Quoting rfields5421 (Reply 24): A more complete CVR transcript is available on the BEA page linked above - with many comments in different English words than the interim report version.

As you say, the transcript in this final report is much more comprehensive than the one they released in the Interim Report. Previously there were gaps in the transcription over 3 minutes in length. This appears to be a nearly complete version of the CVR transcript.

As has been discussed here in the past, BEA has done a very thorough job of investigating the root causes of this event. For example, shortly after the initial disappearance of the aircraft and failure to find the crash site, BEA put together an international working group to aid in finding lost aircraft in areas of the world without radar coverage. That included an analysis of triggered transmittal of CVR/FDR data in the event of an emergency.

That work found its way into the recommendations in this final report on the crash.

I haven't had a chance to fully read the new report, but I did notice it made some further recommendations related to SAR operations and ATC that weren't in the Interim Report.

edit: I'll be interested to get Kaiarahi's input on part of the report that discusses the Human Factors Working Group (1.16.8) and if Mandala can glean any more information from the more complete transcript included in the Appendix to this report for his spreadsheet.

[Edited 2012-07-05 10:41:46]

Quoting rfields5421 (Reply 25): Quoting AirlineCritic (Reply 21): I find it amazing that other crews (not just AF447) often did not identify the unreliable speed situation. And more interestingly, apparently none of the crews performed the unreliable airspeed procedure. The passage you quoted emphasises the need for better, more focused training.

With the 'zero tolerance' approach to failure in the industry, surely the replication of the similar reactions to similar situation suggests a wider problem that needs to be addressed? The correct changes will happen as a result of the AF447 tragedy but I personally find it intriguing that such a disastrous chain of events occurred after decades of industry improvements in control systems and training. Even before AF447, AA587 demonstrated how incorrect training can lead to a horrific crash and I am wondering out loud whether an industry wide audit and indeed reassessment of training method needs to be undertaken to find potentially catastrophic problems that are not apparent in day-today operations.

Example of the CRM and decision procedures in the crew (Section 2.1.1.2):

Quote: In fact, the risk associated with the crossing of the ITCZ was discussed several times by the crew. In particular, from 1 h 45 to 2 h 00, the Captain and the PF noticed that they were entering the cloud layer and discussed the strategy to adopt. To avoid flying in the cloud layer while crossing the ITCZ and therefore to limit flight in the turbulent conditions that he mentioned several times(20), the PF wanted to change flight level and fly above the cloud cover, while recognising that it was not possible for the moment to climb two levels. He made several allusions or suggestions on the flight levels and the temperature from 1 h 35 min 20 onwards. He even considered requesting a non-standard level 360. His various interventions in the minutes that preceded the autopilot disconnection showed a real preoccupation, beyond the simple awareness of an operational risk. Some anxiety was noticeable in his insistence. The Captain appeared very unresponsive to the concerns expressed by the PF about the ITCZ. He did not respond to his worry by making a firm, clear decision, by applying a strategy, or giving instructions or a recommendation for action to continue the flight. He favoured waiting and responding to any turbulence noticed. He vaguely rejected the PF’s suggestion to climb, by mentioning that if “we don’t get out of it at three six, it might be bad”.

and choice of relief pilot (Section 2.1.1.3.2):

Quote: The investigation was not able to determine if the Captain had clearly defined the roles between the two co-pilots during flight preparation and in anticipation of his absence during his in-flight rest time. He did however implicitly designate as relief pilot the co-pilot in the right seat and PF, but did so in the absence of the second co-pilot, just before waking him. If this distribution of roles probably contained no ambiguity for the persons concerned, being in line with the principle in the Operations Manual (co-pilot as relief Captain and PF on the right), it was not however free of difficulty.

[Edited 2012-07-05 11:56:36]

This part is a little interesting. We've talked about how the crew ignored (or seemed to ignore because they didn't comment on it) the stall warning(s) that went on for some time. The report has a section (1.16.8.3) on that:

"Numerous studies have been conducted on insensitivity to aural warnings and they showed that the aggressive nature, rarity and unreliability of these warnings may lead operators to ignore these signals [1, 2]. In particular, in the event of a heavy workload, insensitivity to aural warnings may be caused by a conflict between these warnings and the cognitive tasks in progress. The ability to turn one’s attention to this information is very wasteful as this requires the use of cognitive resources already engaged on the current task. The performance of one of these tasks (solving the problem or taking the warning into account) or of both would be affected."

Additionally, they compare the response of the AF 447 crew to other crews that had similar conditions (1.16.8.4).

"A comparative analysis of reports and statements by other crews based on seventeen events that occurred in similar conditions to those of AF447, two of which are studies in 1.16.2, brought to light the following trends:

1) Analysis of the situation by crews appears difficult
2) Calling on the « unreliable airspeed » procedure was rare
3) Some crews mentioned the difficulty of choosing a procedure bearing in mind the situation (numerous warnings)
4) Others did not see the usefulness of applying this procedure given that in the absence of doubt about the unreliability of the airspeeds, their interpretation of the title of the “unreliable airspeed “ procedure did not lead them to apply it
5) Some gave priority to controlling the pitch attitude and thrust before doing anything else, and
6) The triggering of the STALL warning was noticed. It was surprising and many crews tended to consider it as inconsistent."

#2 is interesting because that is a memory item. #6 may be relevant because it appeared the AF 447 crew ignored the "STALL" warning. So, to some extent there is a common denominator with many of the other events. Perhaps the initial climb, initiated by the PF was the one factor that made this one so much different than the other events.

Quoting DrColenzo (Reply 28): surely the replication of the similar reactions to similar situation suggests a wider problem that needs to be addressed?

The initial BEA report was 'shocking' to the industry - in that about 3 dozen similar events were identified quickly. Such super-cooled icing at altitude was thought to be incredibly rare, hardly worth wasting training time upon the possibility - we learned it is much more common.

1) The review of past possible similar incidents was not comprehensive. Several airlines which fly the A330 / A340 refused to provide data to the BEA of possible incidents. So we know 3 dozen is the low threshold. There are very likely many more which were not included in the analysis.

2) Only Air Caraibes pushed Airbus about the way the crew and aircraft responded to UAS incidents at cruise - i.e. checklist, warnings, etc. And that airline had increased focus on air crew training. Other airlines only began to change some of their training after AF447 and the various interim reports. I have to wonder if anyone but folks deeply involved in avaition safety would have known of the Northwest Airlines incident had AF447 not recently occured.

3) While much has been made of the Thales probe and that the aircraft had not received the newer Goodrich probes - a couple points need to be remembered. Even the Goodrich probe has been shown to be vulnerable to such icing. Also - the really big problem with the Thales probes was in the A320 aircraft series - where there were many, many more icing failures - because those aircraft fly in known icing much more often than the A330/A340 family. That is due to their higher number of cycles on average, and more ascents/ descents through bad weather. Air France was agressive in replacing the probes on that series of aircraft, with the A330/A340 family next in line.

Quoting DrColenzo (Reply 28): I am wondering out loud whether an industry wide audit and indeed reassessment of training method needs to be undertaken to find potentially catastrophic problems that are not apparent in day-today operations.

To me the report - what I've read so far - points to need for a fundamental change in the thoughts behind simulator and flight training. Basically, much more time in training.

Which I fear will not occur. Some training sessions will be modified to include this type incident - but then other important training will be deemphasized, or even skipped completely.

Because the bottom line is that implementing such recommendations will result in noticably higher costs for the airline. And the airlines are going to do everything they can to not only keep from increasing training costs, but decrease such costs.

Quoting canoecarrier (Reply 30): #2 is interesting because that is a memory item.

Maybe the reality is that memory items are becoming like high school students studying for standardized tests. People learn to memorize the answers, but don't learn how to apply the answers to real world situations.

[Edited 2012-07-05 13:02:15]

A long awaiterd document anrd really worth the wait .
For those who don't understand french doc presentation, the final report in English is Here
and the list of appendices is also entry for the respective document : Click on the check mark in the column of the language you'd favour.
This is also long reading and , although we studied this accident at length, there are still some surprises :
- the similarities between AF447 and the other incidents is quite enlightening. A lot more will come from this part of the document as it is too important in terms of ergonomics and CRM, not even talking about SOPs.
- About the crew : the LHS was found in the "stored position. It can only mean that PNF had been in the process of giving up his seat to the captain... How long had the seat been in that position ( it prevents the pilot to have any meaningful input to the flight controls) ?.

I need to read the whole report before making any meaningful input.

Quote: For the cases studied, the recorded flight parameters and the accounts given by the crews did not reveal any application of the memory items from the unreliable airspeed procedure, nor the procedure itself:

Oh how I wished I did not have to learn the memory items for my flight training in the PA28. They were annoying, consumed valuable study time and I never used them anyways. Here's the reason why I still learned them... because they are of vital importance.
If the pilots are unable to recall memory items, then how can you not put (part of) the blame on them? When undergoing pilot training, the company must ensure that pilots are (1) capable of coping with abnormal amounts of stress and (2) that pilots are eager to train for and maintain a set value of internal standards and procedures. I can't see how AF cannot be held responsible for this accident, or at least the way the pilot reacted to a system malfunction.

Quote: The investigation was not able to determine if the Captain had clearly defined the roles between the two co-pilots during flight preparation and in anticipation of his absence during his in-flight rest time.

"Laissez-faire". It's just as destructive in aviation as an autocratic flight deck. Does AF not use the concept of Captain / Senior First Officer / First Officer on long haul? I find it hard to see where the roles and ranks of those three are disputable if such a regulation is strictly enforced by the company!


I am very dissapointed in the way AF treats this issue after 228 passenger have died, largely due to the misjudgement of the crew. If only it were the confusing indications in the cockpit or the abnormal leadership style, but the fact stays that they flew right into a massive thunderstorm, which by all means every pilot knows to avoid.

Quoting Klaus (Reply 20): They never had control of the situation right from the get-go: Failing to even execute the unreliable airspeed checklist leaves no doubt about it.

I think Klaus has a good point here. More than automation the point is the lack of training, in many airlines this maneouvre is performed only until the memory items are done, so the rest of the checklist is not finished.

Hopefully after this it might become a mandatory sim maneouvre

Regards,
Daniel

In section 2.1.2.3 the reports notes:


"It would also seem unlikely that the PNF could have determined the PF’s flight path
stabilisation targets. It is worth noting that the inputs applied to a sidestick by one
pilot cannot be observed easily by the other one and that the conditions of a night
flight in IMC make it more difficult to monitor aeroplane attitudes (pitch attitude
in particular)."

I pointed this out in earlier threads and was told that the other pilots inputs were
"easily" observed.....I guess not. This is particularly significant, especially as though
it appears that the 330 is neutrally stable in Alt 2B law. I wonder if the pilots were
trained for and understood this. I didn't see...or missed...that. I would think knowing
what the other pilot is doing with his joystick would be a great asset....I guess not.

That idea of having tracking joysticks is entirely independent of what the airplane is
doing....even sitting on the ground. One moves...the other tracks....exactly like yokes.
I can't envision any reason...other then cost why this would not be desirable.

Having the way the aircraft responds to joystick inputs, especially suddenly and
while the plane is flying in the coffin corner seems very undesirable to me, but maybe
unavoidable with the loss of airspeed inputs.

Altogether, I thought the report was extremely well researched, documented, and
presented.

One thing I noticed, and correct me if I'm wrong here ... but reading through the CVR transcript which included all the sounds, it was already pointed out that the stall warning sounded unacknowledged some 70+ times.

Another one that caught my attention that started sounding in between the crickets and stall warnings was one that was noted on the CVR as "SV [synthetic voice]: dual input". Which is where the problem of establishing PIC admist the disaster, and that the pilot on the right (as it's hard to determine PF/PNF in my opinion when R/H pilot gives up controls and then takes them over again without saying) ...

Anyhow, how does the 'dual input' alarm relate to this incident, and the all too late realization by the rest of the crew when the R/H pilot stated he'd been pulling back on the stick the whole time?

Though with the mentions in the report about ignoring the stall warning, the one about dual input (assuming it means what it sounds like) might have just blended in with the rest of the noise and confusion in the cockpit, and thus gone unnoticed until the pilot mentioned the fact aloud that he'd been trying to climb the aircraft...

Airbus has flown test flights to determine buffeting levels in configurations corresponding to AF 447 (Section 1.16.5):

Quote: Additional analyses were conducted with Airbus to determine if this phenomenon could correspond to buffet. The identification of this phenomenon is complicated by the fact that the concept of buffet is defined as accelerations at the level of the pilots’ seats and not at the centre of gravity. Airbus subsequently flew special flights to collect more accurate data at high angles of attack and with an aircraft configuration close to that of the accident (mass, flight level, Mach, etc.). These tests made it possible to refine the preliminary correlations and to establish that the level of buffet was considered to be a deterrent by the test pilots when the angle of attack was about 10°, corresponding to normal acceleration amplitude of 1 g at the pilot’s seat. This angle of attack was reached at about 2 h 10 min 57 s during the accident flight. Thus, the stall warning was triggered at 2 h 10 min 51 at an angle of attack corresponding to the theoretical threshold for the measured Mach value. Two seconds later, vibrations that might correspond to buffet appeared. The intensity of vibration probably reached the deterrent buffet level at about 2 h 10 min 57 s.

Apparently, ECAM was not very helpful to the crew (Section 2.1.2.4):

Quote: Once the first actions in response to the perceived anomaly is executed (returning to manual piloting following AP disconnection) and the flight path stabilisation ensured, the philosophy of both the manufacturer and the operator is for the crew to look for additional information necessary to understand the problem and take action. Three seconds after the autopilot disconnection, the ECAM displays no information that is likely to point to a speed indication problem The ECAM mentions a maximum speed that should not be exceeded but does not mention a minimum speed. This could lead crews to suppose that the main risk is overspeed. In the absence of any reliable speed indication, this might lead to a protective nose-up input that is more or less instinctive. ... Thus, having identified the loss of airspeed information, the PNF turned his attention to the ECAM, undoubtedly in an attempt to refine his diagnosis and to monitor any actions displayed. He started to read the messages, and consequently called out the loss of autothrust and the reconfiguration to alternate law. The successive display of different messages probably added to the confusion experienced by the crew in its analysis and management. ... The symptoms perceived may therefore have been considered by the crew as anomalies to add to the anomaly of the airspeed indication, and thus indicative of a much more complex overall problem than simply the loss of airspeed information.

How the crew may have missed the stall warning (Section 2.1.3.3):

Quote: The crew never referred either to the stall warning or the buffet that they had likely felt. This prompts the question of whether the two co-pilots were aware that the aeroplane was in a stall situation. In fact the situation, with a high workload and multiple visual prompts, corresponds to a threshold in terms of being able to take into account an unusual aural warning. In an aural environment that was already saturated by the C-chord warning, the possibility that the crew did not identify the stall warning cannot be ruled out.

Having read this report, one glaring issue seems to point to a lack of handling experience, both in non-normal operations and manual flight. It states that the PNF (the senior copilot) had last performed a manual landing over two months before the accident flight - surely this is unacceptable?? More pressingly, I believe the whole issue of simulator based training needs completely tearing down and starting again from square one. The whole V1 cut/single engine go around/Non precision instrument approach refresher seems just so predictable - and doesn't really reflect real world flying. What pilots should get is a wide variety of failures, encompassing different systems at all stages of flight - but without any briefing of any sort. Surely this would help them to prepare mentally for the shock and surprise of situations such as this - which seemed to almost be the last hole in the cheese.  

Anyway, RIP to all on board, especially the crew who clearly tried as hard as their situational awareness would allow to save their aircraft, right until the very end.

Quoting canoecarrier (Reply 30): #6 may be relevant because it appeared the AF 447 crew ignored the "STALL" warning.

If the crew believed the initial stall warning was faulty (and they certainly had reason to - we know it was driven by corrupted input from the air data sensors), then it's quite likely the subsequent stall warnings were ignored, either consciously or unconsciously. What's curious, though, is that the captain, who wasn't around when the problems began, wouldn't have reason to believe that the warning was faulty, yet he didn't ask about it. Perhaps the actions of the other two, who didn't appear to be focusing on it, led him to the incorrect conclusion that it was faulty and should be ignored.

-Mir

.....an additional thought.

Because the report says that recovery was unlikely once they fell back through 35,000 feet, I'm not sure
they had time between the initial autopilot disconnect and falling through 35,000 to read and use any
check list. That's why we have the "memory" check list to stabilize the situation. This is where they
failed for whatever reason...training, misreading instrumentation, etc.

[Edited 2012-07-05 14:35:17]

Quoting airtechy (Reply 35): That idea of having tracking joysticks is entirely independent of what the airplane is doing....even sitting on the ground. One moves...the other tracks....exactly like yokes. I can't envision any reason...other then cost why this would not be desirable.

System complexity would go way up and reliability would take a significant hit (right down to canceled flights on malfunctions), which doesn't seem like a very good tradeoff when it would have benefits only in the rare and already near-catastrophic situation of a total CRM breakdown as in the case of AF447.

The way Airbus FBW works it would not have any flight-control-related benefits, because the constant spring is already a correct reflection of what effect a movement of the stick will have in normal law (and outside of normal law it sounds even more dubious to engage a backdrive system when parts of the systems are already unreliable).

And whether it would have had any benefit for AF447 is questionable, since neither of the pilots seems to have been aware that they were in a stall all along.

Quoting Mir (Reply 39): If the crew believed the initial stall warning was faulty (and they certainly had reason to - we know it was driven by corrupted input from the air data sensors)

I may mis-remember, but I don't think that is true. The unreliable airspeed indication lasted only for a few tens of seconds and became stable again afterwards. The stall warning depends on it, too, so as far as I'm aware it only sounded when the airspeed indication was reliable again. And was later interrupted only when the PF pulled the aircraft so deep into the stall that airspeed dropped too low to be measured any more, without fault of the sensors at that point.

[Edited 2012-07-05 14:45:42]

Quoting AirlineCritic (Reply 21): I find it amazing that other crews (not just AF447) often did not identify the unreliable speed situation. And more interestingly, apparently none of the crews performed the unreliable airspeed procedure.

If your in the cruise, all you need to do initially is to turn the flight directors off and manually maintain the pitch attitude and thrust setting in response to unreliable airspeed. There is no rush to dive into any checklist or procedure, this is for any aircraft type.

Most cases of unreliable airspeed these days are picked up by the system automation well before the crew does, and self resolve in under 10 seconds. In the old days I suspect they just were never picked up, as those aircraft flew pitch attitude and thrust setting, they just self resolved.

Over reaction to many events in aviation can cause more problem than the original event.

Quoting Klaus (Reply 41): I may mis-remember, but I don't think that is true. The unreliable airspeed indication lasted only for a few tens of seconds and became stable again afterwards. The stall warning depends on it, too, so as far as I'm aware it only sounded when the airspeed indication was reliable again.

But it did sound, in conjunction with the loss of air data indications. That's enough to link the two in the back of your mind, and if you don't realize that you've got valid air data afterward (and I don't believe they did), it's very possible to assume that any stall warnings are invalid as well.

-Mir

Quoting zeke (Reply 42): If your in the cruise, all you need to do initially is to turn the flight directors off and manually maintain the pitch attitude and thrust setting in response to unreliable airspeed. There is no rush to dive into any checklist or procedure, this is for any aircraft type. ... Over reaction to many events in aviation can cause more problem than the original event.

And that makes a lot of sense. Thanks, Zeke. I also think that is the general idea with the unreliable speed procedure:



Paraphrasing a bit, it essentially tells you to turn off the FD/AP and hand-fly the aircraft in the sensible way. Which should come naturally for pilots... I had hoped, at least. You should be aware of your altitude, pitch, power settings... even if the FD/AP takes care of it under normal circumstances. Not that I have to tell you this, obviously. Just making a point that it matches what you said is the natural reaction.

I wonder when the report said that the other crews had not followed the unreliable speed procedure, did they meant that the crews had not explicitly mentioned this procedure, or that they did something inappropriate for the situation? Based on the text, it seems that some crews at least neglected to turn off the PD. But maybe it wasn't necessary in their situations, if the aircraft was still flying the right way. Those crews may have had the situational awareness that AF 447 crew lacked.

[Edited 2012-07-05 15:15:28]

Quoting AirlineCritic (Reply 44):

Read the first line, "if the safe conduct of the flight is impacted", that is not the cruise. You are already level, above MSA.

This is the problem when members of the general public get access to such information, they do not how to use it. You do not do the boxed memory items in such a situation.

Quoting zeke (Reply 45): Read the first line, "if the safe conduct of the flight is impacted", that is not the cruise. You are already level, above MSA.

Ah, yes. Thanks. Question: Does this also cover not necessarily turning off PD/AP, if the crew can observe that they are not immediately causing a problem?

Quoting AirlineCritic (Reply 46):

The checklist is 5 pages long, it is broken up in different phases of flight. The initial part of the checklist which you posted is designed for the initial phase of flight, where it talks about thrust reduction altitude, that is typically 1500' above the runway. It is covering the possibility for example a bird strike that hits the pitot probes. It tells you what thrust setting and pitch attitude to fly to get above terrain to level off and sort it out, we typically fly 15 degrees NU or greater after liftoff, so the attitude is not uncommon for that phase of flight. It is a memory item, as you are too close to the ground to be distracted with a qrh checklists.

Below that entry point, they have another, and in that section you also turn the autopilot and flight directors off, as well as autothrust. It gives you tables of thrust and pitch setting in order to level off. If you are in the cruise, one already knows the pitch and thrust setting, all one needs to do is maintain it.

The a.net discussions have often looked at the role of the stall warnings that were no longer continuous after some time had passed and the aircraft angle of attack was too great. The report states something interesting about this (Section 2.3.3):

Quote: A few seconds after the transition to alternate law, the stall warning sounded briefly, even though the PF’s inputs should have made this warning sound for several seconds. The reason for this is the drop in the measured airspeeds, some of which fell temporarily to below 60 kt, while the angle of attack reached 40°. Furthermore, the drop in measured airspeeds to values of less than 60 kt during the stall caused the repeated activation and deactivation of the warning which may have made it considerably more difficult for the Captain to effectively analyse the situation on his return to the cockpit. However, it was doubtless already too late, given the aeroplane’s conditions at that time, to recover control of it.

I have a quick question. The report says that for low altitude stall recovery the suggested method is 12.5 deg and TOGA throttle. It goes on to say that the PF might have remembered this procedure and used it (the plane was near 12 deg pitch up for most of the time and TOGA for some of it). My question is - how is low level stall recovery 12 deg nose up? I thought stall recovery is always descend or be level.