Fly by wire and control laws for EAB FBW

HomeBuiltAirplanes.com

Help Support HomeBuiltAirplanes.com:

REVAN

Well-Known Member
Joined
Dec 6, 2016
Messages
268
Location
Tucson, Arizona USA
@HotWings - To proceed further with these ideas, it would be appropriate to start compiling a list of potential hardware that may be appropriate for a FBW system to use on a HBA. Then we can start to look at details like how large of a control surface can be driven with an actuator and how many surfaces a plane may need, cost and weight, etc...

Since you started this thread, is this something you have already start to look at? If so, please share what hardware you have found so far, what you like and don't like about it. We can start from your list and build on that looking for COTS hardware to make something viable enough that it could be tried.
 

Hot Wings

Grumpy Cynic
HBA Supporter
Log Member
Joined
Nov 14, 2009
Messages
7,636
Location
Rocky Mountains
@HotWings - To proceed further with these ideas, it would be appropriate to start compiling a list of potential hardware .
You reading my mind? Guess I'm going to have to find a source of real tin foil to replace my Reynolds non-stick aluminum foil hat. :gig:

I sat down to take another look at what is available from the R/C world since I last looked = but decided to look at HBA first. My initial impression is that there has been a considerable advancement, especially from the heli world with regard to robust and apparently high quality servos. Of course I'd get more work done If my research assistant had her own work station.
SANY0038.jpgHer tail can click the mouse at the most inappropriate times........

I'm not really interested in promoting FBW, or probably more accurately, augmented servo control (or ASC) in general as I am in adapting it to one of my projects to reduce weight and build time. After laying awake in bed this morning pondering this I'm at the 90% level that I will go down this path. My needs are on the low end of the scale as far as power is needed since I'm looking at a Vne around 110 mph and control surfaces with less than a 12 inch chord/400 in^2 area - well within the bounds of R/C hardware. A ballistic parachute and better than average cockpit crash protection were already part of the plan.
 

BoKu

Pundit
HBA Supporter
Joined
Aug 15, 2013
Messages
2,871
Location
Western US
...It seems to agree with my understanding that the stiffer the surface is connected the less likely mass balancing will be needed...
One thing that often gets lost in these discussions is that flutter is not a control surface problem, or even a control system problem. It is an airframe problem that results from the interaction of pretty much every part of the aircraft, up to and including the pilot. It is a systemic issue that is best addressed with a systemic approach. A perfectly mass-balanced control surface mounted perfectly rigidly on a wing will not prevent wing flutter if the wing itself embodies structural and aerodynamic characteristics that promote harmonic resonances. In fact, a stiffer control surface might even promote flutter if the stiffness is achieved at the expense of extra mass in the wrong spot.
 

REVAN

Well-Known Member
Joined
Dec 6, 2016
Messages
268
Location
Tucson, Arizona USA
One thing that often gets lost in these discussions is that flutter is not a control surface problem, or even a control system problem. It is an airframe problem that results from the interaction of pretty much every part of the aircraft, up to and including the pilot. It is a systemic issue that is best addressed with a systemic approach. A perfectly mass-balanced control surface mounted perfectly rigidly on a wing will not prevent wing flutter if the wing itself embodies structural and aerodynamic characteristics that promote harmonic resonances. In fact, a stiffer control surface might even promote flutter if the stiffness is achieved at the expense of extra mass in the wrong spot.
Yes. The mass distribution is very important to creating problems, or avoiding them. As for the wing, extra mass at the trailing edge will tend to make things worse. If the aileron needs a heavy balance mass to keep the aileron from fluttering, it can take away the margin of the wing itself, because it adds mass to the aft portion of the wing. If the aileron itself flutters, the wing may survive the event, but the plane will lose the aileron and the control it provided, which may then doom the plane from LOC. So there are really three things to be concerned with on wing design, aileron flutter, wing flutter, and worst case would probably be aileron flutter exciting wing flutter. The tail can flutter also, but the wing is usually the most vulnerable as tails tend to be smaller and stiffer by comparison (but each design is unique).

You reading my mind? Guess I'm going to have to find a source of real tin foil to replace my Reynolds non-stick aluminum foil hat. :gig:

I sat down to take another look at what is available from the R/C world since I last looked = but decided to look at HBA first. My initial impression is that there has been a considerable advancement, especially from the heli world with regard to robust and apparently high quality servos. Of course I'd get more work done If my research assistant had her own work station.
View attachment 67597Her tail can click the mouse at the most inappropriate times........

I'm not really interested in promoting FBW, or probably more accurately, augmented servo control (or ASC) in general as I am in adapting it to one of my projects to reduce weight and build time. After laying awake in bed this morning pondering this I'm at the 90% level that I will go down this path. My needs are on the low end of the scale as far as power is needed since I'm looking at a Vne around 110 mph and control surfaces with less than a 12 inch chord/400 in^2 area - well within the bounds of R/C hardware. A ballistic parachute and better than average cockpit crash protection were already part of the plan.
I think the plane you are describing is a good representation for something like this. FBW should not be thought of as a big or fast plane thing. I've been interested in doing FBW for ultralights simply for the weight savings and also because they almost all use BRS systems, just in case.....

Bottom line, I like the idea of starting with something not very fast and not very big to experiment with these concepts. :)

Let's have at it!!!
 

karoliina.t.salminen

Well-Known Member
Joined
Oct 7, 2008
Messages
407
Location
Finland
The control-law math for incorporating things like stability augmentation is daunting, and I'm underplaying it severely by using "only" that word. Got a mere taste of it in school, and came away deeply humbled. The professor did an extremely simple single-axis example on the chalkboard, and mine was only one of a classroom-full of jaws bouncing off desks. I'd never even seen a couple of those function symbols before.
Every Chinese RC hobby shop has ”stabilizer” units for RC planes, intended especially for FPV pilots.
One does not need to understand the math on integral level as finished algorithms exist and are adaptable for free (open source), waiting for just takers.

Then there is an alternative approach airplane folks may not have considered for some weird reason.
Use of deep neural network for solving the ”everything” of the fly by wire system without programming a single function by hand. While deep learning is not really true AI on implementing thinking machines because deep learning does not really learn or think, it is a pre-optimized universal function approximator. However. For solving the problem of fly by wire, it should be completely fine. What would be needed would be a collection of massive amount of data on control surface movements, cameras, gyros, accelerometers, instruments while human is performing some intended flight condition (eg turn, descent etc.). Then put it number crunching for few weeks with Tensorflow, and there you have it, FBW system without implementing a single cryptic function on it. You will have no idea what the functions inside it are, because DL is a blackbox system, you throw data in and something comes out. But then you go and test it out. And do transfer learning to the model if something did not work as desired. The system will require a nVidia GPU system on the plane, same than in Tesla cars, to run the model.
 

Hot Wings

Grumpy Cynic
HBA Supporter
Log Member
Joined
Nov 14, 2009
Messages
7,636
Location
Rocky Mountains
What would be needed would be a collection of massive amount of data on control surface movements, cameras, gyros, accelerometers, instruments while human is performing some intended flight condition (eg turn, descent etc.). .
Reminds me of an old Star Trek episode:

[video=youtube;NgO_y57_078]https://www.youtube.com/watch?v=NgO_y57_078[/video]

Who gets to choose the pilot and plane to fly for the model?
 
Last edited by a moderator:

markaeric

Well-Known Member
Joined
Jul 31, 2015
Messages
346
Location
Wichita
Just thought I'd mention Dynamixel (http://www.robotis.us/dynamixel/) servos as a potential option. They're oriented towards robotics, but have some useful features which I haven't found in even the more advanced RC units, such as a bidirectional digital bus. I'm not sure of all the data you can get from them, but I have seen mention of position and PWM value to the internal H-bridge (from which you could infer load to some extent). They're not cheap, and their electrical connectors definitely aren't aerospace grade.

One thing that crossed my mind, but I suppose isn't much of an issue since it's not uncommon in the RC world is that if you have multiple servos actuating the same object, a difference in the internal feedback between the servos could cause them to fight each other.
 

REVAN

Well-Known Member
Joined
Dec 6, 2016
Messages
268
Location
Tucson, Arizona USA
I found this with a quick search online:

https://www.servocity.com/140-mm-stroke-44lb-thrust-light-duty-linear-servo

$70 for a 45 pound force linear actuator with 5.5 inch throw and weighing only 3.5 ounces. They have much more powerful (and expensive) actuators also, but I'm still leaning toward having lots of little control surfaces instead of big ones with one powerful actuator. What's missing is the actuator dynamics (not that it could be trusted even if they posted something). It would be on the designer to get the actuator and run some gain and phase tests to determine what can be expected from the actuator in a dynamic setting. It may not be that important unless the plane is running a closed-loop stability algorithm, but a Bode plot for the actuator will be needed to model the actuator for any kind of system analysis. Since it is used on RC planes, it seems unlikely it will be so slow as to cause PIO type issues in an open-loop system.
 

Marc Zeitlin

Exalted Grand Poobah
Joined
Dec 11, 2015
Messages
890
Location
Tehachapi, CA
I found this with a quick search online:

https://www.servocity.com/140-mm-stroke-44lb-thrust-light-duty-linear-servo

$70 for a 45 pound force linear actuator with 5.5 inch throw and weighing only 3.5 ounces.... Since it is used on RC planes, it seems unlikely it will be so slow as to cause PIO type issues in an open-loop system.
Ummm, this servo has a 140 mm stroke and a speed of 8mm/sec. So if you're willing to wait 17.5 seconds to go from full throw in one direction to full throw in the other direction, well, have at it, and good luck controlling your aircraft. I would argue that for emergency situations where evasive maneuvers may be required, lock to lock throw should take under one second. Maybe there's some standard for this somewhere, but I'd be willing to bet you that 17.5 seconds is at least 10X over the limit.

Now, just as a reference point, SS2 has redundant electric actuators for the horizontal stabs - once it goes supersonic, the stick forces get so high that the pilot uses the trim button to change pitch AOA through the electric actuators. The actuators were custom, and contracted out for $1M each, IIRC. I also recall that they actually ended up costing over $2M each. Yeah, my eyeballs fell out of my head, too, when I heard the #.

So, there's some actual data for what it takes. Yeah, SS2 is a lot bigger and faster, but the concept is no different. A $70 actuator isn't going to do it and if you want to go 10X faster lock to lock so as to actually have some reasonable response to stick inputs, you will have 1/10th of the force (10X geartrain), so 4.5 lb rather than 45 lb. I don't know of a plane where a 4.5 lb. actuator could be used to run a control surface, but I'm sure I could be educated on the matter.
 

Hot Wings

Grumpy Cynic
HBA Supporter
Log Member
Joined
Nov 14, 2009
Messages
7,636
Location
Rocky Mountains
but a Bode plot for the actuator will be needed to model the actuator for any kind of system analysis.
Making me do more homework!?! :speechles

I think some of the guys here oh HBA have said they using this companies products for trim control. They might be fast enough for main control surfaces, but intuitively I have a hard time imagining that.
 

Pops

Well-Known Member
HBA Supporter
Log Member
Joined
Jan 1, 2013
Messages
9,010
Location
USA.
Added --- I used RC servos of 350 oz,inch torque. Installed in the SSSC of 485 lbs EW that cruised at 80 mph. Servo had lots of torque for the use of moving the trim tabs. No problems in the 2 years of use.
 

Hot Wings

Grumpy Cynic
HBA Supporter
Log Member
Joined
Nov 14, 2009
Messages
7,636
Location
Rocky Mountains
Maybe there's some standard for this somewhere, but I'd be willing to bet you that 17.5 seconds is at least 10X over the limit.

ASTM for LSA:

It must be possible to reverse a steady 30° banked
coordinated turn through an angle of 60°, from both directions:
(1) within 5 s from initiation of roll reversal, with the airplane
trimmed as closely as possible to 1.3VS1, flaps in the takeoff
position, and maximum takeoff power; and (2) within 4 s from
initiation of roll reversal, with the airplane trimmed as closely
as possible to 1.3VSO, flaps fully extended, and engine at idle.
 

REVAN

Well-Known Member
Joined
Dec 6, 2016
Messages
268
Location
Tucson, Arizona USA
Ummm, this servo has a 140 mm stroke and a speed of 8mm/sec. So if you're willing to wait 17.5 seconds to go from full throw in one direction to full throw in the other direction, ....
Yup, that's way too slow. This must be something they use for landing gear, not flight control. I didn't see that at first (overlooked the obvious). I was looking for gain and phase data published at different command amplitudes, but I'm thinking that will be rare to find for these actuators. However any real design will need it. The ones using the hardware are the ones responsible for conducting the testing to verify that is should work in their application.
 
Last edited:

Marc Zeitlin

Exalted Grand Poobah
Joined
Dec 11, 2015
Messages
890
Location
Tehachapi, CA
ASTM for LSA:
Thanks for the pointer - the old version of Part 23 had almost the same thing in 23.157. The new version of Part 23 doesn't have that requirement. But in any case, a 5 second 60 degree roll from one side to the other will allow maybe 1, 2 seconds MAX for lock to lock motion.

Thanks again for the pointer.
 

Hot Wings

Grumpy Cynic
HBA Supporter
Log Member
Joined
Nov 14, 2009
Messages
7,636
Location
Rocky Mountains
Thanks for the pointer - the old version of Part 23 had almost the same thing in 23.157. The new version of Part 23 doesn't have that requirement. But in any case, a 5 second 60 degree roll from one side to the other will allow maybe 1, 2 seconds MAX for lock to lock motion.

Thanks again for the pointer.
First time I read this standard I thought it was the definition of "wallow". ;)
 
Top