http://img254.imageshack.us/img254/4839/beap.jpg

Your theory seems a little bit forced . As can be seen on the graph the selected VS value drops ceased to occur at the start of the climb, so it’s improbable to be related to post flight memory damages or data corruption.

If indeed are generated by improper value of the data used for graph, the source it’s highly probable to be a soft/process related to the flight.

So, what can be , and relevant or not for a/c behavior?

Hi Turbine D,

It seems to me that PF may have simply memorised 3 attitudes and power settings, and forgotten the "Below Thrust Reduction Alt" bit.
Once he had selected TOGA power, then he placed the aircraft at 15 degs pitch and waited.

Dohhh!!!

Hi,

A small question about the behavior of the aircraft (A330)
What is the difference at low and high speed for roll control (aileron effectiveness)
I suppose the roll control is more effectiveness ( more sensitive ) at high speed .. ?

infrequentflyer789

Thanks for a meticulous and thoughtful reply. I quote the initial BEA text when they sought to describe PF's actions at a/p drop. One can fall back on the dual language, made an issue here, but I prefer to place more confidence in the text as translated, to avoid weasel wriggling.

Some thoughts: The PF's actions were initially "Mostly Nose Down, with alternate roll left and right." Nose down is allowed when the bus is intitially dealing with Overspeed, but gradually ND becomes unavailable, and instead, it is replaced by auto NU, adding to Pilot input, if any.

The THS is also unavailable in Overspeed mode, so that explains the "loss" of THS TRIMOUT after the initial .7 degree TRIMUP. So we are left with a climb of robust nature, and little ND, until just at STALL, where the THS, reinvigorated, cycles all the way NU. Is this from "Memory"?

What does the Bus do if it is in Overspeed after identifying and reacting to it, but then speeds are lost?

Keep climbing? What is the nature of the annunciation of Overspeeding?

I notice your discomfort with a connection twixt UAS and the STALL. I share it. I also am not at all convinced the Pitots misbehaved in any way.

Just as the VSIs were reading "pegged" and "true", these probes could have been legit in their sampling, and the computer confused a "discontinuity" in airspeed with "discrepant" readings at each probe. The a/c apparently can react to these discontinuities as if it is discrepancies, instead. What if the initial airmass shear was additive to IAS? Overspeed determination? What if diminishing in IAS? If sudden, as shear is, could the drop have been 60, 80, 100 knots? Could such a drop in IAS have triggered the "zipper" at VS/SELECT prior to loss of autopilot? There was a WINDSHEAR REPORT on the ACARS, as there was also TCAS alarm?

I mention TOGA and 15 degree PITCH UP not to include it as an option at altitude, but to include it in the PF's muscle memory, and training. It is also possible he selected TOGA to attempt to get the NOSE UP, having not realized the NOSE was UP as high as it was already.

For three minutes, the Pilot was seemingly convinced the a/c had "some crazy speed". The PNF was not convinced, but neither was he ready to take the a/c into a Steep NOSE DOWN recovery. When Captain returned, he never seemed convinced either way, and his presence may have been a fatal distraction to the two at the helm.

So it is possible yet to retain an open mind. Demonstrably it is also predictable to judge the pilots incompetent in the court of easy chair PPRuNe jurisprudence.

Hand of God
 

Re-posted by someone:-


OK. I give in. Aerodynamics had no part in this incident:-

It was:-

The Hand of God.

Or perhaps you might prefer - A Hand Of God? Since; He has presumably more than one?

Agree with that, and for what my opinion is worth (not a pilot, but I know, or used to, a bit about building stuff that flies) it still looks confusing. And it has no examples for cruise (as you say).

"If safe conduct of the flight is affected" is a critical decision point (memory items vs. level-off) yet seems to me to be vague and not defined anywhere. Plus all those nice big graphics of the horizon at 15deg pitch up, and toga...

Does "safe conduct affected" just mean "if no danger of terrain impact" ?

If I immediately have problems controlling roll, is safe conduct affected and should I therefore apply the pitch memory items ?

If in RVSM airspace and I'm struggling to level off, is safe conduct affected and should I therefore go to pitch up :\ ??? [ mind you I should be getting out of RVSM anyway without an A/P ?]

If different pilots on here have different answers, then yes, it's all dangerously unclear.


I'm still not sure 447 ties with any of the procedures though. If they were going for the memory items, and recalled the wrong pitch/power, then FD should have been turned off (in, I think, all versions) - and it wasn't.

On the other hand, if they were going for "level off and troubleshoot"... where does the climb come from ?

It could be misapplied proceudre, it could be inadvertent - I think PNF's comments lean towards the latter, but not conclusive.

What might clinch it woudl be the details of the rumoured (posted on here) AF sim tests shortly after the accident. If most of their crews crashed when faced with the scenario on the sim, then something systematic is wrong at AF and not (only) with the guys on the night. If we knew why they crashed in the tests, then we might be much clearer about why 447 climbed.

I wonder if that info (assuming we believe the rumours) will ever see the light of day, or will AF manage to bury it ?

Me, I'll go for numbers every time (blame it on engineering or software background, take your pick). Text has been written by a frenchman, edited by a lawyer and translated by a linguist - the numbers have far less room to wriggle.

As to overspeed, it could have been - but I don't think it was. Overspeed is shown in the trace on P111, and doesn't trigger in the time region shown, which includes the A/P drop.

Also, PF pitch actions were NU first (see same page, below the line is stick-back, above is forward) - whatever the text might say :)

I think they probably did, but my concern is more that beyond causing A/P drop and Alt Law, pitot failure may have no bearing on what happened. Which means that any other failure leading to alt law could lead to the same result - and fixing the pitots is a convenient scapegoat that leaves larger problems unaddressed.

I judge them not (or try not to) - and am not qualified to do so anyway. I try to understand their actions from a user-interface / controls / human-factors perspective, and I await the human factors report with interest.

If I have come to any judgement so far, it is that there were major systemic operational failures that put this crew up there (along with other AF crews if the sim test rumours are true) unprepared to cope with this scenario. And it was a known scenario, precisely known, but clearly not assessed - until after the crash. To quote AF (from Press Releases ):But did they train, or even assess, their A330/40 crews to handle a loss of airspeed in cruise ? It would appear not (according to the BEA).

Was discussed in several posts some time ago, e.g. http://www.pprune.org/tech-log/45687...ml#post6621079 and http://www.pprune.org/tech-log/45687...ml#post6618515 and http://www.pprune.org/tech-log/45687...ml#post6621079

One reference point is that the ADs allowed for replacing only two out of three probes with Goodrich initially, leaving one Thales.

Didn't find my original remembered source, but did find others including this:
EASA and FAA Airbus Pitot RequirementsWhich probably means everyone was right in that argument - could have grounded planes but didn't because risks were assessed and timescales were managed.

Regardless of what a certain person will make up, the text is pretty unambiguous in this matter. Just DFTT.

Solved, not, discussed at most as just reminded by infrequentflyer789 here.
Diagnostic also kindly sent me some references of interest, but more related to the spiky traces phenomena in general. here and here.

I share your analysis alexd10.

IMO this has to be seen as a general comment only. As a matter of fact, the flight was not proceeding in NAV anymore but in HDG.

The FD/AP VERTICAL MODE is a critical parameter for the all event, after, but also before the AP disconnect.
BEA ... where is the trace ?

Vertical Speed
 

To me, there is a much bigger issue than the "sawtooth" trace of selected vertical speed values prior to 2:10:06 (in reference to the graph in posts #1186 and #1195) -- WHY is the selected vertical speed shown as being constant in those graphs after 2:10:42? If what the PF saw on the VSI during that period was a constant, slightly positive value, completely unresponsive to his pitch inputs, it seems likely that he would conclude that the VSI reading was erroneous.

Hi,

I asked before

And no answer ..
I asked this because in the BEA report N°3 the PF tell " seem's we have a crazy speed"
If the PF know how react the A330 to the roll command ( he give many .. reference PF SS mayonnaise graph) he must also know if it was high or low speed ...
Logic ?

Hi jcjeant,

99.99% of the PF's manual flying experience will have been in Normal Law, where stick deflection controls a roll rate independent of air speed. When he was faced with ALT Law, he had direct control of the ailerons. You get full deflection of the ailerons with full side stick - independent of the airspeed. The roll rate at high Altitude is very rapid, I've never experienced this this for real - but I am reliably informed that it is "very twitchy".

PF seems to have been struggling to maintain wings level whilst he was over controlling. Whether he was flying fast or slow could not be part of his evaluation - because he probably had no previous experience of ALT Law at high FLs to base his judgement.

Hi jcjeant,
When the PF made that remark, the real airspeed was about 125 kCAS: BEA Interim Report no.3, p.76: At 2 h 12 min 04, the PF said that he thought that they were in an overspeed situation, perhaps because a strong aerodynamic noise dominated in the cockpit.

Yes, and that was reported. I assume: no change reported, then no change occured in vertical mode. You assume differently, that's your call. :)
Let's hope for more traces in the final report.

It's constant... for some time... but changed again at ~02:12:50
:confused: :confused:
Didn't you mistake selected vertical speed (shown on the MCP) and the current vertical speed (shown on the VSI) here?

CONF,

I admit to being baffled by the SEL'D VS trace.
The 'zipper' before the AP disconnect looks like an FDR or postprocessing graph software artefact (as suggested), but why does it jump to +5000ft/min shortly after the disconnect, and then drop abruptly back to approx. zero?

I seem to remember somebody saying earlier that it was not the 'raw' MCP setting, but an 'intermediate' parameter.

Isn't it a red herring, if the AP/FD was in ALT CRZ mode?

I'm not 'au fait' enough with the Airbus FCS.

Hazelnuts39. It would be interesting to imagine a wind tunnel with the forward ten metres of A330 fuselage mounted level and subjected to 200knots of airstream impacting the belly at 35 degrees AoA.

PF makes his comment, and it was discussed at length, re: motive, CRM, etc. Clearly, at least one airman was baffled by the noise.

A discussion of the zipper trace would be interesting. An understanding of vertical speeds, commands, and events that had an effect on autopilot would be progress. Thus far, no possibilities are broached, and I for one would be grateful if some one would take it on.

A novice' question might be: the VS select 5,000fpm UP just post a/p drop; is that related to the ensuing climb? Because that would mesh with Pilot's inputs to produce 7,000fpm.

Is there evidence of the absence of the a/c having a role in the climb? It would be necessary prior to conclusion that the PF acted alone.

Thanks, Lyman, for confirming you do not have the slightest notion of aeronautical engineering...

Give it a break, mate... you're trying to play way outside your league.

The airstream in windtunnels is horizontal (except in spin tunnels where it is vertical). You would need a relatively small airplane model to mount it at 35 degrees AoA without getting too close to the tunnel walls. What would you be interested in? I would be interested in the pressure near the static port, and in the pitching moment at various angles of elevator and THS. I take it that the noise registered on the CVR is generated by the flow separation on the wing. And why 200 kt? The airplane wasn't stalled until the speed was much less.

Though I'm curious about the cause of the 'zipper trace', I don't see any effect of it on the airplane. As the AB simulation shows, the airplane behavior is entirely consistent with the SS control input. There is no delay in the response in pitch attitude.

HN39. "It would be interesting to imagine....." 200 knots? At 15,000fpm down, and 60 knots forward....... I am interested in what might be considered the type and level of the acoustic environment in the cockpit.

Chris:

I am less embarrassed by my apparent lack of Ae, than I would be to own your lack of imagination.

Let's go for another round, shall we?
 

My apologies, sir, I indeed wrongly attributed the quote to you. Sorry. I'll do my best not to let it happen again.

Interim 3, page 59. Oh, wait a minute...

So you know about them!

I have some bad news for you, sir. If you think procedures are confusing as you sit comfortably in front of your computer screen, there is no chance in Detroit you'll be able to recognize the one to be applied at 0400 AM at FL 330 when hell breaks loose. You can't be an Airbus pilot unless you radically improve your airmanship. As Boeing procedures, despite indignant and unsubstantiated cries they're otherwise, are not radically different from AB's, you probably won't cut it right now as Boeing pilot either. Or Embraer pilot. Or Tupolev pilot. Or any airliner pilot.

Good news is: given solid & undamaged head mounted computer, clean bill of health, good hand to eye coordination, lots of effort and determination and considerable time, one can become a pilot.

Good thing they were designed with pilots in mind, not PPRuNers.

Flying aeroplanes is tough job! One of the requirements thereof is to promptly tell the difference between "below acceleration altitude (tipically 1500 ft above ground level) and "above FL100". Not for easily confused.


Do you have a slightest idea what was the topic you brought on and we were discussing?

It was flight instruments (ATA 31) and autoflight (ATA 22), not FBW (ATA 27)!

Regarding EFIS/AP/FD there is not much difference between Airbus and the rest of the world. Heck, Piper Cheyenne has more FD buttons and not much less FD modes than FBW Airbi. I don't see that allegedly better solution applied anywhere!

Not applicable as aeroplane was in ALTN law. No overspeed protection there.

There shouldn't be any connection. If it weren't for utterly paradoxical pilots' reaction, that is.

No. It's a tough life in the cockpit, when one has to know all the time what is safe and what unsafe. This game is not for faint hearted. Or feeble minded.


You might want to have a look at interim2, pages 100-103. 37 330/340 crews faced with the same problems as AF447's did. In real life, not sim. All survived. Surprise, surprise.

I have never used anything but ALT CRZ in cruise, never heard of anyone who didn't, interim 3 is pretty clear on it and anyway I don't think that "Superiority of using V/S mode in cruise instead of ALT CRZ on A330 series aroplanes" thesis is going to get anyone that coveted PhD, no matter how original the research will turn out to be.

It's critical... just how? As long as AP is engaged, it's straight and level. When it drops out, aeroplane behaviour is consistent with the sidestick input in ALT law. Do you dispute that?

Clandestino. Can you school me on Overspeed, and A/P reaction? Also, what of a sudden loss of speed (Indicated) that would require a/p remediation? How does the autopilot control these deviations from "straight and level"?

If the computer senses and controls A/S with Pitch, it obviously stays in NORMAL LAW for the duration. "there is no Overspeed Protection in ALT LAW." If Overspeed prot had latched, and the IAS were lost, what would be the a/c's reaction? What does the a/c do with "Overlap".

IOW, with the a/p's manouvering limits, what occurs to the "gradual" diminution of Control LAW? Say the a/c was in a protection, and the a/p limited out, what then? If in NORMAL LAW in OVERSPEED protection, what does the a/c do with a conflict? How long would NORMAL LAW obtain, after UAS had tripped the a/p?

Lyman,
Except where airspeeds are erroneous due to either blockage of pitots or due to extreme AoA causing large errors in the pitot and static pressures, they correlate closely with groundspeed derived from IRS and GPS. That correlation excludes sudden large loss or increase of airspeed with meteorological origin.

@Clandestino:
With respect, sir, you (and the BEA) do not know that those "37" other crews faced "the same" problems as AF447 (how many of the multiple elements of each situation are you comparing?). Of course those events listed in interim 2 appendix 7 are believed to be icing events of 2 or more pitot probes, but that is the only similarity which the BEA state. As the BEA explain in interim report 2, there is insufficient data gathered for most of the other events, for detailed examination. Only 13 events had enough information for such examination (see interim report 2, section 1.16.3, PDF page 51 onwards).

As I said to you on 31 Aug the behaviour of some other crews where such detail is known, is not without fault either - so the degree to which AF447 is an exception is far from black & white.

Of course AF447 are the only crew to crash, but that is a very superficial view, IMHO. Look deeper into section 1.16.3 which I mention above, and several similarities with (incorrect) crew behaviour are seen amongst other crews too.

Let's not get into a reciprocal slanging match (again?) but with more than 30 years in "Ae", I think I know the difference between unfounded imagination and knowledge better than you do, it seems.

Clandestino

I accept your apology regarding the "hamster" mix up.

As to the rest of your remarks directed towards me, well, normally I would ask some questions regarding the small bits of technical input you mentioned. However, I don't think you are the right person to ask anything of, given your mental disposition.

I will say, your apparent suffering from cerebral flatulence that you mentioned, could be a problem, a charm school might help, but you should ask your doctor what's right for you.

Hazelnuts. Hmmm. Groundspeed? GPS? Inertial? I am entertaining transient (brief) discreps. Only as little as 30 knots can cause an ADR problem? Here is what I mean by overlap. Are you saying then that the airspeeds were only erroneous, and can have no basis in reality if their effect was to cause error?

Big airplane, lag to velocity? Besides, if reversing, there is a possibility meteorological challenges can go under the "Radar"? (Discovery).

Chris: I most quickly grant you your observation re: "unfounded" imagination. Imagination, by its nature, is vulnerable to ignorance. If I am ignorant of the platform, I welcome your corrections, no slang intended.

:ok:

Bloomberg:

Air France Crash Book Shows Pilot Confusion

The first book investigating the Air France 447 crash...
Publishing what he says is the first full transcript of the pilots’ voice recordings, French aviation author Jean-Pierre Otelli describes a scene in the Airbus SAS cockpit that is dominated by confusion, a lack of coordination, and denial among the flight crew as the jet plunged through the night sky toward the ocean surface. Otelli, who specializes in aviation safety, publishes his book “Piloting Error, Volume 5” today.

Air France Crash Book Shows Pilot Confusion - Bloomberg

Documentation

Flight Crew Training Manual Airbus Air France
AFR A330/A340 FLEET PIR-010. P 4/4
FCTM 09 JUL 08
PDF_F_FCTM_AFR_TF_F_EU__20090113_FCTM.pdf

quote Rudderat:
Hi jcjeant,

"99.99% of the PF's manual flying experience will have been in Normal Law, where stick deflection controls a roll rate independent of air speed. When he was faced with ALT Law, he had direct control of the ailerons. You get full deflection of the ailerons with full side stick - independent of the airspeed. The roll rate at high Altitude is very rapid, I've never experienced this this for real - but I am reliably informed that it is "very twitchy".

PF seems to have been struggling to maintain wings level whilst he was over controlling. Whether he was flying fast or slow could not be part of his evaluation - because he probably had no previous experience of ALT Law at high FLs to base his judgement. "

Correct as far as alternate law is in play. In normal law, the roll efficiency starts to decrease above Va (250 knots in the A320 family) to protect the airframe and wing structure. That seems little known to pilots, but when you are under 10000 feet at 250 knots as mandatory, you have the best possible roll efficiency for traffic avoidance (relatively high speed with full aileron deflection).

I learnt recently that the autopilot in cruise "trims" the airplane for 0,9g, but of course maintains 1g through elevator position. When you loose the AP, the plane has a natural desire to climb...Not very useful, but sharing what I learn!

Literal transcription of CVR
 

Jean-Pierre Otelli's book: "full transcript of the pilots’ voice recordings" acc to the Bloomberg/autor or " a part of a literal transcription of the Cockpit Voice Recorder" acc. to an inflamed BEA ?
http://www.bea.aero/en/enquetes/flig...ctober2011.pdf

Had anyone read it? And found some relevant for accident?

Hi Narval,
The only aircraft I know which did that was the L1011 during an Auto Land Approach. (AL selected)
Every other aircraft I've flown, when the AP drops out, remains in trim and continues along the same trajectory. I can confidently report that's true for the A320 series in Normal Law. (If it does what it says on the tin - then it should be the same in ALT LAW also)

Hold up - I remember reading something like that about the A320 (which has a marginal tendency to climb), but also that the same situation is not the case on the A330/340, which hold their trajectory. Is this the same thing we're talking about here?

alexd10, the book doesn't contain a "full" transcription.
=> BEA is right to describe it as "a part of a literal transcription of the Cockpit Voice Recorder".

The book is not a "revolution" in my opinion, but come with some indications. Note that the BEA is right when it wrote:
"This transcription mentions personal conversations between the crew members that have no bearing on the event, which shows a lack of respect for the memory of the late crew members."
Indeed, most of the "new" elements of the transcription in the book are more or less personal and/or not related to the flight. :hmm:

But, in the same time, the BEA forgot (?) to publish (or delayed it until the final report?) some parts of the CVR which are, from what I understand, relevant to the flight. The most striking example:
[edit] translation to english, I just tried with google and it's not so good, so here we go for non-french readers:
[/edit]
=> This is not present in the BEA's report.
=> This is what the two F/O said to their captain when he came back...


On the other hand, nothing appears in the transcription in the book from 02:10:15.9 to 02:10:27.
I.e. it's missing the "Alternalte law Protections (lost?)" from the PNF at 02:10:22, which is present in the report from the BEA.

I think Otelli had access to a different version of the transcription (a previous draft, perhaps?) while the BEA has total access, of course, but choose not to publish it in extenso (for the moment, at last).

Overall, the book shows two things:
- A crew may be more "relaxed" he ought to have been to fly in the ITCZ (prior problems)
- A crew that does "not understand" the situation, and whose actions are beyond logic or procedure ... This is best shown in the book that the report of the BEA.

That's after a quick (and uncomplete) reading.

NARVAL. If the THS is trimmed to .1 g less than elevators command 1g at cruise in a/p, will it chase .9g if the a/c neglects to cancel that order after the loss of auto pilot? Is that perhaps why it stopped at less than full UP at the STALL? It hadn't zeroed at the start of its uninterrupted migration to 13.9 degrees NU? Without that cue at the start of UP toward the STALL, was it flying anything other than g? Not following the manual input?

AZR. fwiw, "je ne comprends riens..." was part of a leak, early on. It was not claimed by BEA, but BEA did in general rue the release of unapproved data in a memo. No source could be identified for it, but it was discussed here. So technically it is not "new", only re-released, and still not disavowed by BEA......

If it is accurate, it is a bold condemnation of the aircraft and its pilots.

Yes, and the aircraft. For those who disagree, I can only point out the public side of the behaviour and docs exhibited by the BEA. Review them, and take note how the interface among the principals has been a design/build from the outset, and it is yet to reach QA. It is not a report so much as it will be a creation.

One guess as to the goals of the "ARTISTES"

From the appropriate document: (this may have been presented previously)

Max roll rate in Normal Law: 15 degrees/second

Max roll rate in Roll Direct: approximately 20-25 degrees/second (varies with speed and config, but this figure IS MAX)

If you feel this is 'twitchy', be careful (and smooth).

I could get a roll rate of about 60 degrees/second out of a 727 with partial spoiler deployment (increased spoiler roll differential)

That's minimally 'twitchy'.

Max roll rate for a T-38: 720 degrees/second

That's 'TWITCHY'.

http://i45.servimg.com/u/f45/11/75/17/84/af447_21.png


- The part after the AP disconnect has some logic :
Whenever FD reengage, they do it basic mode, which is HGD – V/S.
The selected V/S commanded by the FD will be the current V/S at the time of the FD reengagement.
There is almost good correlation between both FD and V/S SEL V/S traces, but not entirely.

- The part before the AP disconnect raises questions :

• Why the SEL V/S is cycling from zero to -5000FT/MIN ? That’s not commanded by the crew.
• When did start that regular cycling ?
• The V/S trace in red is partially masked by the green line, but seems to indicate an oscillation getting stronger, up to the AP disconnect.
  
  http://i45.servimg.com/u/f45/11/75/17/84/af447_23.png
  
  
• It could indicate that the AP was trying to follow that cycling SEL V/S ?
• What does indicate the AP/FD VERTICAL MODE trace for that very period ?
• What does indicate the AP/FD VERTICAL MODE trace for the all event ?
• Where is the AP/FD VERTICAL MODE trace ?

If procedures had been followed, the F/Ds should have been switched off until the problem was solved.


Because the red trace is masked it's hard to tell, but it could equally be the lumps and bumps you'd associate with turbulence. Some confirmation that SEL V/S is not used in the mode the aircraft was in would be useful - but there are plenty of explanations, such as the FDR using that channel for something else at that point - I'm almost sure the negative peaks indicate a bad value rather than anything selected though.

Let's reserve judgment until the final report before we get our knickers in a twist though.

CONFiture

Please note in your second graph above, (the truncated version), That the a/c is shown oscillating in vertical speed up 1000fpm, down 1000fpm.
In autoflight. This is actual vertical speed, not selected.

The Selected trace covers it up, but I think it may be important.

edit: Sorry, this is repetitive, and Dozy squeaked my out. I'll leave it, so there is a record of my notation. Not like me to jump the gun before reading your complete post (as if).