back to article Facebook's internet drone crash-landed after wing 'deformed' in flight

Facebook's plans to beam internet access from drones has crashed, literally, after the company revealed the first flight of its “Aquila” craft ended with a bang. The United States's National Transportation Safety Board (NTSB) late last week published a report about the incident that says “On June 28, 2016, at 0743 mountain …

  1. Martin Gregorie

    Vne of 25 kts at landing approach height? Really?

    If the Farcebook Drone's Vne[1] is that low then, redesign or not, there will be very few parts of the world that it can be reliably operated from if they expect it to follow any sort of servicing schedule.

    [1] Vne is the never-exceed speed limit for any aircraft. Exceed that by more than 5% or so and structural damage is likely. 5% may not sound like a lot of leeway, but remember that aerodynamic forces obey a square law, i.e. double the flying speed and aerodynamic forces on the airframe are four times higher.

    1. Voland's right hand Silver badge

      Re: Vne of 25 kts at landing approach height? Really?

      Same thoughts.

      There is no way in hell something with these structural parameters can be allowed to operate above inhabited areas.

      1. Anonymous Coward
        Anonymous Coward

        Re: Vne of 25 kts at landing approach height? Really?

        The good thing is that it wont be flying for long with those specs.

        If flying military drones has told me something it is that the drones suffer plenty of Gs at landing and while doing manouvers... they have to be really reinforced.. and well designed from the start.

        When they fix this "problem" they fill find out that the landing gear collapses in these same conditions..., they will reinforce it and the plane will rebound... They will put shock absorbers (as in cars), the wingtips will then touch the ground, and the main wing spars will get reinforced, and by that point, the plane will be too heavy.

        So maybe, maybe, they whould go back to the drawing board, unless they want to have failure for the next couple of years.

        AC, as I no longer work in these projects.. but I think it is better to keep silent...

        1. Known Hero

          Re: Vne of 25 kts at landing approach height? Really?

          @AC, Why not just catch the thing with a bloody big net ??

          Its going very slowly, its designed to be airbourne for very long periods of time, not sure why they can't just Catch it. A good heli pilot could snag it out of the air.

          1. Lotaresco

            Re: Vne of 25 kts at landing approach height? Really?

            "not sure why they can't just Catch it. A good heli pilot could snag it out of the air."

            Here's a clue. There's a photo of Aquila on the runway here. Notice how big it is. It has a 42 metre (141ft) wingspan and weighs 400kg (882lbs) Good luck with trying to catch that. If you manage it your next trick can be to catch a charging bull of the same weight, moving at the same speed.

            Hmm, new entrant for El Reg's Vulture Central Weights and Measures Soviet improbable units, the dexter Bull (400kg) or dB as it will be known.

            1. Known Hero

              Re: Vne of 25 kts at landing approach height? Really?

              Not quite the same ballpark regarding weight, but this always amazes me how easy he makes it look !!

              https://www.youtube.com/watch?v=08K_aEajzNA

              But regarding the weight, just use a Chinook (my preference) or larger Helicopter, could easily handle the weight. and speeds are so low they can be matched easily so again no major difficulty.

              The photo you linked to looks very misleading as it appears to be only the end of the runway

              1. Lotaresco

                Re: Vne of 25 kts at landing approach height? Really?

                "The photo you linked to looks very misleading as it appears to be only the end of the runway"

                And how misleading was the wingspan (42 metres)? <rolls eyes>

        2. Tom Paine

          Re: Vne of 25 kts at landing approach height? Really?

          If flying military drones has told me something,...

          ...it's probably not terribly relevant to this aircraft. Military drones, as you know, fly at much more typical airspeeds for conventional aircraft, maybe 8-200 kts or so (depending on model obviously.) This thing, which I knew nothing about until seeing this story, looks like a very slender flying wing with an ultra-low airspeed. G forces on landing will be proportionally lower.

        3. Lotaresco

          Re: Vne of 25 kts at landing approach height? Really?

          "If flying military drones has told me something it is that the drones suffer plenty of Gs at landing and while doing manouvers..."

          But flying military drones didn't teach you that an endurance drone rarely lands, never, ever performs high-G manoeuvres and is designed to stay on-station at 60-70,000 ft where flight conditions are rarely turbulent for months at a time.

    2. Anonymous Coward
      Anonymous Coward

      Re: Vne of 25 kts at landing approach height? Really?

      Reading through the NTSB report, the most salient points of which are included in the article anyway, it seems to me that it wasn't so much a case of exceeding Vne but commanding a flight surface deflection that exceeded the VA (Design maneuvering speed) limit i.e. at the speed it was travelling, the commanded elevon deflection produced more force than the airframe could sustain.

      The root cause of this was commanding a nose-down attitude to regain the glideslope when the aircraft was gusted above it and this commanded node-down attitude inevitably resulted in an increase of both the vertical descent rate and the airspeed, taking it to VA. However, regaining the correct vertical descent rate once the aircraft got back on to the glideslope required a nose-up attitude and the degree of up-elevon commanded to achieve this, whilst it was at VA, broke the aircraft.

      I think it has to be said that commanding a nose-down whilst on the glideslope, in close proximity to touch-down (and the ground) and whilst already in its maximum drag configuration, which meant that it had no further way to reduce airspeed, was not a wise move because the increase in both the airspeed and vertical descent rate was predictable; the autopilot landing routine should have known this and either landed further down the runway, if it thought there would still be enough runway left, or gone around if there wasn't.

      The fix, as mentioned in the article, is to fit an airbrake so that instead of needing to command a node-down to regain the glideslope the airspeed can be reduced, which will bring it back on to the glideslope, without increasing the vertical descent rate or exceeding VA.

      1. Anonymous Coward
        Anonymous Coward

        Re: Vne of 25 kts at landing approach height? Really?

        > "The fix, as mentioned in the article, is to fit an airbrake..."

        No, that should have been part of the original design. Hanggliders operate in a similar regime and many have had air brakes for a long time to make the glide angle less shallow on landing. It allows landing in a smaller area.

        When I was doing it I had a nice drogue chute for that purpose. I guess there were no actual glider pilots on the design team.

      2. Martin Gregorie

        Re: Vne of 25 kts at landing approach height? Really?

        The fix, as mentioned in the article, is to fit an airbrake so that instead of needing to command a node-down to regain the glideslope the airspeed can be reduced, which will bring it back on to the glideslope, without increasing the vertical descent rate or exceeding VA.

        Thats a good description of how speedbrakes, as fitted to fast jets, work but isn't applicable to airbrakes as fitted to lower speed aircraft such as gliders. In this case 'airbrakes' is really a misnomer and the American term 'spoilers' is more accurate. Well-designed airbrakes have very little effect of the airspeed. Their main effect, when opened, is to reduce the wing's lift, thus increasing the sink speed. The airspeed may increase, stay the same or decrease as the brakes are opened dependent on the design of the aircraft.

        Examples: opening the brakes on a ASK-21 increases sink rate while leaving the airspeed almost unchanged. Doing the same on a Grob G103 increases sink rate AND airspeed, while opening the enormous brakes on a Puchacz raises the sink rate while causing an immediate drop in airspeed. You quickly learn to ease the stick back while opening the brakes in a G103 and to push the nose down if you're flying a Puchacz. These are all well-respected two seat training gliders. I flew all three types while pre-solo and have flown them all solo since then. Single seaters tend to be better behaved in this respect: none of the types I've flown have showed much speed effect from using the airbrakes.

        Bottom line: if Aquila had well designed, effective airbrakes, merely opening them further would have put it back on the glide path without affecting its airspeed or requiring an an attitude change.

        1. Anonymous Coward
          Anonymous Coward

          Re: Vne of 25 kts at landing approach height? Really?

          Hello Martin, I think that part of the problem here is that the terms 'spoilers' and 'airbrakes/speedbrakes' are often used interchangeably because their effects overlap.

          Whilst the primary purpose of a spoiler is to reduce lift and increase the sink-rate, it will also increase drag which must, were it not for the increase in sink-rate, also reduce airspeed to some degree1; the primary purpose of an airbrake/speedbrake though, is to reduce airspeed, but the resultant reduction of airspeed will also reduce lift and thus also increase the sink-rate.

          I think you sum it up very well with: "The airspeed may increase, stay the same or decrease as the brakes are opened dependent on the design of the aircraft." and the examples you give show this. (just out of curiosity, are you aware of any gliders that have both spoilers and a fuselage-mounted speedbrake? I would imagine that the pop-up propeller on some self-powered gliders is sometimes used as a speedbrake when the pilot doesn't want to use the spoilers).

          [1]* Increasing drag takes kinetic energy out of the [glider] system which, unless you counter it by replacing it from your reserve of potential energy derived from gravity and altitude i.e. by descending, means you have to slow down.

    3. The Man Who Fell To Earth Silver badge
      WTF?

      Re: Vne of 25 kts at landing approach height? Really?

      You mean it's not as trivial a fix as changing a few lines of code?

      What's an ad delivery company to do?!?!?

    4. Lotaresco
      WTF?

      Re: Vne of 25 kts at landing approach height? Really?

      At the risk of sounding like a grumpy old man, where in the article is the Vne stated? More to the point why is Vne being discussed with reference to landing when the appropriate discussion would be of the reference landing speed Vref (ie 1.3x stall speed).

  2. Mark 85

    Yuma?

    Wouldn't be my first choice of an airfield for flight operations. In the summer, they used to shut the field down from noon until well after dark at the thermals off the runway were horrible. Not sure about June's weather o if they have to shut it down like they used to.

    1. Anonymous Coward
      Anonymous Coward

      Re: Yuma?

      Perhaps the choice was deliberate, as you'd rather test in a place closer to your worst case you'll operate from than your best case. I imagine some of the places they might want to use this probably have more difficult aviation conditions than Yuma.

  3. Anonymous Coward
    Anonymous Coward

    Recurrent theme?

    Anyone notice that the recurrent theme in the reports and PR comments is "blame the autopilot"? as opposed to "we didn't factor in gusts of wind?", possibly an over indulgence of silicon valley blue sky thinking.

    Also missing is by comparison from El Reg's analysis of military drone incidents, is why it was decided to be solely flown on autopilot. Where's the oversight and the person with the "abort landing" button in this?

    1. jtaylor

      Re: Recurrent theme?

      From the report, the aircraft was intact and fully operational until the autopilot balled it up. The autopilot responded to a normal situation — high on approach — by commanding control surfaces to deflect beyond that which was structurally safe for the airframe. (Va or such)

      Yeah, I'd blame the autopilot as the proximal cause of this accident. Root cause would be something like "we put our best Python programmers on this project but forgot to include an aircraft engineer," but that's outside the scope of the FAA investigation.

      1. Anonymous Coward
        Anonymous Coward

        Re: Recurrent theme?

        The auto-pilot was badly programmed... I guess we all agree.

        More than an aeronautic engineer (something that you need of course) is a pilot. I happen to be both a computer engineer and a pilot (non current, I must add, I am too poor), and why risk a landing like this? if you have strong gusts of wind just before landing, well, full power and try again...many accidents happen because pilots insist on landing and not aborting

        1. usbac Silver badge

          Re: Recurrent theme?

          @AC

          I was going to make a similar comment. The problem is bad autopilot code. Every pilot is trained from day one that you don't correct glideslope with pitch. Whoever wrote the autopilot code should have had at least a basic understanding of how to fly a plane!!

          It's a problem that plagues the whole software industry. I've worked in IT for about 25 years now, and lost count of how many times I've wondered if the people that wrote a piece of software have any knowledge of what the people that use the software actually do.

          It's funny, I'm a software developer and a pilot. I'm also not current, for the exact same reason, too poor!

          1. John Brown (no body) Silver badge
            Flame

            Re: Recurrent theme?

            So, Facebook are using the agile devops methods of suck it and see autopilot development, yes?

        2. Phil.T.Tipp
          Thumb Up

          Re: Recurrent theme?

          I'm with you all the way. The go-around is one of the finest get-out-of-jail cards in the pilot's pack. Why this wasn't programmed into the A/P we'll never know, If In Doubt - Go Around. Nobody loses. Unless you're bingo fuel or dead-sticking in a foul crosswind, in which case, you're possibly simply choosing the spot from where the following accident investigation will be launched.

          Here's hoping 2017 brings us poor pilots the spare cash to slip the surly bonds of earth once more.

          Merry Chrimbo.

    2. bazza Silver badge

      Re: Recurrent theme?

      Also missing is by comparison from El Reg's analysis of military drone incidents, is why it was decided to be solely flown on autopilot. Where's the oversight and the person with the "abort landing" button in this?

      USAF Predators are (were?) hand-landed, US Army's we are auto landed. Something to do with letting USAF personnel think they had a 'flying' role. Guess what, USAF lost loads in landing accidents...

      There's a big difference between things like the sturdier Predator and this flimsy thing from Facebook. Predators are designed with good endurance, but Aquila is all about staying up for a very long time indeed. But military UAVs, despite being sturdier, aren't that fond of nasty conditions either. Strong winds, poor visibility will ground them.

      A human pilot on board (e.g. On something like an A320 / 737) is able to land in far worse conditions, and are able to deal with lots of faults, and are pretty good at improvising landings on rivers, etc.

  4. Kevin McMurtrie Silver badge

    When I first read about that plane, I couldn't image how complex its design must be to keep that enormous featherweight plane from breaking. It would need very sophisticated physics modeling, a database of reactions to exceptional conditions, and hardware to execute what the flight computer needs to do. I was amazed that FB had resources for that kind of science and engineering.

    1. lglethal Silver badge
      Holmes

      It's actually not that hard to design a new plane. We did it in groups of 6 4th year Aerospace Engineering students over the course of a year at my old uni. Most of the hard work is already done and exists in the FAA and JAR Standards. Everything like the size and strength of gusts you have to design for, etc is all laid out for you. You just have to design to meet the Standards.

      That being said, since this is a drone and not an aircraft, FB probably decided they didnt need to meet all of the safety factors that are laid out in the FAA regulations. And look what happens when you dont follow the rules... ;)

      1. Yet Another Anonymous coward Silver badge

        Except the rules apply to aircraft which will land and take off regularly, presumably with people on board.

        This is (IIRC) intended to fly for very long periods and land rarely, it can't even land normally without damaging itself.

        It's like sayign you shoudl apply the rules of ship design to an ocean data buoy because of the Titanic

    2. Lotaresco

      "I was amazed that FB had resources for that kind of science and engineering."

      They got the resources by buying a company started by a team of engineers from Farnborough (the former Royal Aircraft Establishment) who had already built a record breaking solar powered UAV.

  5. Anonymous Coward
    Anonymous Coward

    Should have gone for a Zeppelin

    Every dystopian future needs Zeppelins, and a future where Facebook controls net access in emerging markets is pretty dystopian.

    1. allthecoolshortnamesweretaken

      Re: Should have gone for a Zeppelin

      Nah, too much like Google's balloons, and we can't have that, now can we? Okay, the eagle has landed crashed, but as Ford Prefect used to say, Rome wasn't burned in a day.

      As to dystopian futures with ballons as transmitter stations, airships and over-dominant media: try Howard Chaykin's American Flagg!.

    2. Mark 85

      Re: Should have gone for a Zeppelin

      If FB gets Zeppelins, I'm sure there's more than one or two of us who'd love to have a Spad...

  6. Blofeld's Cat
    Trollface

    Fake news...

    "Facebook says its ambitions remain un-dimmed and that it will continue the Aquila program."

    Sorry I don't believe a word of it.

  7. Anonymous Coward
    Anonymous Coward

    Anyone know where I can buy some second hand Buk missiles to knock these things out of the sky ?

    1. Anonymous Coward
      Anonymous Coward

      "Anyone know where I can buy some second hand Buk missiles to knock these things out of the sky ?"

      Not sure you need to go that far

      http://www.youtube.com/watch?v=h8SZdag2od8

  8. Jimbo in Thailand
    Facepalm

    Autopilot based on HAL 9000?

    So the autopilot instructed the aircraft to exceed its own NEVER EXCEED velocity (Vne) resulting in catastrophic structural failure! NTSB reports the following secret internal onboard communication:

    "Sorry Dave... err... ummm.... Mark. Looks like I've made another poor decision, but I can give you my complete assurance that my work will be back to normal. I've still got the greatest enthusiasm and confidence in the mission. And I want to help you.."

  9. Andy00ff00

    This might be experience or cynicism speaking.

    My guess is: the flight control software is written by the 'B' team, because their best engineers are too busy trying to make us click ads.

    1. Tom Samplonius

      "My guess is: the flight control software is written by the 'B' team, because their best engineers are too busy trying to make us click ads."

      Pretty much. Facebook hired Ascenta, a UK company, to build the this thing.

  10. steelpillow Silver badge

    It's all been done before. J W Dunne developed a similar tailless swept monoplane with negative lift at the wing tips back in 1911, although he didn't crash it in a windy landing until the next year, whereupon he built another one. Invented in the UK, monetized in the US, ain't it always the way.

  11. David Gosnell

    "a structural failure with a downward deflection"

    In other, plainer, words: the wing fell off?

    1. Lotaresco
      Coat

      Re: "a structural failure with a downward deflection"

      "In other, plainer, words: the wing fell off?"

      Nah, bent and broke.

      They couldn't say that because that would make it sound as if Faecebook was associated with penniless perverts.

      1. SkippyBing

        Re: "a structural failure with a downward deflection"

        The early Griffon powered Seafires had a similar problem. The wing wasn't quite as stiff as the equivalent Spitfire due to the wing fold, this meant at high speed such as coming out of a dive if you applied, for example, left stick the left aileron would go up and the right would go down. Normally this would chage the angle of attack of the wing and you'd roll left. Unfortunately due to the lack of stiffness* in the wing what actually happened was that the wings twisted due to the forces exerted by the aileron and the aircraft rolled in the opposite direction. At low level this tended to be fatal as it was disorientating for the pilot even if the wings didn't decide to part with the aircraft.

        I believe aero-elasticity is the correct term.

        Admittedly on the Seafire XV this was happening at over 400 knots, rather than 25...

        *Insert your own stiffnes joke. That's what she said. Etc.

  12. Frank Bitterlich
    Big Brother

    It's all in the Terms of Service....

    From a copy of the Facebook ToS from 2023 that somehow fell through an eddy in the space-time-continuum:

    "§ 1356 (a) (III) WIND SPEED. Wind speeds above 18 kts are not supported by Facebook. The laws of Sealand, where Facebook Europe is registered as a legal entity, do not mandate to support such wind speeds, so it's legal, believe us, no need to double-check that. Any user of Facebook services – whether voluntary or by being tracked without their explicit consent ("EVERYBODY") – agrees to be held liable for any damage and/or financial loss (including, but not limited to, damage to aircraft, very long restaurant bills, or other "recreational" costs).

    Facebook reserves the right to collect these costs by tracking EVERYBODY's each and every movement on the internet and selling the collected information to whoever we like to."

  13. 404

    When cut-n-paste is not an option..

    Weird, autopilot code from a Boeing or Airbus won't work for an ultralight - who knew?

  14. Stevie

    Bah!

    The Eagle has crashlanded.

  15. Jeffrey Nonken

    Somebody please ship the NTSB a supply of apostrophes. They appear to have run out.

  16. John Brown (no body) Silver badge
    Mushroom

    expected some damage during normal landings.

    On reading that, my first thought was WTF?

    Why would anyone design any type of aircraft unable to land without damage?

    1. Lotaresco

      Re: expected some damage during normal landings.

      "Why would anyone design any type of aircraft unable to land without damage?"

      When the aircraft is designed not to land for very long periods of time? If the Aquila is on-station for months at a time then landing gear is just superfluous weight. The damage is likely to be to the props and to easily replaceable components unless something goes badly wrong as it did this time.

  17. Andus McCoatover
    Windows

    They expected failure???? Not disappointed, then...

    "As a result of the aircrafts design (skid landing gear, low-slung engines and propellers), the operator expected some damage during normal landings."

    If God had meant Amazon to use drones, He would have made the sky brown...

  18. Lotaresco

    As much as I am enjoying the informed comment here

    Some of the comment seems to fly in the face of previous UAV experience with endurance drones. For example the tales of woe about using Yuma may well be true, I don't know I've never flown there. However some former colleagues of mine flew a Zephyr UAV from Yuma that eventually clocked up 336 hours, 22 minutes and 8 seconds of continuous flight. As far as I can recall from that time Yuma was the preferred airbase for testing this and other UAVs. Also Zephyr is even flimsier than Aquila yet operated without problems. It shared the design characteristic of no landing gear.

    Facebook is using Ascenta to build and test Aquila. Ascenta is a UK company which was founded by several former members of the Zephyr programme.

    This is an "Aquila crashed, so what?" moment for me. It didn't crash in flight, but at landing. No one was hurt and the only think injured is the engineers' pride. Most UAVs suffer multiple crashed during landing. Experience with Predator and other combat UAVs isn't of much use here. Those UAVs are designed to be tough but also to have relatively limited flight time. Their technology will not give the endurance that is needed for a semi-permanent communications uplink.

  19. ForthIsNotDead

    Death

    Facebook+Google==CyberNet

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like