There was indication to the pilots of unreliable airspeed information.

deSitter,

That's a pretty harsh indictment of a lot of good people. People are not 100% perfect at flying planes or designing and writing software. Most pilots and software engineers are very committed to continuous improvement in our respective professions. In the case of AF447, great care needs to be given to the point of intersection of these two disciplines precisely because, as you point out, errors or omissions (whether piloting or programming) can lead to people dying.

Back to basics
 

Just to try to sort out this man-machine-interface stuff, can we imagine the very same initial event in a more conventional airplane, say, a B757? In cruise at night, George is in control, and all of a sudden all airspeeds go wild and George gives up. What indications are available on the panel, and what is the PF supposed to do?

grizzled, thanks well said.

Again, I think we're still missing too large pieces of the puzzle to tell what really happened. However, from the precious few bits of information that we do have, what makes you come to that conclusion?

Attitude + Power = Performance

It is basic stuff that sometimes tends to get obscured in all the magic. It is also the foundation under which the "Unreliable Speed Indication" drill and checklist were written.

Bienville, you are correct. I am not a pilot. I simply read.

"From 2 h 10 min 05 , the autopilot then auto-thrust disengaged and the PF said "I have the controls". The airplane began to roll to the right and the PF made a left nose-up input. The stall warning sounded twice in a row. The recorded parameters show a sharp fall from about 275 kt to 60 kt in the speed displayed on the left primary flight display (PFD), then a few moments later in the speed displayed on the integrated standby instrument system (ISIS)."

I read that as the pitots froze up and the stall warning happened. I am not reading the "left nose-up" as being a violent movement such as would stall the plane. I am also reading this as the aircraft presuming 60kts is an obvious stall so let's warn the pilot he's about to stall. (There being no real unique stall warning.) That's the moment the fit hit the shan.

Now, what sequence do you see happening there? It is worded a little ambiguously. But it does seem to say the stall warning was coincident with the abrupt reduction in indicated air speed. And I take it as dead wrong. You cannot stop an aircraft that rapidly. (I read it as "seconds or less" not "tens of seconds" for the "sharp fall" from 275 to 60.

I certainly do not read a left-up control stick movement such as to the left around the storm (context, remember) as being the precipitating event to a real stall with no drop in air speed followed measurable time later by the "sharp fall".

If I am wrong, please clarify the quoted paragraph as you read it.

Period between 0200 and 0600 is called "window of circadian low" for a few very good reasons.

Relevant time zone was UTC+2.

I switched onto the big bus just some weeks ago, so I decided to shut up until now... but how some people here speculate on things like the sidestick commands reversing or protections going crazy, I just want to add one thing: Ockhams knife - the most simple solution is the right solution - for some reason somebody just screwed up.

Probable Cause:
Unreliable airspeed leading to the PF reacting to a stall warning with a prologend nose-up command several times, aggrevating the stall.

Contributing Factors:
(1) Lack of training
(2) Man-Machine Interface
(3) Sidesticks not linked -> insufficient monitoring
(4) Icing of the pitots
(5) Fatigue
---- or ----
(6) Somebody beeing plain stupid (reacting to a stall warning with a pitch up is plain stupid) and shouldn't have been let near a cockpit

Just my two cents....

Bienville, to address the response to the stall I very drily note you are correct. The correct response to a real stall is drop the nose (any way you can) and gain speed. Why did the Continental pilot die with a recorded 150# pull on the stick as his plane stalled into the ground? Why have many other pilots responded to the stall warning with a pull up and goose the engines move? Is this because the stall warning is taught as "the plane cannot stall, this is a warning it might (contradictory here but a paraphrase of what I've read here) stall so pull up and push throttles forward some to compensate."

As I read messages here, on this thread, there is no real "you are now stalled" warning for the A330-200.

So as I see it, based on messages here, the pilots reacted to training and pulled up. A slight bit should have sufficed. It didn't work. So from there it turned to poo rapidly.

And if I had to involve a pilot error in the picture that is the moment. They were not thinking when the stall warning happened or THEY would have seen it was spurious. (IMAO the computer should have seen it was spurious.)

edit: (I must say I am impressed by your persistence in your claim that the first stall warning indicated a real stall. It stalled from around 38500' when the airspeed really did get too low for that altitude. The first warning was, as I indicated, purely spurious. If not howinheck did the plane slow down fast enough to enter a real stall? No indication of such a deceleration is made.)

edit2: I note that the 2:10:16 interval notes the pilot flying made a nose down stick movement. So the initial reaction was correct.

1. Check the pitch attitude and thrust.
2. If pitch attitude or thrust is not normal for phase of flight:
Autopilot disengage switch . . . . . . . . . . .Push
Autothrottle disconnect switch . . . . .. . . .Push
F/D switches (both) . . . . . . . . . . . . . . . OFF
Establish normal pitch attitude and thrust setting for phase of flight.

First stall warning was false
 

JD-EE wrote:You are absolutely right. The first stall warning was false. The flight dynamics were fine (though I wonder why the computer caused the roll that had to be corrected).

If there was a first false stall warning, would it cause the pilots to question subsequent stall warnings?

JD-EE is right on. There was no stall initially, just a false stall warning due to a completely impossible "loss" of airspeed.

The stall was created by the PFs reaction to what was not a stall. The computer made it unrecoverable by trimming full nose up.

Hi bear,
0210:05...The airplane began to roll to the right and the PF made a left nose-up input.
0210:51...The thrust levers were positioned in the TO/GA detent and the PF maintained nose-up inputs...The PF continued to make nose-up inputs. The airplane’s altitude reached its maximum of about 38,000 ft...
0211:40...The PF made an input on the sidestick to the left and nose-up stops, which lasted about 30 seconds.
[French: 0211:40...Le PF exerce une action sur le manche en butée à gauche et à cabrer, qui dure environ 30 secondes.]
"en butée à gauche et à cabrer"= full deflection, stick left and nose-up (during 30 seconds).

Real stick imputs chanels are recorded by the DFDR independently of the "computer" imputs. It is how one can see that both pilots are possibly giving "conflicting" imputs, whatever the result is on the "computer" and control surface orders.

triggering of stall warning
 

Just a question as I could'nt find any reliable information about it.

What sensors do exactly trigger the stall warning? To my knowledge stall warning depends primarily on AOA sensors (taking into account slats/flaps settings)

Or is any other data also processed by the stall warning (air speed?)?

Assuming that the stall warning primarily depends on AOA sensors, and given that you have no/confusing/unreliable/disagreeing air speed data, why not fully trust or prioritize "stall warning" over all other warnings like potential overspeed warnings?

I mean that stall warning would then be based on sensors, which you still could have faith in , whereas overspeed warning would be based on sensor, which obviously show problems. Problems, which brought you into that situation...so why trust any warning coming from them?

Or do I miss something here?


Yeah,sure, all that fancy laws and protections left the pilot no other choice but pulling on that stick until the bitter end. :hmm:.
I followed your earlier 'elaborations' about software engineering with some sympathy as I partly can support your 'findings' from my own experience. But given your last post here, you just seem to be another one trying to adjust reality to your own 'agenda' with a given result of the 'blame game' from the very start.

@lateott

No, it was genuine - The stall warning is based only on AoA which was valid (except later when the speed dropped below 60kts)

>> The airplane began to roll to the right and the PF made a left nose-up input. The stall warning sounded twice in a row. <<

With a nose up input at this flightlevel with a heavy A/C you are approaching the critical AoA really quick. And with the roll the load-factor increased leading to reaching the critical AoA even faster

engfireleft and Zorin_75, thanks for your answers:
Now, for those questioning the Airbus MMI, how is the information different on the 'bus panel? Attitude and engines N1 are available as well, so what is this nonsense about the A330 MMI having prevented the pilot(s) from correctly reacting to UAS condition?

Please excuse me for a moment while I stop choking on laughter.

There that's better. Now it seems to be that you made up the fact that the plane stalled at the time of the first stall warning without carefully reading the paragraph that defined it. You have a preconceived notion and are driving it into the ground with impressive arrogance.

And to me it looks like you "just make stuff up" to support your notion. You certainly are offering no evidence that the plane COULD have stalled at the time you indicate. What slowed it down far enough? (I will admit that the English in the report could be clarified a little, particularly the paragraph titles. I suspect it's a stylized phrasing for their report purposes. The "From 2 h 10 min 05" indicates "data from" that time shows the following information.)

Svarin, I have no idea which airline trained you, but during my initial A330 training, I clearly recall using the manual pitch trim in direct law. This is to expose the pilot in manual pitch trim and allow us to get a sense of how sensitive the pitch trim really is (and it is bloody sensitive). We fly an approach in direct law with manual pitch trim only, as part of the initial type training / exposure to the airbus flight control law, although this is not part of the regulatory requirements, it is still good to see and very good exposure.

Also on many occasion during our simulator training,our simulator instructor had told us on many occasion that when we are in the coffin corner, and if we had a approach to stall situation, with TOGA thrust being auto apply at Alpha Floor, and with THS being full trim, we might not have enough elevator authority to counter the pitch up by TOGA thrust. Therefore a small amount of forward push of the THS might be required to help with the nose pitch down for the stall recovery, 99% of the time we do not need this, but we were told it is available and do use the THS if needed.

Just my two cents!

What's not clear from the report is why the PF made any stick inputs at all. My understanding is that, even in alternate law, the computers maintain the pitch and roll attitudes previously commanded until the stick is moved, so presumably the aircraft would have continued straight and level, assuming that was the case when the a/p disengaged.

"From 2 h 10 min 05, the autopilot then auto-thrust disengaged and the PF said "I have the controls". The airplane began to roll to the right and the PF made a left nose-up input."

Would the flight control system countered the roll without the PF inputting left nose-up? And what led him to think that he should pitch up at all? If (as reported) he was flying from the right seat, the airspeed he was seeing was not recorded on the FDR, so perhaps there's insufficient information to fully judge what he was responding to.

But it does appear, if my understanding of the Airbus FBW system (gleaned mostly from these threads) is correct, that the aircraft would have continued in straight and level (and unstalled) flight absent the left nose-up input by the PF.

Sorry for the silly questions: in ALT, the sidestick controls load factor demand, correct? Once the a/c is stalled, that load factor demand cannot be met. In fact, g would have been <1 at some point. What happens then if the sidestick is in its neutral position? In particular, would the autotrim be active?

Readers Digest Condensed Version - Rush to Pilot Error
 

The post volume is definitely up since Friday's release, but I don't know that we can say the same for the post quality. We were all hoping for "the answer" to be released on Friday, and I think a lot of people, especially some of the general media seem to feel it was, and that the answer was pilot error.

What we got from the BEA was basically a condensed version of what happened, and it leaves with a pilot taking what appears to be an unexplained action, 30s of pitch up.

Coming from a technical and software background (machinbird and others can probably tell from where) one might think I would be quick to jump on the pilot error bandwagon, but I am not.

If we look at this from the standpoint of Input --> Decision --> Output we have, I feel, only partial insight to what the inputs to the pilots were, we don't know, from what was given so far, what the decision making process was, and we have at least some of the outputs, the pitch, roll, throttle commands but not a real understanding as to why those choices were made when they were made.

Hopefully when a more complete release of the CVR and DFDR information is made we will gain greater insight as to the decision making process and "why" of the pilot actions. I find it hard to believe that if they appear counter intuitive that we are not missing parts of the picture either from the brief release, or limitations to what data is collected in the complete record.

These pilots had a very vested interest in solving the problem they were presented, I strongly suspect there was a lot more discussion and information than we are currently being given. Though the BEA may have been wanting to help stem the amount of leaked information, they may have added to the amount of misinterpretation, or rushed judgement.

I also hope the moderators will ensure that several of our new posters who share a common viewpoint do not originate from a single IP address.

Caygil asked: "Would it be out of question to create an automation graceful degradation, which:.."

There is a patent filed, since the accident, on just this idea.

The more I think about it the more I suspect that the plane could fly perfectly well using only GPS and inertial data if it's looking to maintain altitude, attitude, and speed. Abnormal conditions would be indicated by the precursor conditions to either stall or mach buffet. Actual stall would be indicated by altitude loss. So for modest periods of time the plane could probably do very well on such data as it has. I'm not sure retrofitting this "learning" software into the ancient (in computer years) computers aboard an A330. But it certainly could be incorporated into new designs.

Well, I would be careful with such conclusions. I asked that question before, but to my knowledge stall warning is based on AOA sensor, not air speed. And crusing at that altitude you dont need anything close to "stop an aircraft" to approach stall, you only need to lose a "few kts", which could well be the result of even only a moderate pull up maneuver.
So for me there is no reason to doubt that even the first stall warnings were correct.

:bored:...pull up? Excuse me. I have no idea where you get that from,,,,but you will hardly find anyone in the world proposing that as 'stall procedures', except those with suicidal tendencies...

If I am misreading some recent posts, I apologize, but there seems to be some confusion about stall warning and airspeed. The sudden drop in indicated airspeed was due to ice particles blocking the pitots, not due to a real speed reduction. Stall warning begins when the AoA exceeds approx. 4 degrees, and that threshold was exceeded at several times. AoA reached 16 degrees at 38000 ft, indicating that the airplane was definitely stalled at that point.

Stall warning, followed a few seconds later by uncommanded left/right rolling motions probably accompanied by heavy buffet is usually accepted as an indication of "you are now stalled".

Until a fault tolerant method is developed for obtaining airdata, I hope UAS drills become far more comprehensive in the sim. For three rated pilots to miss the recovery is shocking, assuming BEA isn't leaving out some important tidbits.

Oh, and patch the bugs in the stall warning system for low airspeeds and/or AoA... No doubt that design "feature" should be reclassified as a bug.

Lastly, the man machine interface really needs some review regarding sidestick, thrust lever, and trim positions. When you are 3 minutes from doom you should be given enough information as quickly and comprehendible as possible to take over from Otto when he can't figure it out.

Certainly one would think so.

In this case we see that Attitude+Power was not obtaining the Performance one would normally expect.

The crew seem not to have realised they were in the stall regime while all us of in our armchairs are at a loss to understand why they did not.

I have seen somebody in a steam gauge sim with white on black AH invert the ac at 16000' and fail to realise what he had done until after he "hit" the ground. Fixation can take over, and deafen you to the guy yelling "roll left" directly in your right ear.

Here it seems the entire crew were aware they were on the way down without knowing why or anybody offering any ideas on how to change the situation for the better.

garage years
 

many thanks for link on pitch and AOA. as a SLF myself i have followed the AF447 thread from the beginning on PPRUNE. may i ask what is TOGA and the ISIS display. thanks:ok:

"But it does appear, if my understanding of the Airbus FBW system (gleaned mostly from these threads) is correct, that the aircraft would have continued in straight and level (and unstalled) flight absent the left nose-up input by the PF."

Great question, I trust a knowledgeble poster will answer for us.

Thanks in advance

And that, in a nutshell, is what needs explaining.:sad:

OPENDOOR, I figure good sense disagrees with your rather broad statement. The inertial system can tell if the plane has actually decelerated or accelerated by over 200 kts or even 50kts and the mother of all tail winds overtook the plane. So there is a valid software check on the stall warning.

Furthermore, AI seems to think there is enough information in the inertial and GPS systems to keep the plane flying while the pitots recover. The plane has a tremendous amount of inertia. So one minute or less of simple pitch and altitude with comparison between inertial/GPS data on ground speed can diagnose whether they are in a large wind or not. If they are in a savage tail wind they will cover more ground than they should for the past ground speed. The difference is the speed of the tail wind. The same fits for any other strong wind. If you are not deviating from the previous course with no control changes then your airspeed cannot have changed.

It appears on a closer look that my naive question from a couple years ago has an even better answer than I'd thought at the time. It may give a bumpy ride for a couple minutes. But it will keep the plane going very nicely. It can tell you if you did hit a wind, what you need to do to correct for it. And the inertial system should be able to tell you quite rapidly what is going on.

However, the GPS alone is indeed useless. It's lag is too large. It can be cozened into giving you real velocity figures. The accuracy may not be as much as you like. (Three GPS antennas and differential phase tracking GPS can give you plane attitude, on large planes, fairly accurately, too.) GPS's main fault is that it tells you where you were not necessarily where you are. That's why you have Kalman filters and inertial navigation systems in the picture.

(And, yes, bienville, I do have a half a notion of what I am talking about with GPS.)

GY, maybe this answers some questions?

This is the important part: "The PF made nose-down control inputs and alternately left and right roll inputs."

The PF did indeed stick the nose down. It didn't work.

Inexplixable Happens
 

An acquaintance of mine was once making a routine test flight in a helicopter under development. In the brief, he had been reminded by engineering that high-inertia rotor blades were installed (not part of the test) and therefore not to let the rotor rpm get too high. During the flight, which involved shutting down an engine, he was presented with an unanticipated failure/condition with the rotor tachometer going to and freezing at an overspeed indication. He raised the collective to get the rpm down and fixated on the indicated rpm. He persisted in holding the collective up even though it wasn't producing a change in the indicated rpm and the rotor had slowed well beyond the point that he should have first recognized that it was below normal. (Apocryphally, to the point where he should have been able to see the individual blades instead of a disc...)

Presumably Airbus and Air France are putting line pilots into a simulator and exposing them to the same instrument indications and aural warnings to see what they do.

My understanding is that the side sticks are not connected and have limited range of motion so the pilot not flying and the pilot(s) observing have no idea what the control inputs of the pilot flying are. If the pilot flying were doing the inexplicable, they wouldn't be able to tell?

dear Foster 23
TOGA = Take-off Go-Around (thrust setting)
ISIS= integrated stand-by instrument system
:ok:

Sir John:
 

The 'apparent' actions of the crew are truly a mystery and until the BEA releases a more detailed report (sometime next year?), there is no way to divine the truth of what happened with the coarse data and spotty information provided so far.

takata, I add the PF made a nose down movement at 2:10:16. That's an important item and should not be ignored.

Post #237 (Svarin) and post #260 (mm43) are both recommended reading.

Re #237 : Svarin's speculation has a certain uncomfortable ring of truth. Protection mechanisms can be a doubled edged, and can often bite when one of those 'unknown unknowns' comes out of the woodwork. Post #4 in "Not good in an ETOPS/EROPS TWIN" is an excellent description of a similar conflict between man and machine.

http://www.pprune.org/tech-log/43214...rops-twin.html

ihg, there is a difference between "a stall is coming" and "you are stalled". When near the ground do you want to nose down BEFORE the stall?

They weren't near the ground. But were they trained for high altitude actual stalls?

HN39 - they hit a strong updraft?

JD-EE

The nose was never down! Have a look at post #260.

I notice a number of regular posters are quiet, which is to be expected.

BJ-ENG

Thanks for noticing!