4

I was reading about Air France 447 and came across the line:

The stall warnings stopped, as all airspeed indications were now considered invalid by the aircraft's computer due to the high angle of attack.

In short, due to pilot error, the aircraft (A330) reached a 40 degree angle of attack while under Alternative Law (due to icing conditions) and stalled at 52 kts. Further confusion came from the stall warning suppressing itself after some period of time.

Roughly 20 seconds later, at 02:12 UTC, the pilot decreased the aircraft's pitch slightly, airspeed indications became valid and the stall warning sounded again and sounded intermittently for the remaining duration of the flight, but stopped when the pilot increased the aircraft's nose-up pitch.

It would seem, if the aircraft was stalled, and the computers knew the angle of attack was still far greater than normal, that it's most reasonable to assume the aircraft would still be in a stall even after other sensor data became unreliable. Also, if no airspeed sensors are reading valid data... seems it's most reasonable to assume the aircraft is in a stall... or on the ground (but wheel load sensors can corroborate when that's the case) - not just give up and assume everything is fine.

It seems to me, that perhaps this stall warning is at least partly at fault for this accident.

Why did Airbus design the stall warning system to suppress warnings in this situation?

SnakeDoc
  • 2,438
  • 3
  • 26
  • 43
  • AF447 didn't stall at 52 KIAS - the 52-KIAS reading was erroneous, due to the blocked pitot tubes (as evidenced by the fact that, 29 seconds later, the airspeed reading jumped back up to 223 KIAS, while the aircraft was still climbing). It stalled later, after the airspeed indications had become valid again. – Vikki Jan 27 '19 at 01:59

2 Answers2

5

Stall warning was inhibited below 60 knots airspeed, in order to eliminate nuisance warnings during taxi and other situations. It is well below the stall speed of an A330 in cruise configuration:

  • landing speed with flaps extended is in order of 135 knots;
  • stall speed in landing configuration would be in order of magnitude of 120 knots;
  • stall speed in cruise configuration way above that.

It was indeed a very tragic circumstance, one of the factors preventing the flight crew from recovering from their situation. The solution however, perhaps in addition to enabling stall warnings for very low speeds in cruise configuration, is to train flight crews in recovering from a fully developed stall. During recovery, the load factor must remain under 2.5 g, the structural limit for an airliner.

This is now mandated by regulation authorities, and the first systems have been certified for training use at airlines. FAA has published 14 CFR part 60 Change 2 for upset recovery training. Alaska Airlines had the first Part 121 simulator certified for Full Stall Training under Part 60 Change 2, in April 2016. The simulator is a B737-800.

Koyovis
  • 61,680
  • 11
  • 169
  • 289
  • 2
    Is this official/authoritative reasoning? I ask, because it doesn't make much sense. If it's a nuisance, why not have a button to silence it, so pilot knows the stall didn't go away on it's own. Or simply have wheel loading sensors to determine if the aircraft is on the ground or not (inhibiting the warning when on the ground only). The flight crew didn't realize they were developing a full stall partly because the stall indicator silenced itself when nosing up (the opposite of what was expected). A full stall will be at low airspeed, which is precisely when you need the warning. – SnakeDoc Jul 07 '17 at 20:32
  • 3
    @SnakeDoc because encouraging the pilots to cancel frequent warnings/alerts can cause them to miss something important. – fooot Jul 07 '17 at 20:35
  • 1
    @fooot I would hope stall warnings aren't frequent when at cruise (at least when the computers think the aircraft is at cruise). – SnakeDoc Jul 07 '17 at 20:37
  • 1
    The backup speed tape now provides direct AoA indication. In a fully developed stall from cruise altitude at speeds lower than 60 knots, you need way more than a beeping horn. But yeah you do have a point for the cruise configuration. – Koyovis Jul 07 '17 at 20:38
  • 1
    @Koyovis Part of the reason a full stall was developed was due to non-linked (physical or otherwise) flight sticks. One pilot was pushing slightly down (correct), and the other was pulling hard back (incorrect), causing the plane to climb until a full stall was developed (they had no idea they were giving conflicting inputs). The pilot continued holding back on the stick since the stall warning went away, and they had no airspeed indicator to make them believe they were in a stall anymore. Several things compounded, but seems this indicator could have helped. Just my opinion :) – SnakeDoc Jul 07 '17 at 20:44
  • @Koyovis Can you cite something official for your answer so I can accept it? – SnakeDoc Jul 07 '17 at 20:44
  • Added reference to changed FAA regulation, and a link to the first solution implemented. – Koyovis Jul 07 '17 at 21:15
  • 2
    @Snakedoc yes agreed on the stall warning and the tragic circumstances. There is a question about adding the stick inputs on this site. – Koyovis Jul 07 '17 at 21:34
  • 1
    @Koyovis Thank you! I'll do some more reading up on the stick inputs... that still boggles my mind. I might just not be an Airbus fan ;-P – SnakeDoc Jul 07 '17 at 21:36
  • 1
    @Sankedoc it is my firm belief that the sticks shoukd be coupled. Other than that I do appreciate the design philosophy of Airbus, their safety numbers are excellent. As are those of Boeing. – Koyovis Jul 07 '17 at 22:25
  • 1
    @SnakeDoc The airspeed indicators were actually working fine after the initial icing had cleared, the pilots just didn't believe them afterwards. I too think the sticks should be coupled but that's not a perfect solution - an ATR recently suffered severe damage after the control inputs differed (https://www.atsb.gov.au/publications/investigation_reports/2014/aair/ao-2014-032/). The better solution is better crew awareness, procedures and training. – Ben Jul 07 '17 at 23:30
  • 1
    @Ben agreed. Proper procedure is for one pilot flying. – Koyovis Jul 07 '17 at 23:38
  • 1
    How do two pilots, in the same cockpit, not notice that (a) they're both giving control input and (b) that they're giving different control inputs? Man, everything about how they crashed this plane is crazy. – Ian Kemp Mar 27 '18 at 22:37
  • 1
    @IanKemp: Possibly because, in the A330, control is via sidesticks on the window sides of the pilots' seats, making it a lot harder to notice what control inputs your copilot is making than it would be on a plane with a conventional yoke (like, for instance, an A300, or any Boeing airliner ever). – Vikki Jan 27 '19 at 01:47
5

The stall warning needs to be an immediate attention-getting alarm that you react to promptly and accurately. If you routinely hear it on landing roll-out or during taxi, and 9,999 times out of 10,000 your reaction has been to cancel it, then on that one time (or less) in a career that you're hearing it for real, there would be a real risk of cancelling it instead of reacting to it. Thus, the need to inhibit nuisance alarms.

The problem on AF447 was two-fold: the defective pitot tube heating setup, and the gross mishandling of the aircraft by the FO. If the pitot tube heat hadn't failed, no accident. If the FO hadn't made control inputs that run counter to every training, every instinct, every habit of flying pilots throughout the world for the last 100+ years, no accident. What that pilot did was so far outside of what's expected that no amount of "normal" systems or procedures was going to do very well dealing with it. You simply don't train for what to do or what to look for in a case where the other pilot does something utterly irrational that defies all common sense and training.

There are things that Airbus aircraft could do differently so that the other pilot knows what sorts of inputs his partner is making, and there are always things that can be done differently in fly-by-wire and warning systems. But at some point, you have to make assumptions that the pilots have been trained and that they won't do things entirely opposite to that training -- especially when you're already in a degraded state without full protections and full air data inputs available.

Enabling nuisance stall warnings & building a habit pattern of cancelling them isn't going to make things better.

Ralph J
  • 51,356
  • 17
  • 157
  • 249