SaturnV

jcg and mr. optimistic, we do not know that the passengers were not instructed to be seated and belted. If the PF reduces airspeed to 0.8 for turbulence penetration, would you not expect him to have the seat belt sign on?

On TW800, about 100 bodies were recovered floating on the surface, and its very likely the fasten belt sign was still on when the center tank went.

bearfoil

"Can't release something"

full stop.

Mr Optimistic

SaturnV, thanks, yes I would have thought so too. So the early finds were not indicative of passengers not seated or unbelted and the CC were either seated or unprepared in the rest facility I presume with the impact forces and immediate structural breakup doing the 'selection' ? Don't like to think of any other scenario.

Mr Optimistic

Bear, don't know what the cloudbase was like but at some stage they must have seen the ocean and maybe got an horizon. Bet that prompted an utterance or two.

Khashoggi

The radalt and gpws would have given them some new clues as to the impending disaster. They were monitoring the altimeter tape so it also.

On a moonless or a moon obscured night, getting a horizon may not have been possible,

Either way, at 10000+ ft/min descent the number of seconds for the radalt info and the gpws alerts would have been minuscule.

Svarin

Gentlemen,

computers will always do what their programming tells them to do. Sure.

First caveat is "programming" takes human beings, who make mistakes, be they engineers in a design committee, or pilots on a flight deck.
Second caveat is when physical failures affect the computerized system.
Is it not obvious in our present case that a "physical" fault occured : wiring ?
BEA offhandedly ascribing it to "PRIM2 rejected ADR1" is a feeble excuse for an explanation. No mention of "wiring" ?

I have reliable confirmation from the relevant documents that this failure :

WRG:ADIRU1 BUS ADR1-2 TO FCPC2

is indeed a "wiring" fault, between ADR1 and PRIM2.

This fault appears "simultaneously" (look at ACARS and CMC time stamps) with the acknowledgement by the system that speeds have gone haywire :

PROBE PITOT 1X2,2X3,3X1

These faults are unrelated : wiring + icing. How can this be ?

Has this strangest of dual failures ever been foreseen ?
How can one expect the design to operate as advertised in this likely un-designed for situation ?
Does anyone have reliable information as to the exact state of PRIM2 operation after this combination of faults ?
How can anyone be certain that PRIM2 did indeed latch Alternate 2 Law ?
How can one be sure it did compute identical to PRIM1 & PRIM3 when it is obvious its state is different ?
What happens if PRIMs differ in their assessment of the outlier ADR ?
What happens if, after their 10 seconds ADR cross-checking "window", PRIMs end up in different states, deciding on different flight laws ?
Has anyone considered what could happen to PRIM2 if it lost datalink with ADR1 at the very moment when the PRIMs tried to eliminate the outlier ADR ?
Has anyone studied carefully the first BEA interim report and extracted the PRIMs behaviour at the start of UAS sequence ?

I did. PRIM2 could very well have reverted to Normal Law. With erroneous data fed into it. Definitely not designed for. But quite possible.
Please, do prove me wrong. Do not recite the manufacturer's mantra, I know it and I do not care what the salesmen say. Sh't happens but it does not sell.

So many people posting here assume that the computers reacted "as expected", confusing it with "as programmed". I am deeply disturbed at seeing so many people who work in science and technology embrace blind religious faith into man-made technology. Where has the scientist's healthy doubt and critical judgement gone ?

Any high-tech item is an artifact : man-made. It is never perfect. It fails, even if rarely. It errs. It does unexpected things.

I am of the opinion that in this instance, statistically remote, it did unexpectedly err, compounding a surprising but otherwise manageable situation into something far more dangerous and incredibly difficult to understand in a limited amount of time.

The initial pitch-up, more than 10 degrees nose-up attitude, up to +7000fpm vertical speed, +3000ft altitude deviation, has been discussed quite a lot. Such discussion is always along the lines of :
- deliberate pilot action : then why ?
- unintended pilot action : then how ?

Outraged whys and hows all over. Naive reaction to a very narrow extract of all the available data, delivered to the press at a commercially critical time. Understandable coincidence.

So few are those who suggested it could have been undesired, an unexpected reaction from the aircraft itself, its flight controls computers confused.

Why is this possibility unacceptable for so many involved people ? All aircraft types, from all manufacturers, have at times suffered serious design-induced malfunctions which caused crashes. Such problems are rare. All modern jet airliners are statistically safe. Why the reluctance to acknowledge a very specific flight controls problem ? Why the brand of blasphemy when such problem is invoked ? Why the need to invoke "pilot error" when questions point towards the manufacturer and its design ? This blame-shifting attitude is contrary to the necessary openness and honesty that could improve flight safety. I find it very much akin to religious intolerance.

If pilot error is to be found in this case, it is twofold :
- calling for the Captain to come back to the flight deck used up resources which should have been entirely allocated to flying and cross-checking instruments, control inputs and results between PF and PM, especially in the early stages of the accident sequence. This could have allowed identification and isolation of a flight controls issue.
- failure to quickly and decisively use manual pitch trim, against training, in a flight law that usually excludes its use, to effect the desired attitude and trajectory if sidestick was found insufficient. This could have maintained a safe path against the undesired actions of confused computerized flight controls enough for extended troubleshooting.

These are real things that can really be corrected both in the attitude and flying skills of modern age airline pilots. These can improve flight safety. But the initial 3000 feet climb, 10 degrees and more pitch-up is unlikely to be "pilot error", even though some powers that be, and hordes of blindly faithful technology worshippers would prefer it to be so.

RR_NDB

In respect to FBW, Airbus SAS philosophy and computers with it´s intrinsic complexity (software testability issues, etc.) the redundancy implemented is mandatory for it´s advanced planes.

The use of the 3 Pitot´s (identical) subject to the same problem (recorded in many other similar cases) being scanned by the System and "capable to trigger" a major a/c change makes me concerned.

I think this is a point (before better available Pitot´s) to be reviewed.

Sounds like System Designers used the idea of Redundancy in an almost "wrong sense".

In summary: To scan and process an information of this magnitude of importance and "switch laws" based on multiple "non reliable sensors" does not seem to me a safe approach. Better perhaps to "warn only" and present the separate info to the 3 crew (LH, RH and ISIS).

During "short period Pitot failures" this could be adequate.

And i insist, Pitot´s problems for this new advanced planes deserves immediate attention and this not seems to me a technological problem. IMO it´s just a product (specs, cert. etc) issue. Not critical in other less "advanced" a/c.

Mr Optimistic

Svarin, yes I recall that you raised the 'wrg' report some time ago without getting much response. Your opinion on the apparent NU inputs ?

Probabilities are just that but the CVR would make it clear:

Option 1: one rare failure plus maybe large sub-optimal response (not proven)
Option 2: two independant rare failure plus maybe small sub-optimal response (not proven).

Time will tell.

Perhaps significant attention was being paid to trying to get the autopilot and or computers back on line (you have already suggested this some while back). I noticed in the N Atlantic upset, the report seemed to indicate that hand flying was started only after attempts to re-engage the autopilot failed. Seemed the wrong way round to me. Note I am SLF with perhaps too much to say, so hope no offence.

Edit: Didn't BEA also say something along the lines that the a/c systems were not at fault (excluding pitots one assumes).

Machinbird

Svarin,
Masked faults in aircraft are nothing new. We realize you may be on to something, but we all have to realize that BEA is investigating this accident, and if they have integrity, then they will address this question in their final report. Let us hope they do.

With the highly reliable systems of today's aircraft, you can fly thousands of hours until a random event unmasks a hidden fault. Perhaps they ought to swap out which PRIM has the lead position just to unmask such potential faults. Of course, you have to give the crew the ability to quickly switch back to a stable configuration should a fault show teeth.

Heathrow Flyer

Actually the opposite, as reported by AVweb:
"At no point" on the cockpit voice recorder "is the word stall ever mentioned," Chief Investigator Alain Bouillard said in an interview.

So one can reasonably conclude that the crew did not know that they were in a stall.

RR_NDB

Hi,

Mac

Svarin, your dense post requires extra commenting, to be made later.

Machinbird

When you look at a typical Coefficient of lift curve in a publication, they always stop just past the Clmax point. See example below:


Did you ever wonder why they don't carry on further to the right? It is because past that point, it gets a lot more complicated, varying with aspect ratio, thickness ratio, and Reynolds Number.
A typical extended curve looks more like this:

The double humped curve shown above is rather typical actually. Now we have a clue where the notch in the Viper pitching moment curve comes from.

Fortunately the wind turbine industry has need of post stall airfoil performance and someone has studied the subject a bit. The following pdf file is on a slow server, so give it some time to load.
http://home.comcast.net/~shademaker/...tStall44XX.pdf

I researched this topic a bit because I was interested whether or not AF447 was likely to get stuck in a pitching moment notch of its own. Maybe so, but the THS lower aspect ratio than the wing indicates decreased likelihood of having as much of a notch as the wing might have, i.e. more nearly linear performance.

bearfoil

Svarin


Svarin

"So many people posting here assume that the computers reacted "as expected", confusing it with "as programmed". I am deeply disturbed at seeing so many people who work in science and technology embrace blind religious faith into man-made technology. Where has the scientist's healthy doubt and critical judgement gone ?"

The answer is to pose 'silly' questions, and plant the seed. Engineers and Pilots both have to have the answer, and be first. It takes patience, but in an endeavour where Time is cheap, it is crucial.

First input, then, was in Normal Law?

just sayin'

RR

"Sounds like System Designers used the idea of Redundancy in an almost "wrong sense"..."

They were on the right track when they RFP'ed the software to isolated teams, encouraging anomalous, (but equitable), performance.

Think about it. Three Thales Probes. Let's eliminate two, and retain one. Gone is the need for constant "voting" and the very real possibility of fatally confusing the systems and pilots on the "Altar" of "guess we got that figured out".

With one Thales, one has binary reliability, if it packs up for any reason, we have BUSS. BUSS came after THREE THALES, so it was an acknowledged improvement in the name of critical safety.

Complex is not always the best. If one starts to incorporate some real analysis, the need for complexity disappears. Wait, doesn't that make _____ look bad?

boy howdy. So here comes Pride, Apathy, and Money.

I say NORMAL LAW or Manual flight. Making up stories is starting to get expensive, in machinery, and human souls.

Mr Optimistic

HF, thanks. Difficult to believe the word wasn't mentioned when the stall warner triggered.

Machinbird: there was a change of course at some (late) stage as shown in the BEA graphic, so something was able to effect change.

Svarin: could 'wrg' refer to a failure in message protocol rather than anything physical ?

Machinbird

With a full readout of the DFDR AND CVR recorders, I hope BEA is still trying to get info out of the unprotected quick access recorders. It may not be trivial at all, but will probably be well worth the effort. May even open up new areas for the investigation. (Of course that can be a bit inconvenient when the report is almost ready to publish.)

Comparable to seeing things with an earth bound telescope compared to the Hubble telescope.

Svarin

Machinbird : thanks for the expanded curves, greatly interesting and informative.

Mr Optimistic : Yes, indeed. I tend to incline towards software versions that could communicate very well under all normal circumstances but failed to do so under this very specific ADR-checking 10 seconds process. To be verified and pursued further, but I lack the relevant documents. Of course, software incompatibility is not part of the design assumptions, thus it could only appear as a "wiring" fault.

Mr Optimistic

Svarin, thanks. That is not a 10^-9 probability, I won't bet against you.

VicMel

From LazerDog, #1729:

From DozyWannabe, #1730
There seems to be some suggestion of a possible problem with altitude data from ACARS Message 8 (WRN/WN0906010210 344300506NAV TCAS FAULT). According to BEA Interim report No.2, section 1.16.2.4.1, analysis of message 8 , the most likely reason for the generation of this message seems to be, “...is a monitoring process internal to the TCAS which applies to the standard altitude parameter. The latter is received from the active transponder (it can thus be the altitude elaborated from ADR 1 or 2) and is submitted to a “credibility” test. In actual fact the TCAS elaborates an altitude prediction that it compares permanently with the altitude received.”According to section 1.6.11.6 a compensation factor is applied to standard altitude dependant on Mach but in the example given, for a difference between Mach of 0.8 and 0.3, the compensation to standard altitude is only 0.7%, whereas an 'incredible' gap between a measured and a predicted altitude would IMO be of the order of 2 to 5%, say around 1000' or more.
Another possible reason for the PF NU input?

SaturnV

Khashoggi,

From first interim report on the experience of AF459, flying about 35 minutes behind AF447. (Bolding mine.)

As AF447 did a loop in the last several minutes, the cockpit view would have become oriented back toward the core of the meso convective complex.

The Lufthansa 744, flying 20 minutes ahead of AF447, also observed no lightning.

infrequentflyer789

Actually, both are "true" in that both "this is a stall" and "at no point is the word..." have been reported in different media "leaks" from teh investigation.

So, one can resonably conclude that at least half the journalists reporting "leaks" from the investigation and actually just making it up.