a MOOG valve to control the governor on his T62: youtu.be/Zk1ZujAkjsY
MOOG proportional valves are high-precision and very, very expensive. I think bog-standard hydraulic proportional valves should work equally well here.


I would suggest it is far easier to write your own code than modify MegaSquirt which is specifically designed for a car.

I wrote a touchscreen supervisor for my GTD-350 and am currently extending it to be a complete control-by-wire FADEC on my Nimbus. I used a ARM CORTEX-M4F microcontroller running FreeRTOS (porting to chibiOS now). I used MAX31855 thermocouple interfaces and for the 2-40mA pressure transducers I used a discrete instrumentation amp layout with a precision sense resistor bridge rather than a module. One trick I used was unity-gain buffers on the input to the instrumentation amps to boost the CMRR and stop generator/starter noise getting in there (obviously this just makes a two op IA into a 'full' IA, but full IAs are getting thinner in the supplier catalogs). Design to automotive standards is a must obviously.

HTH! Jon

[rexhunt]

Hi Cursorkeys,

Have you done a writeup on your control system?

It sounds like something I would like to implement once I get a good stable engine built.

Cheers,
Rex

[wolfdragon]

FYI the 4L/min translates to roughly 400 lbs/min, which while extreme for one injector, is easy to accomplish with multiple injectors, and yes I mean automotive

www.summitracing.com/parts/acc-74120i/overview/ does 120 lbs/hr
www.summitracing.com/parts/acc-74160/overview/ does 160 lbs/hr

I have looked into this, pwm'img these isn't especially hard, you just need a snubber circuit to kill the back emf from toasting your driving FET

I have partial plans and code (using Atmel AVR for the brains) for a legitimate FADEC for a DIY GT, it's just hard to get the time (and with the shutdown, the money) to properly dev this thing...

[turbochris]

just vary pump speed it's much safer than a solenoid PPvalve/injector and more stable. If you want to use a pump driven by N1 like the nimbus has using a small valve in the return line may work better than a big valve in the feed line. The pump in the nimbus is just large enough to accelerate the turbine it don't take much to bleed enough away to make it go under equal and decelerate. Another spring loaded electomagnetically latching arming valve that slams shut on power loss is also a good idea.

[wolfdragon]

I think what he's getting at is trying to not use a servo-valve as they can get quite expensive.

With proper feedback from the engine, failsafe routines aren't terribly hard to implement, it's just a matter of making sure all your controls and their logic(s) are also failsafe (much harder to do)

[cursorkeys]

Oct 15, 2013 19:11:02 GMT -5 wolfdragon said:

FYI the 4L/min translates to roughly 400 lbs/min, which while extreme for one injector, is easy to accomplish with multiple injectors, and yes I mean automotive

www.summitracing.com/parts/acc-74120i/overview/ does 120 lbs/hr
www.summitracing.com/parts/acc-74160/overview/ does 160 lbs/hr

I have looked into this, pwm'img these isn't especially hard, you just need a snubber circuit to kill the back emf from toasting your driving FET

It's the pressure mostly that I was getting at, you'd be operating at maybe 10x the design pressure and I don't know if they would take that?

Plus you'd be dead-heading the engine's HP fuel pump unless you used Chris's idea of (ab)using the governor feedback to control the main fuel flow.

Although on the drive side there are some nice parts like the LM1949 which provides proper peak/hold and over-current features. BJTs/Darlingtons are (IMHO) also a far better choice than FETs/IGBTs for driving highly inductive loads. I did a FMA for a client once on some FET-based motor drives that were failing after a couple of hundred hours. The parts were well within their SOA normally and had a diode clamp, but the dv/dt on turn-off was so fierce it was damaging the gate oxide before the diode conducted sufficiently and the FETs slowly became leakier and leakier. A 1GHz scope had trouble seeing the spike. With more careful snubbing it would have been fine but the client just changed to BJTs and no further problems. I was quite shocked really at the results and have been much more wary of designing with FETs since.


On the fail-safe side for the software you can use windowing watchdog timers and try to stick to a, testable, design methodology that would please DO-178C; probably quorum-of-three hardware-software design is a little too hardcore for this I've always liked hardware fail-safes such as force-guided 'safety' relays (you use one set of the contacts to either wire back to the processor for state-violation detection or as a hardware interlock/both) and mechanical intrinsically-'off' safeties like normally-open solenoids.

Nope, no writeup yet. I've just recovered from swapping processors from Microchip PICs. The PIC32 was just too slow to drive the display acceptably and had no external memory bus. I got double-buffering working by abusing page-flipping on a Solomon Systech TFT driver but it was hacky. These Cortex M-4 parts are awesome and much better suited to driving displays and talking CANbus.
I'll definitely release my design when finished, it's pretty modular. Nothing very fancy though, I've done the electrical design to automotive standards on immunity and the different bits just talk to each-other via CANbus 2.0. The current graphics library I'm using needs changing as its not licensed for use with the ARM parts I've swapped to but apart from that the software is nearly done.

The plan is to control my Nimbus via an inner EPR control loop driven by a 'power demand' control and I've got industrial pressure transducers hung off each station of the engine to get that data. I also have an EGT thermocouple and a fuel flow transducer so those will be the outer control loops. And two accelerometers that get acquired by 16-bit ADCs and then FFT'd, they should make some very interesting graphs as to what vibrates at what speed etc... It's wonderful fun designing it anyway, at least doing design work for yourself there are no annoying things like product price-points to have to hit!

Edit: I have this pic of the display when it had a PIC32 driving it: www.cursorkeys.com/index.php?page=ignitor-replacement-3
It looked ok but I could only get 3Hz or so for full screen updates, even with DMA and a 16bit port trying to get more than 4MB or so of data per second out of a PIC32 is too painful.

[djs500]

Thanks for all the input
The only successful run on my gtp-70 it was running about 100 psi fuel pressure at 90% speed. The fuel supply to the fuel nozzles is supposed to be around 400 psi.
And cursorkeys you are correct the automotive fuel injectors won't function properly at higher pressures (above 100psi or so)
If using Pwm fuel injectors the pump wouldn't be deadheaded as you need to run a fuel pressure regulator to bleed off the pressure.

The garret gtp-70 uses a pneumatic thermostat to control the fuel pressure. If I retain the stock fuel pump and controller I could use a pneumatic 4-20 ma pressure regulator in place of the pneumatic thermostat.

The thing is I have 2 engines, 1 stock needs bearings and another without all the accessories, no fuel pump, pneumatic thermostat, missing 1 fuel nozzle.
I can easily put 1 together from the 2 but it would be nice to have both running.

I like the idea of a variable pressure electric pump. I was thinking about using a fuel nozzle from oil burning furnace which will run in the 100 -200 psi ranges at 7 gallons per hour.
All this thinking has me wondering if I couldn't use a fuel pump from an oil burning furnace, and the fuel nozzles, but control the pressure by bleeding with a pwm injector.

The hydraulic control valve sound like it would work great, but the $800 or so, I see them for on ebay, still has me thinking.

[cursorkeys]

Oct 16, 2013 22:46:34 GMT -5 djs500 said:

I like the idea of a variable pressure electric pump. I was thinking about using a fuel nozzle from oil burning furnace which will run in the 100 -200 psi ranges at 7 gallons per hour.
All this thinking has me wondering if I couldn't use a fuel pump from an oil burning furnace, and the fuel nozzles, but control the pressure by bleeding with a pwm injector.

Why not just just used closed-loop control on the pump?

You use a pressure transducer and a PID Controller to control the pump speed to get the desired pressure. The only 'trick' with PID controllers on pumps is to allow a fixed bleed round the pump of maybe 5-10% of the pumps capacity to avoid the pump ever completely stopping turning. I've successfully used that method on petrol with a fuel injection pump and it works great.

[wolfdragon]

use the temperature coming out the back end and P2 as the fuel pressure will want to stay relatively stable so you know your flowrate thru the injector