Air France 447 — Summarizing the Summary

Warning: For those with an aversion to aviation esoterica, or a low boredom threshold, proceed no further.

(Go here and here for TDD's two previous posts on AF447.)

It took two years since the AF447 smashed into the middle of the Atlantic to find its surprisingly well preserved digital flight data recorder (DFDR) under nearly 13,000 feet of water, but only a matter of weeks to arrive at some preliminary findings.

I am going to present this in two parts, the first summarizing and translating the factual elements of the preliminary report relevant to those who want to know what happened, while being less concerned about how they got to what, or background details not specific to AF 447. (The next post will concentrate on the why behind the what. Inevitably, it will be more speculative.)

What follows is distilled from 150 pages, all the while hoping not to create more questions at the end than readers arrived with.

• AF 447 was operating at FL350 (approximately 35,000 feet), Mach 0.82, with a pitch attitude of 2.5 degrees. These are normal cruise parameters.


• The crew altered course 12 〫 left to avoid weather, and slowed to Mach 0.80 due to turbulence (the standard turbulence penetration airspeed, used whenever the subjective estimate of turbulence exceeds "light".)


• Two minutes later, roughly simultaneously, the autopilot and autothrust disengaged and the airspeed on the primary flight display went from 275 to 60 knots. Then airplane began to roll right, followed by the PF (pilot flying) making a left nose-up input. Shortly thereafter, the airspeed on the standby instrument system also failed.


• Because of the lost airspeed reference, the aircraft reverted to the Alternate Law mode of the fly-by-wire flight control system. This mode provides no envelope (pitch attitude / angle of attack) protection. Stall warning is available, but in a degraded mode. Without valid airspeed, the flight control system assumes a speed of Mach 0.3, and generates a stall warning angle at a 10 degree angle of attack.


• The airplane's pitch attitude continued to increase beyond 10 〫 and the airplane started to climb. Vertical speed reached 7,000 feet per minute until dropping to 700 fpm as the aircraft slowed. Roll angle varied between 12 〫 right and 10 〫 left. After roughly one minute, the aircraft reached 37,500 feet, and the primary airspeed suddenly went from 60 knots to 215 knots and M 0.68. Angle of attack (AOA) was 4 〫.


• The stall warning then triggered again, with an AOA of 6 〫 and increasing. The horizontal stabilizer, which controls pitch, went from 3 〫 to 13 〫 over the next minute, and remained there until impact.


• Fifteen seconds later, the standby airspeed returned and was consistent with primary airspeed. The airplane reached 38,000 feet with pitch attitude and AOA being 16 degrees.


• At this point, the very high induced drag attending high AOA caused a rapid loss of airspeed and, without any pitch correction from the PF, there was a further increase in AOA to 40 degrees and descent increased to 10,000 feet per minute. To visualize this, hold a pencil with the tip roughly 15 〫 above horizontal, then move the pencil downward at an angle 40 〫 below horizontal. The aircraft had several roll oscillations reaching as much as 40 〫, after which the PF made an input to the left and nose up stops lasting 30 seconds.


• About thirty seconds after the aircraft entered its terminal descent, both the PF and the Pilot Monitoring said they had no more indications, and the throttles were placed in idle.


• About fifteen seconds later, the PF made some pitch-down inputs, AOA decreased, speeds became valid, and the stall warning resumed. However, while the AOA decreased, the pitch input was insufficient to unstall the wings


• After thirty more seconds, the PF said "We are going to arrive at [10,000 feet]."


• Fifteen seconds later, both pilots made control inputs, then the PF said to the other pilot "go ahead, you have the controls."


• The aircraft hit the water four minutes and thirty seconds after the initial loss of airspeed. The last recorded aircraft parameters were a descent of 10,900 feet per minute, ground speed of 107 knots, and pitch attitude 16.2 〫 nose up. During the mishap sequence, the aircraft heading changed 260 〫, from roughly 010 〫 (just East of North) to 270 〫 (due West).

The event precipitating this crash was flying into an area with some combination of super-cooled precipitation (small water droplets can be well below 0 〫 C and not freeze) and ice crystals in sufficient volume to overwhelm the anti-icing capabilities of the pitot probes (which sense the pressure created by the aircraft's movement through the air). The pitot probes on AF 447 were fully functional, and complied with certification requirements.

An unusual number of airspeed loss incidents in the A3xx (320 / 330 /340) series due to water ingestion or icing during heavy precipitation, but within the allowed operating limits of the airplane, caused a 2001 Airworthiness Directive (AD) to replace the original Goodrich manufactured pitot probes with uprated Goodrich or Thales probes. The latter were installed on Air France A330s.

Further A3xx airspeed losses led to a 2007 Service Bulletin (SB) recommending replacing the Thales probes with a newer Thales design that was in response to A320 water ingestion at low altitudes. Unlike ADs, SBs are advisory.

Because Air France only had problems with its short range A320s, it elected to replace the original Thales probes on the A330 / 340 series only following service failure.

In 2008, A340s suffered 7 high altitude airspeed failures. Air France asked Airbus if the uprated Thales probes would remedy these failures.

Airbus replied that they were developed to deal with water ingestion; therefore, they had no increased ability to deal with icing. Further, Airbus also stated there was no way to completely eliminate icing risk, that all the installed probes exceeded certification requirements, and that there was a procedure for incorrect airspeed indications.

In early 2009, there were two more airspeed loss events, including the first on an A330. Air France then decided to replace all the probes on the A330 / 340 aircraft as soon as the uprated parts became available.

Reiterating "for want of a nail" horribly, the first batch arrived six days before the mishap.