Understanding Air France 447

by Palmer, Bill

It is not likely that someone would ask another person if they were OK, unless they had a reason to believe otherwise. This exchange may be a clue that either or both first officers were not well rested.

  Captain Dubois powered his seat to the full aft and left position to allow himself to get out and Robert to get in. As the seat moved, its distinctive motor sound is heard on the voice recorder. When Robert took the seat, he never moved it any farther forward to the normal range. That would have been a comfortable position to sit in to put his feet up on the foot rests that hang below his instrument panel. His seat was found in the far back position in the wreckage. From this aft position it would have been very awkward to fly the airplane should he need to—and he would need to.

  First Officer Bonin provided a minimal briefing to First Officer Robert prior to the captain’s departure. The turbulence ahead and the inability to climb due to temperature were mentioned. Bonin asked the captain to remind them of the HF communications frequencies currently in use and assigned ahead. Captain Dubois provided them and then left for his break at about 02:02.

  After leaving the cockpit, a pilot will typically use the lavatory, perhaps chat with the cabin crew for a few minutes before taking a nap in the rest area. As the captain is said to have had a companion on board, that would be another reason for him to not head straight to the bunk area.

  Eight minutes later, at 02:10:04, the airspeeds become unreliable and the autopilot disconnected.

  At 02:10:53, 49 seconds after the autopilot disconnected, the two first officers knew they needed help, and a call to the crew rest area was made, where they assumed the captain to be. Five more calls were made over the next 34 seconds. Unlike the banging on the wall that First Officer Robert acknowledged calls to the bunk with, the calls for the captain went unanswered. Robert asked “[expletive] where is he, eh?” “[expletive] Is he coming or not?”

  The autopilot disconnect alarm is loud enough that you can hear it in the bunk, but not usually in the cabin. Had the captain been in the bunk and heard it, he would have known that was an unusual situation. Nobody turns the autopilot off at cruise, in fact with 1,000 foot vertical separation at altitude, its use is required.

  Seconds after the autopilot alarm sounded, the airplane pitched up from 0 to 11° nose up, the g load increased to 1.6 g’s, the vertical speed increased to 5,200 feet per minute, and the airplane roll oscillated left and right. The captain would have felt and noticed the increased g load of the pitch up and repeated side to side rolling motions. That may well have prompted his return to the cockpit without being called, to see what the heck was going on. If he was back in the cabin, and not in the bunk, he most likely did not hear the other two pilots calling the bunk area, which explains their frustration in the cockpit when he did not return immediately upon their calls.

  At 02:10:56 a female voice was heard on the inter-phone saying, “Hello?, yes?” No response was given and the cabin handset was hung up. Three more calls were made and at 02:11:18 two knocks were heard on the wall. Two more call chimes were recorded and another female voice on the inter-phone.

  At 2:11:42 the captain returned to the cockpit. He had been gone only 11 minutes, and only about a minute and a half from the first call to the crew bunk. I suspect he came back on his own initiative, for when he returned he did not inquire, “What do you want”, but “What are you doing?”

  In less than two minutes from the time the autopilot disconnected until the captain returned, the airplane had climbed from 35,000 feet to a few feet shy of 38,000, reaching a peak vertical speed of about 6,900 per minute (an extremely high and very unusual climb rate - over 10 times what would normally be used.) The pitch attitude had been as high as 18°, equal to the typical maximum pitch after takeoff.

  At the captain’s arrival, they were descending back through 35,000 feet and the pitch was about 15° nose up. The normal pitch attitude for this phase of flight is 2.5°-3°. The angle of attack had increased from about 3° to over 30°, but that was not displayed, and the airplane was descending at at almost 10,000 feet per minute in in a deep stall. In addition to the turbulence from the storm, the airplane was shaking due to the stall.

  In response to Captain Dubois’ “What are you doing?” inquiry, the two first officers stated, “I don’t know what’s happening,” and “We’re losing control of the airplane.” Both statements were true.

  Panic, Confusion, or Fatigue?

  Several articles have concluded that ‘Panic brought down Air France 447.34’ While that makes for a great headline, I do not think it has been proven that an actual state of panic existed.

  Panic is defined as, “a sudden overwhelming fear, with or without cause, that produces hysterical or irrational behavior, and that often spreads quickly through a group of persons or animals.”

  The Final Report states in paragraph 2.2.1, “In the first minute after the disconnection of the autopilot, the airplane exited the flight envelope. Neither of the two crew members had the clarity of thought necessary to take the corrective actions. However, every passing second required a more purposeful corrective piloting input.” The report also referred to “the startle effect and the emotional shock at the autopilot disconnection.”

  Confusion and being overwhelmed by an unexpected flurry of alarms and alerts that simply exceeded the pilot flying’s ability to cope with it all, is a likely scenario.

  It was dark and they were experiencing moderate turbulence. No visible outside cues existed. Simultaneously, there was an autopilot disconnect siren, the “cavalry charge”. The airspeed dropped low and momentarily disappeared, the flight directors disappeared. The indicated altitude dropped 300 feet due to the erroneous airspeed data. “Dings” sounded alerting the crew to the ECAM messages telling of the reversion to Alternate Law and the loss of autothrust. Soon after, the C-chord (indicating a departure from the selected altitude) sounded along with a synthetic voice announcing “STALL STALL.” (which sounds like “STOLE” with a British accent).

  First Officer Robert asked, “What is that?” What he is referring to is unclear, but he could be referring to the stall warning that had started sounding one second earlier. He had not had stall training since his A320 checkout nearly 10 years earlier. Additionally, “stall” is announced in English to the French speaking crew.

  The flight controls degraded to Alternate 2 Law. In the roll axis, the airplane will no longer maintain a constant bank angle (or wings level) automatically. The ailerons were under direct pilot control. The roll sensitivity had nearly doubled that in Normal Law, up to 25°/second.

  The airplane rocked to the right and required an immediate correction to the left. Why Bonin also pulled roughly half-way back on the sidestick is the great mystery. Did he overcompensate for the 300 foot loss of indicated altitude? Was he trying to recall the loss-of-airspeed training exercise performed months earlier? That event in the simulator was performed after takeoff, where the proper pitch attitude was 15° nose up. The proper action in this case would be to maintain his current pitch attitude of about 3°. It is likely the stall warning was never encountered in that training scenario.

  He may have been trying so hard to control the roll, that the pitch simply went unnoticed until Robert pointed it out.

  Having duplicated the situation in a simulator myself, the pitch up action that Bonin used requires little effort and happens quickly. There are a number of alert sounds that go off and it would be easy to see how one would tune them out in an effort to load shed in what could be a quickly overwhelming situation. When the pitch up was first made, the stall warning sounded almost instantly, but only briefly. As the indicated airspeed fell below 60 knots, the stall warning was disabled. Once the g load from the initial pitch up subsided, it would have silenced on its own anyway, as the angle of attack returned to a normal range. Within a few seconds the pitch attitude was at 15° and the vertical speed in excess of 6,000 ft/min. Once established, the nose up attitude would have been maintained effortlessly (even han
ds off) due to the stability of the flight control pitch law.

  Due to the loss of reliable air data, AF447 the flight control law degraded to Alternate 2 Law, and the airplane provided no pitch-down tendency on its own to help correct the errant control inputs and decaying speed. Airspeed depleted quickly as the airplane converted that airspeed to altitude, climbing from 35,000 to nearly 38,000 feet in less than 30 seconds.

  For the crew of AF447, the ever-reliable flight directors disappeared and reappeared several times. But after the first time they reappeared, their pitch guidance mode had changed from altitude hold to maintain the vertical speed that existed each time they came back on. This was indicated, but it is easy to miss. The flight director commands are not recorded, but their commands were calculated as part of the investigation. Did Bonin follow them? The recorded pilot commands and calculated flight director positions seem to indicate that is a distinct possibility. The second time they reappeared the vertical speed was 6,000 feet per minute up, and that was the flight director target for 10 seconds before they disappeared again. Subsequent appearances provided guidance for a 1,400 ft/min climb.

  Seconds later, Robert in the left seat was instructing Bonin, “watch your speed,” who then replied “Okay, Okay, I’m going back down,” but he did not.

  Robert persisted, “According to that we’re going up. According to all three you’re going up, so go back down.”

  Experience tells me that when the flight director is on, it provides a powerful cue, and is easily looked upon as providing the right thing to do. But the flight directors, which do not give stall recovery guidance, continued to provide pitch up commands while the stall warning sounded continuously and the descent rate increased.

  Their solution was to call for the captain.

  The stall warning had gone off repeatedly and Bonin applied TOGA thrust. There was no increase in the actual amount of power. One minute after the autopilot disconnected, the stall warning sounded continuously. Bonin said, “I’m in TOGA, eh.” It is apparent that he was expecting to power the airplane out of the stall situation like it can at 5,000 feet. Selecting TOGA at that high of an altitude provides only minimal additional power above the cruise setting, and no more than the climb setting that the thrust levers had been at. It is as though he had forgotten what a stall is and learned the wrong lesson on what causes the recovery to work. A stall is most effectively recovered with a pitch reduction to quickly reduce the angle of attack. Power, especially the minimal increase available at that altitude is not going to make any significant difference.

  Attempting to level off as the altitude approached 38,000 feet only resulted in the continuing loss of airspeed. In that condition it is aerodynamically impossible to increase or even maintain airspeed with the available thrust. The decaying airspeed, and lack of nose down command sufficient to reduce the angle of attack, only resulted in a rapidly increasing angle of attack as the stall deepened. Soon after, the attempt to maintain altitude degraded into a rapidly increasing descent despite a somewhat constant nose up attitude. Reducing the thrust, caused the nose to pitch down. The situation was then far beyond any training ever practiced in the simulator, or even imagined.

  First Officer Robert knew what Bonin was doing was wrong. He could see that Bonin was having a hard time determining this for himself. Why did he not he take over at this point? He should not have allowed Bonin’s incorrect inputs to continue when it was obvious to him that the inputs were critically wrong and he could have done a better job. It was not until they passed 10,000 feet that the dire nature of the situation became all too apparent.

  In response to Robert’s command to climb, Bonin admitted he had been pulling back for some time. Only then did it reach a point where Robert felt he must take over as Bonin had been doing the exact wrong thing for a long time without realizing it.

  It is apparent that they did not trust the instruments. Robert switched the data source for not only the air data, but for the attitude and heading as well. This was not an annunciated step on the ECAM. There was no mention of an attitude failure. He just said, “I’ll put you in A-T-T,” changing the attitude and heading data source to another inertial reference unit. At this point he had given up on the ECAM procedure and was trying to think of things that would work. He later asked the captain to try to reset flight control computers. That was never a solution for Alternate Law, and it was equally ineffective in this case.

  When the captain returned they did not provide a chain of events, they only said that they didn’t know what was happening and were losing control of the airplane. Robert stated they didn’t understand anything and had tried everything.

  As the stall deepened the airplane began a very rapid descent. A descent many times faster than they have likely ever seen before. They must have thought, “How can that possibly be?” and the altimeter (if it was working) must be wrong.

  At 02:11:58 Bonin stated, “I have a problem, it’s that I don’t have vertical speed information.” The vertical speed at the time was in excess of 15,000 feet per minute. The vertical speed indicator has a maximum display range of 6,000 feet per minute. If it was operating, the vertical speed pointer would have been pegged at the bottom of the display and changed from green to amber due to the excessive value. It may have also disappeared altogether, as will happen when the indication is in excess of 20,000 feet per minute. A value the designers apparently assumed was too high to be encountered.

  Even with the captain present, the three pilots had a hard time determining if they were in fact descending. At 02:12:27 as they descended through 18,000 feet at 15,000 feet minute, the flight recorder tracings show the indicated/recorded vertical speed to be erratic. Robert made contradictory calls: “You’re climbing” then “You’re going down down down.” Bonin asked, “Am I going down now?” Then the captain said, “You’re climbing.” They were not.

  It is possible the “climbing” and “going down” calls refer to pitch attitude and not altitude, or the erratic vertical speed, but it seems clear that the crew could not get a grasp on what the airplane was actually doing and why it would not respond to inputs.

  Bonin asked, “What are we here? On alti, what do we have here?”

  The captain declared, “..it’s impossible.” What did he see that contradicted what he thinks the airplane was doing that it would be impossible to be true? Could it be that he could not reconcile the nose pointing up, with full power applied, and an altimeter that was showing an almost unimaginable descent rate? He was also not made aware of the 3,000 foot climb, the prolonged stall warnings, and the 3,000 foot fully-stalled descent that had occurred prior to his return.

  For some time they were battling control of the roll, trying to get the wings level, often with little success, despite full control inputs.

  There were also the apparent contradictory indications from the stall warning. Once the angle of attack reached 45° as they descended through 34,000 feet, each time the nose pitched down and the angle of attack reduced, even slightly, it allowed the pitot-static system to measure more airspeed, bringing the stall warning back on.

  Could it possibly have seemed better to them when the nose was higher and the airspeed indication low or blanked out, because the stall warning was silent? Did they somehow believe that the stall warning was erroneous? Other pilots who encountered loss of airspeed indications, and had transient stall warnings, reported that they thought the warnings were false, because they had not strayed too radically from level flight. For them, the stall warning was incongruous with the stable pitch attitude and power settings they had flown. The stall warning margin is narrow at that speed and altitude and is not difficult to trigger due to turbulence or minor pitch inputs.

  They may have been so overwhelmed by trying to keep the airplane upright, by the other alarms, by trying to communicate with each other, with calls from the cabin, and calls for the captain, that the stall warning was tuned out altogether.

  In the AF447 transcript,
no explicit mention of the stall warning was ever made by the crew. However, the selection of TOGA thrust followed the stall warning by five seconds and remained there while the stall warning continued for another 49 seconds. Two seconds after the stall warning stopped the thrust levers were retarded to idle. During that time Robert exclaimed, “But we’ve still got the engines, what’s happening?” The thrust levers were returned to TOGA 46 seconds later after a subsequent series of intermittent stall warnings, and Bonin remarked “Okay, we’re in TOGA.”

  To me it paints a picture of an initial loss of situational awareness that they were not able to recover from until a very low altitude. Their loss of awareness of how they drained the airplane of energy in a 3,000 foot climb, and the apparent contradiction of a 15,000 ft/min descent with the nose pointed up with full power, made the indications of that descent incomprehensible until it was too late to do anything about.

  Their actions indicate that they thought that all that was required for stall recovery was the application of full power, and when that alone was not working, they did not know what else to do.

  At the heart of the matter seems to be the crews inability to comprehend the situation. Could fatigue have been a factor in the crew’s inability to analyze and correct the situation?

  Even though Bonin deferred to comments and suggestions Robert had made (the presence of ozone, St.Elmo’s fire, the slight diversion for weather), Bonin was the pilot in command. Robert would have had to reach a given threshold of intolerance before simply offering tips and observations to the pilot flying would turn into taking command of the airplane. By the time the captain arrived back in the cockpit Robert seemed to be a confused as anyone when he said “But we’ve got the engines, what’s happening? Do you understand what’s happening or not?” Only when Bonin stated “I don’t have control of the airplane anymore now, I don’t have control of the airplane at all,” did Robert respond with “Controls to the left.” This could have meant that he was taking over, but at the time the airplane was rolling past 30° to the right. Bonin had already had is sidestick full left, and Robert made left roll inputs himself for a few seconds. Bonin did not relax his full back and left inputs at all, and the airplane’s bank increased towards 40° to the right.