Air France 447

This may be the most spectacular accident and investigation in the history of aviation. AF447, an Airbus, departed Rio de Janeiro Galeāo enroute to Paris (CDG) on 1 June 2009 with 228 persons on board. They encountered weather over the Atlantic about four hours after departure. The aircraft never arrived in Paris and the search began. Within five days, floating debris and bodies were found on the surface of the ocean. Four undersea searches in the area over a period of two years finally yielded the debris field on the sea floor 12000' below the surface from which the black boxes were eventually recoverd. Air France 447

In my opinion, the report will resonate more with pilots than non-pilots, as the disconnect between deck angles, angle of attack, airspeed, and power settings against seat-of-the-pants sensations will be more easily grasped. As an old stick-and-rudder, bell cranks and cables kind of pilot, I am a little unnerved by the notion of fly-by-wire and side stick controllers. A lot of pilots with a whole lot of big iron experience aren’t, so my opinion means nothing. Nevertheless, when you read the report, or more importantly, when you read Bill Palmer’s excellent treatment, “Understanding Air France 447”, you may understand why I’m a little concerned about aircraft software that makes choices about whose stick to believe and how those sticks may operate and react depending on which software law is in force. Bill is also an A330 pilot, so you can believe him.

The consensus seems to be a combination of compromised instrumentation, beginning with the icing of pitot tubes (from the storm through which they were flying) causing instruments (principally airspeed and altitude) to not properly render. Initially, that triggered an autopilot disconnect, as well as a switch from Normal Law automation to Alternate Law, and then Alternate Law 2. That exacerbated a condition of a pilot seeming to forget how to fly on instruments (a vast oversimplification), or at least in an Alternate Law mode in which he might not have recognized the differences and limitations of the controls.

Further, as roll excursions as well as extreme pitch excursions took place, airflow to pitot tubes (only temporarily iced) and angle of attack (AoA) sensors were compromised yielding conditions which not only gave misreading displays, but also rendered warnings either intermittent or cut them off entirely.

An attempted rescue by another pilot (these were both FOs, while the Captain was in the rack—normal for long overseas flights) wound up at cross purposes as confusion reigned in the cockpit. The chaos between the two seated pilots was in turn aggravated by the flight controller “law” that caused one pilot’s inputs to over ride the other’s. Essentially full forward stick on one side and full aft on the other. Not a good condition.

As they were on the way down (not discernible by the instruments due to the pitot icing) the Captain came in to, “what’s all this, then?” The aircraft was unrecoverable by this point and, in fact, the whole episode from initial auto pilot disconnect to impact consumed all of four minutes. The aircraft climbed too an altitude near 380,000’ and then plunged to the sea at rates exceeding 15,000 ’/minute. Take a look at the FDR (Flight Data Recorder) data to see how bad it was.

This was a very complex set of circumstances. Software driven airplanes involve a lot of code and documentation to understand them. I always thought I might have been able to fly big iron, but I don’t know if I’d have ever been able to grasp these modern ones.