Air France 447: What the Black Box Tells Us

Here's how to think about the release today of a report based on "black box" data from the Air France crash into the Atlantic Ocean two years ago. Or at least how to start thinking about it. I wrote a number of previous items about the crash soon after it occurred. (Plus, see updates below.)

Thumbnail image for WeatherAirFrance.jpg1) This tells us something, but not everything. The main info in this report, from my perspective, is that the pilots kept trying to pull the airplane's nose up, even as it was entering a stall. I'll explain why that matters in a minute. But there are many things it doesn't address or resolve -- including, as I mentioned soon after the crash, whether a known issue in rudder-control with this model of Airbus plane had any bearing on the crash. Or, how the pilots ended up in the middle of a thunderstorm (right) that other flights were avoiding.

2) The "head pilot was resting" theme probably doesn't matter. Many news reports led off with the info that the flight's captain -- the most experienced, lead pilot on the flight -- was taking his scheduled rest when the problems began, and that the least experienced of the plane's three pilots was at the controls as problems intensified. Anything is possible, but my guess is that it didn't matter. All members of this kind of crew would be highly trained. Moreover, the captain was back in the cockpit within 90 seconds of serious trouble (with the autopilot disengaging), so he would have been part of the discussion about what to do. Deciding on the proper reaction would be more important than executing it with hands on the controls, so the captain would presumably have been involved when it mattered.

3) The plane "stalled," but not in the way you think. The great impediment to accurate coverage of many airplane crashes involves the world "stall." Its normal meaning, to 99 percent of the reading public, is that an engine has stopped or failed. Engines do sometimes fail on airplanes, and in some cases can even stall in the normal sense. But the "stalls" and "stall warning" signals mentioned in the blackbox report mean something entirely different.

An "aerodynamic stall," which maybe is the term we should always use, involves the angle of a wing as it moves through the air. The term of art here is "angle of attack," and it measures how sharply the wing's edge is angled up into the oncoming wind. Bear with me for some illustrations, from this excellent explanatory site. This one shows what angle of attack means.


The next sequence shows what a "stall" means, in aerodynamic terms. In the top illustration, the wing is almost horizontal to the oncoming wind, with a very low angle of attack. Let's say it's zero degrees. It produces no lift.

In the middle drawing, the angle of attack is higher -- let's call it eight degrees. At this angle, wind flows over and under the wing in a way that produces more lift than at a lower angle.

But then look at the third illustration:

There the angle of attack is higher still -- let's say, 15 degrees. But instead of producing more lift, it produces much less. The angle the wind would have to follow across the wing is too steep. Instead the airflow is disrupted and the wing (not the engine) "stalls."

The transition from a high angle of attack, to a too-high angle, can be fairly abrupt. You pull back on the controls, raising the nose of the plane and increasing the angle of attack. You get more lift, and more lift -- and then suddenly you get dramatically less. The wings start to shudder, as they are approaching a stall and losing lift; and then, in a fully developed stall, there's a "break" as the plane stops flying and the nose drops to point straight down to the ground. There are lots of variations involving type of plane, whether you're in a turn, and other factors. But the main point is, an airplane "stalls" not because its engines fail but because the pilots have increased the wings' angle of attack too much. This also means that the plane's airspeed is too low.

So when you read frequent references in the Air France report to "stall warnings" etc, they don't mean that there was an engine problem of any sort.* They mean that, for whatever reason, all three members of a professional flight crew responded to warnings that the plane was flying too slowly/had too high an angle of attack -- by deciding to pull back on the controls. Which leads to:

4) This report raises a new question. The new info from the black box concerns, among other things, the "control inputs" the pilots were applying during the last stages of the flight. What they were doing with the throttle to control power, with the ailerons (to roll right or left), with the rudder (to yaw the nose from side to side), and with the elevator (to pitch the nose up or down). Without the black box there would be no way to know those things for sure.

And the main puzzle, as several of the initial stories point out, is why a team of experienced pilots would have kept pulling back on the controls, to increase the nose-up pitch, when the stall warnings were going off. This is a puzzle because being trained to do exactly the opposite is practically the foundation of learn-to-fly courses. If a plane is losing speed and threatening to stall, you recover by pointing the nose sharply down and adding power (plus other things). This reduces the angle of attack, builds air speed, and allows the wings to start providing lift once again.

Every pilot has done this in practice time and again through his or her flying career. "Stall recovery" drills are part of every basic flying curriculum, every recurrent competency drill, every bit of familiarization with a new airplane. I had not flown an airplane for several months because of my recent stay in China. So when I went out this past weekend for a recurrent-training flight, the instructor put me through a series of stall-recovery drills -- exactly as I expected him to do. [Update: several readers have pointed out that among the many differences between flying a small airplane and a big airliner, especially one with the Airbus's "fly by wire" control system, is that in some circumstances the right response in an Airbus can be to raise the nose. At this point I'll acknowledge that I've reached the limits of my first-hand knowledge, and add this to this list of things about which we await more info,]

Why this didn't happen in the Air France cockpit is the next stage of the mystery to explain. I'm sure there was some reason -- these were trained, experienced pilots -- and it probably had to do with the inaccurate or contradictory airspeed indications, the confusion of being in a thunderstorm at night, specific recovery protocols for the Airbus, etc. But it's not yet fully clear, and may never be, what that reason was.

There is more to say about this tragedy at some point -- including the role of the pitot tubes, what auto pilots can and can't do, Airbus-v-Boeing control differences, and similarities to other airline disasters -- but that is what I have time for now. And, of course, sympathies to all affected by the tragedy.
* My guess is that the NYT story on the black box report may have initially confused the meanings of "stall," based on this correction:
This article has been revised to reflect the following correction:
Correction: May 27, 2011

This article has been revised to show that the engines of the aircraft operated normally throughout the doomed flight.
And in fact, the original version of the story, from a cache, definitely confused the meanings of "stall" and referred to pilots' efforts to "re-start the engines," whereas in fact the engines worked fine throughout.

UPDATE: A reader who is a pilot sends in this note, which makes sense to me.
>>I thought this AP article by Elaine Ganley and Jill Lawless was particularly well written - an accurate and easily understandable portrayal of the aerodynamic principles involved, using the proper terminology, and an overall balance presentation of what likely happened and whether the pilots responded appropriately.
A key passage for me is: "Just over two minutes before the crash, Bonin is heard to say, 'I don't have any more indications." Robert says: "We have no valid indications.' " If that was the case, at night with no visible horizon, I think even an exceptionally well trained and experienced professional pilot would have lost situational awareness and would not have been able to discern the pitch attitude of the aircraft. They may be faulted for not diverting around the storm, but I don't see how they can be blamed for their actions in the cockpit once the problems developed if they had lost all their instruments.<<
Update-update: Comment-on-the-comment after the jump.

An aerospace engineer in Virginia writes in:

>>A comment on this comment:
Reading the BEA report, there's no indication that the crew lost all instruments, simply all instruments dependent on the pitot-static system.  Attitude, derived from a combination of rate and attitude gyros should still have been reliable.  The key difficulty faced by the crew was that the standard practice of using pitch + power to maintain safe flight without air data didn't seem to be working and the problem is that it won't if the aircraft is already stalled.  With the airbus' FBW [fly by wire] system and passive stick, the crew would have none of the force or buffet cues through the side-stick that might have told them this. 

At the same time, if the attitude of the aircraft is nominal [normal], power is nominal, but vertical speed is indicating -10,000ft per minute, the most likely cause is that the airplane is stalled.  Yes, the VS indications could have been bad also, but it's less likely to be wrong than airspeed (requires only static pressure) and the crew had already tried the standard response.<<

Also, here's another and very clear explainer on the whole aerodynamics front.