AF 447 Thread No. 7
 

AF 447 Thread No. 7

Thread part -

(a) #1 starts here and finishes here. Posts = 3890
(b) #2 starts here and finishes here. Posts = 2537
(c) #3 starts here and finishes here. Posts = 2071
(d) #4 starts here and finishes here. Posts = 1061
(e) #5 starts here and finishes here. Posts = 1978
(f) #6 starts here and finishes here. Posts = 1683


Links to the various BEA reports are given below. If I have missed any of the useful papers, please PM me with the URL and I can include it.

(a) BEA site - French, English
- Report link page - French, English

(b) Interim Report (No, 1) Jul 2, 2009 - English

(b) Interim Report No. 2 Dec 17, 2009 - English
- Update Dec 17, 2009 - French, English

(c) Estimating the wreckage location Jun 30, 2010

(d) Wreckage search analysis Jan 20, 2011

(e) Briefing and associated update May 27, 2011
- Briefing - update French
- Briefing - update English
- Briefing - update German
- Briefing - update Portugese

(f) Interim Report No. 3 July 2011 - French, English

Miscellaneous pertinent links -

(a) Airbus Operations Golden Rules
(b) ALPA FBW Primer
(c) C* and Civil Transports - Cranfield
(d) Longitudinal Flight Control Design - RAeS
(e) Longitudinal Stability: Effect of High Altitude and CG - Boeing
(f) pitot static system performance - USN (Pax River) FTM
(g) The Problem of Automation: Inappropriate Feedback and Interaction, Not Over-Automation. Donald A. Norman UCSD
(h) Upset Recovery - 16MB zip file
(i) Ironies of Automation. Lisanne Bainbridge UCL
(j) Cognitive Capability of Humans. Christopher Wickens Uni Illinois
(k) Trust in Automation: Designing for Appropriate Reliance John D. Lee, Katrina A. See; Human Factors, Vol. 46, 2004
(l) Training for New Technology. John Bent - Cathay Neil Krey's CRM site


Search hint: You can search PPRuNe threads with a filter in Google by using the following search string example -

ths af447 site:http://www.pprune.org/tech-log/

This will search for mentions of THS in the AF447 threads of tech log only.

Just change the THS in the string to whatever you want to look for. This allows one to search for any term or phrase of interest throughout the threads.

Adding the site:URL end part is the magic that restricts Google to only searching in Tech Log.

This filter technique is absolutely wonderful and can be used generically to find things of interest in PPRuNe - appears to work OK in the PPRuNe search function as well.

In respect of Google searching, JenCluse has added some suggestions -

a) indenting the text block with a (one) Tab, *and*

b) emphasizing the fact that it is a search text block with some manner of . . .
<SearchText>, or
"SearchText", or
?SearchText?, etc

DozyWannabe, thanks for your last post (#1681) in the previous topic.

One question about it: you didn't mention thrust settings during your tests.
What were they? Did you change them during the simulated flights?

On searching for text strings in this long discussion, via Goog
 

Mr Tullamarine:

May I suggest thet your suggestion ref searching, with it's example text, be modified slightly?

I's suggest that the generic entry "ths af447 site:http://www.pprune.org/tech-log/" be modified slightly, by:

a) indenting that text block with a (one) Tab, *and*

b) emphasizing the fact that it is a search text block with some manner of . . .
<SearchText>, or
"SearchText", or
?SearchText?, etc

I'm sure you get the picture :-?

Thanks for the thanks folks. For convenience, I'll link to my original "weekend sim research" post here:

http://www.pprune.org/tech-log/46062...ml#post6793521

@xcitation - Having conferred, we loaded extra fuel so that the FMGC showed MAX ALT FL379. C of G was 32% MAC. The ROD in our experiments maxed at approx 6,000ft per minute, with the VSI needle turning amber in the PFD. One of the reasons I hope someone will perform a later experiment will be to see how leaving the recovery till later in the sequence will affect the ROD, and hopefully also find out how a 40% MAC CoG will affect things. The caveat here is that the later you leave it, the further outside the tested flight envelope you go, and the more divergent the sim's performance from the real thing will become.

@AZR - Initially, autothrust dropped to Thrust Lock. We pulled the thrust levers back to match the thrust, but as we moved them the thrust increased slightly. The TRE then deliberately staggered the TLs slightly to induce a roll to the right which we trimmed out with rudder and the slightest touch of aileron.

After the initial NU pitch increase (induced with approximately half back-stick, as in the DFDR traces), we triggered a very short "G" induced stall warning as we climbed, then when the real warning sounded continuously (as happened in the AF447 scenario) we applied TOGA and held 10 to 15 degs pitch on the sidestick - during which full deflection was required in order to come close to maintaining it - as I said, the nose wanted to come down naturally if I released pressure for even a split-second.

Hope this helps!

Further details I've just been reminded of - the stall stabilised at approx 180kts IAS on the sim control with the nose-up trim at 3 degrees (the A320 hard limit). With full nose-up trim the stall was similar, but stabilised at approx. 160kts IAS. The Stall Warning was not only clear, but so loud that the TRE had to cancel it with the Emergency Cancel button in order for us to hear each other. On the second (full nose-up trim) experiment, all I had to do was briefly glance down to my left to see the trim roll forward - smoothly and *very* quickly - following recovery via sidestick pitch down.

DW: Thanks :)
OK, I understand you never did (or had to) go idle to get the nose down. That's reassuring.
Now, does that apply :
- on an A330 sim
- on a real A320
- on a real A330

?

No problem. As for your question, I suspect there will only be minor differences - but the short answer from my end is - that's for other people to find out. :)

Don't under-estimate the sim.

My boss, the famous RD, said to me: "you can learn a lot about the aircraft in the sim".

True, I'm telling you.

Possibly if a poster might provide pictures of an A320 and an A330, we could verify by looking at them that they have similar stall characteristics.

The purpose of the experiments was not to provide any definitive data on stall characteristics, they were simply a technical exercise to determine if there were any obvious system problems or design deficiencies that would make recovery difficult, which it would appear there are not. We can say that with a significant degree of confidence for the A320 and a reasonable degree of confidence for the A330, as the systems designs for both share common parentage in terms of flight control requirements.

Aerodynamics is not my specialty and I'll defer to those who know better, but if I were to make an educated guess, given that a lot of the same people were involved in the airframe developement of the A320 and her larger sisters, I'd say it's a reasonable assumption to make that they are fairly docile in stall characteristics across the range. The TRE's opinion was that the A320 seems extremely docile in the stall if the sim behaviour was anything to go by - and this guy seemed to know what he was talking about - appearing to be a pilot's pilot of the old school through and through, with types from every pretty much every major manufacturer of the last 40 years under his belt.

Dozy,
Good on you for an interesting experiment.:ok:

I gather you didn't get to a high enough AOA to experience the invalid airspeed condition (or that the sim was not programed to duplicate that feature). One wonders how the little 'Bus would handle that.

The sim is not programmed to handle that (as I understand it's it's a little-understood phenomenon to start with), and even if it did we didn't have enough time to take it that far outside the envelope while I was there. My testing colleague ran them to a greater degree the week before and said that the sim outputs relayed valid speeds (despite the failed air data situation relayed to the PFDs and flight control logic) all the way down, or at least as far down as they went. Just to be clear, this test was simply intended to demonstrate that nothing was untoward with the systems, there was no unintended knock-on effect from Alternate Law affecting the controls and that even with full nose-up trim wound in it should be possible to effect a recovery with sidestick input alone *as long as the situation is recognised in a timely manner*.

One more test which confirms once more that aircraft or its systems are not to blame (if we forget that there still had the problem for a short blockage of pitot tubes) in the minutes that followed the withdrawal of the auto pilot and power levers
The plane is once again not to blame .. it's work perfect
Everything converges again to the way pilots reacted to the event
If they react well .. results will be likely those of the tests
We know that was not the case
So .. stay on the playing field:
Pilots
Air France
Regulation offices (DGAC .. BEA ... etc..)

DW, does the sim produce a lot of noise when “stalled” and held in the stall with high RoD?

(It was reported in an earlier thread that somebody who had heard the CVR said that the air rush noise in the AF447 cockpit was very loud, and may have added to the PF confusion by reinforcing any belief he might have had that their problem was overspeed, as that too would have been noisy; and the pilots had probably never heard either noise before in the aircraft, and may not have heard both in the sims.)

Not particularly, there was a noticeable increase in volume and we were getting bumped about a fair bit - but I can tell you that in that simulator the stall warning was LOUD when it activated, certainly loud enough to get my attention even with the headset off. The only way to compare relative volume would be on the CAM track of the CVR. Hopefully the CAM and HOT sources will be made clear in the transcription of the final report.

It is more complex than that.
With full back stick, the PF managed to silent the STALL warning and to temporarily obtain a V/S between zero and -2500ft/min (it is not clear which V/S is recorded ...)
Of course both PNFs have no idea what are the inputs of the PF for that time.

Vultures won't fly much on twisted data.
Find something else.

never say never,

your first comment show that with a bit adrenalin in your hand the spring in the SS is not so strong as expected from the designers, and this can play surely a role for the beginn of AF447s zoom climb

the second one..... if this cancel button would exist also in a real A330, I´d say they would have pressed it for the same reason.......

:confused: Dozy's saying that it was even stronger than he'd expected... It does. And it would be clearly evident on the CVR, wouldn't it?

grity:

Zorin_75 is right - it was *more* difficult (schwierig oder schwer - nicht leichter) than I expected to hold the sidestick all the way back. And the EMER CANC button is on the A330, in the same place on the ECAM panel - I've highlighted it here in yellow for you.

http://i1088.photobucket.com/albums/...panel_A330.png

well, I have to ask for pardon, for my mistranslation :ouch:

Stick Force, Detent
 

The term 'detent' with SS controls keeps garabbing my attention.

So please excuse my ignorace of how the Buss's SS are rigged.
Reference point is the fly by wire stick on A-7 Corsair II aircraft.

The stick was mounted on a set of force bars with transducers.
Electronics package was set to ingnor stick forces under 4 pounds.

There was very little if any actual travel as I recall.

That's OK - English isn't your first language. I made some quite embarrassing mistakes translating the initial French CVR transcript from Interim Report #3, so I can't be too hard on you.

But he did have consistent stall warning for almost a minute, and a reliable V/S for just over a minute (if my reading of the traces is correct). The altimeter continued to unwind with the standby clearly visible from the right-hand seat, providing a secondary method of checking the aircraft's vertical trajectory.

Please don't think I'm criticising the PF, by the way - I'm just saying that those indications were there if he'd known where to look.

Sometimes the simplest solutions can be obscured in a crisis situation - all they had to do was ask. The question is, why did they not think to do so?

@gleaf - The detents refer to the thrust lever settings, not the sidestick. The Airbus sidestick allows for a greater degree of travel than those generally fitted in fighters, however those comparing them to a computer joystick would really have to get their hands on one to understand that while the comparison works from a distance, they are obviously much stronger - chunkier and stiffly-sprung than that. They have to be to stand decades of constant operation.

IMO, the most important parts of the accident are being ignored, some through ignorance, and some through active effort.

My drum? Again, the intitial 10 seconds (manual), and the ten before.

The crew inherited a maneuvering aircraft, for one or the other of two basic reasons. The AirSpeed was unreliable, or the a/c autopilot was unable to keep up with turbulence. Either way, the a/c was therefore in or on the verge of Upset.

All the interesting blather about an aircraft's performance after the fact is not relevant, it is after the fact. Post Facto.

Actively unaddressed? A rerun WITHOUT TURBULENCE factored in? A lack of notation from BEA re: the "Zipper"?

The further the pendulum swings, the more blatant is this "unknown" zone.

The entire basis for the "conclusions?" Unrelated to the onset of Upset, imho.

Commanded or no, the climb qualifies as Upset. What initiated the Climb?

Crew, in response to instruments.

Ambulance Chasers? Fanatical Toadies? Nil/Nil. fmm.

Side stick
 

Salute gleaf!

You can find a good link and description of the 'bus sidestick by searching one of our earlier threads.

With no small amount of time in the SLUF, I can attest to the use of force transducers for the stick grip. Seems like breakout force was very low and the 4 pounds reference might have been when the physical stick began to move. For FCF's we would brace the stick with our knees and one hand, then exert force on the grip. With CONTROL AUG enabled, the control surfaces would move. Sure made formation flying a lot more precise.

That stick grip was used on the original Viper and worked the same way - no physical movement. Later jets, like from number 30 or so, had 1/8 inch of movement.

@DOZE

I take it that the increasing force per degrees of stick deflection was impressive?

back to our regularly scheduled discussion...

For the record, the turbulence encountered by AF447 was described as "MODERATE", which has a very specific meaning - namely :

I'm sure that given the thousands of FBW Airbuses criss-crossing the globe on a daily basis for 23 years, most of them must have encountered "MODERATE" turbulence at one time or another. As yet there have been no reported incidents of such turbulence inducing an autopilot disconnect, but there have been several cases of disconnect relating to unreliable airspeed. Therefore it's quite reasonable to state that the most likely cause of disconnect was unreliable airspeed, and given the ACARS log and the DFDR records it becomes practically certain that unreliable airspeed was the reason.

As for ambulance-chasers, an example was the Aeroperu 757 crash, in which the initiating event was the inappropriate use by airport ground staff of silver duct tape to cover the static ports. The lawyers still elected to go after Boeing (who, true to the edicts of ambulance-chasing, had by far the deepest pockets), despite the manufacturer having done nothing materially wrong.

@gums - Yes, the build quality of the sim was very impressive in general. There's a reason why the top-drawer *home* sim component sidestick costs approximately $3,500 (thus the real deal I experienced likely costs more) and home computer controllers cost 1/100th of that. :)

Otelli Book
 

Good Morning, Ppruners,

I have eaten all of my prunes this morning, and feel quite regular!
Has anyone found a source for the Otelli book, in English, for sale in the U.S.?

I personally feel that Otelli is out-of-line in drawing any conclusions at all at this juncture, and he comes off as an ambulance-chaser, but that doesn't mean I wouldn't be interested in reading the book.

All I can find is the French version, for $50!

Thanks for any feedback.

From one who "makes" those noises...
 

As noted previously, I am one of "those" engineers responsible for developing the sound models used to generate the sounds in simulators (many, many types now). There would be no vertical speed component contributing to the noise signature in the sim, unless the company producing that sim extrapolated way outside of the available sound data supplied with the aircraft data package. The key driver for aeronoise is CAS modulated by Mach factor.

James R.Chiles

Interesting discussion. Congratulations to Dozy for his field research. I do not doubt any second that what he experienced was very close to what did the crew of AF447.

Still this is no proof that he (and me, and BEA and some other pro Airbus people) are right about the accident. You still could argue that the simulator did not simulate what the real aircraft experienced in her dire situation. That the simulation of the case is just an interpolation of known aircraft situations. That the AF447 was so far away from known that they experienced something completly else.

The only thing that Dozy proofed is what we (Airbus people) already knew: stick movement were consistent with aircraft movements, Airbusses are pretty easy to handle and stall recognition, recovery and UAS situation are possible to handle.

But I'm sure he is right. And that an A330 is behaving very similar if not easier than a A320.

The simplest solution has a name : Fully visible control columns - they tell without talking.

What do you wanna ask ?

What can you reply ?
"I am full left - half back - a third left - full back - neutral - a touch forward - mainly right - full forward - back left .........................."

It is highly unlikely that control columns would have made any difference. Go and read the accident report on the Airborne Express DC8 stall crash which involved control columns - they didn't make any difference to the outcome. Absorb that report and remember it before you post - the control column is another red herring. A proper and comprehensive instrument scan and properly executed CRM are what would have helped. The fact that (as Dozy identified) no one asked is highly significant and indicates a serious breakdown in communication at a vital time, just when communication should be clear and frequent.

The fact that a PM would need to ask the PF what are his inputs on flight control commands is highly significant how the concept of the sidesticks by Airbus is poor.

Read and absorb my previous comment regarding the DC8 event and remember it before you post.


Sidestick is a sure way to waste valuable information in a multi crew operation.

Dozy never replied - Where such statement is erroneous Old Carthusian ?
Please comment.

Stability in the stall
 

There were a few comments earlier that AF447's long stall had been very stable, with a high AoA, although with a fairly steady bank. When flying as a SLF when I have been able to see a wing tip through a rear cabin window across the aisle during Take-off, at what I would guess was Vr, the wing tip lifts aerodynamically, thus increasing the static dihedral. With the very high AoA AF447 had, might not this aeroelastic effect have tended to keep the aircraft stable, awaiting some other control input ?

The aircraft that I flew, attempted to have more rigid wings. (The biplanes even had Flying wires and Landing wires between the wing struts !) And later, I would be unable to see the wing tips on Take-off, even had it been safe to do so. With more rigid wings I would guess that we got a rougher ride in any turbulence for a given TAS.

( This "wing-tip-lift" is not unique to A or B, it is just that I have seen it there, on both, as an SLF .)

Lateral stability in a stall
 

My experience at and beyond stall AOA in swept wing jets (trying to hold the nose up in a low speed scissors) was that centering the stick laterally and using gentle rudder inputs had the best effect in keeping the wings level. But even then the wings rocked a lot and it needed bigger and bigger rudder inputs to pick the low wing up.

I'm wondering if all those spoilers on the A330 wing more than counterbalance the adverse effect of the ailerons and make it practical to reverse the roll using the stick, particularly once well past the stall AOA? Or is it just the rudder effects from the BYDU and pedal inputs that rolled AF447 back the other way.

- I'm pretty convinced you are right. The apparent 'restorative' effect of aileron on 447 in a fully stalled condition has puzzled me for a long time and goes against 'conventional wisdom'. I now believe that the commonality of large spoiler deflections with aileron means we may have to re-write the book on the effect of roll control at the stall..

Hi BOAC,

I don't think so.

On page 96:
At 2.12.45. "The copilot sidestick is to the left in stop position. The roll angle changes from 12° right to 41° right in 3 seconds then fluctuates between about 20° and 40° right (period of 10 seconds)
"Hey you’re in … get the wings horizontal Get the wings horizontal"
"That’s what I’m trying to do"
"Get the wings horizontal"
At 2.12.59:
"I’m at the limit … with the roll"
"The rudder bar"

It strikes me that the roll control worked the wrong way during the stall (which agrees with current wisdom). The Captain ordered the use of the rudder bar, which then picked up the wing.

I agree with Linktrained and think it was simply the massive side slip which they had with 40 degs of bank (due no radius of turn) and the dihedral effect which caused it to stabilise.

Quite possible.

Can someone remind me please? We know that fore and aft SS demands are 'processed' by HAL according to some 'law', be it 'g'/attitude or whatever . What does HAL think a pilot wants when the SS moves over in whichever *** 'law' the beast was in at the time? Roll Rate/bank angle/acceleration in roll?? IE When PF moved the stick to the left, what did HAL think he wanted - and gave him?

Roll control is direct law, stick position -> surface position.

SS commands and actual control surface movements
 

Salute!

Beg pardon, but roll commands do not command a 1 for 1 deflection of either the spoilers or ailerons in Alt1 or Alt 2, only in "direct law". That's what my copy of the FCOM shows.

The commands except in "direct" are roll rates, and the "gains" are set by various inputs and other control surface settings. So no direct deflection of surfaces according SS deflection.

This is a fairly standard implementation for FBW jets. The rudder will move to help rolling and minimize adverse yaw. In the Viper, HAL would command the rudder to very large deflections when rolling at max rates and slow speed.

At high AoA's, I go with 'bird. Use rudder and not aileron for roll. Spoilers can reduce adverse yaw and such, but my understanding is that the outboard spoilers on the 'bus are inhibited once gear up. So the inboard spoliers may not have helped as much as rudder.

The CVR shows the PNF mentioning the "rudder bar" when PF had full SS deflection for roll and the jet was not responding. Here again, knowing you were stalled or approaching a stall might have changed the PF stick inputs. However, in view of most PF stick inputs and PF's possible lack of experience flying at extreme AoA , then I can see a bit of the rationale for roll inputs, but not pitch.

Sent by an old FBW pioneer.

OK - seconds out - round 2

Zorin Roll control is direct law, stick position -> surface position.

Gums roll commands do not command a 1 for 1 deflection of either the spoilers or ailerons in Alt1 or Alt 2, only in "direct law". That's what my copy of the FCOM shows. The commands except in "direct" are roll rates, and the "gains" are set by various inputs and other control surface settings. So no direct deflection of surfaces according SS deflection.

Which FCOM/understanding is correct? It could be important in a confused cockpit. Not a lot of time to open an FCOM volume? Let's get it right?

Ding Ding Ding.

Oh yes- a round is 3 minutes I think - about the time from 38,000 to sea level at a high rate of descent .