noske

I am aware that whatever Langewiesche writes cannot be a primary source of technical information. He's good at telling stories for a broad audience without making too many errors on the technical details. No more and no less.

But the bizarre summary quoted by Machaca ("As Langewiesche discovered, the pilots had become confused..."), the sensational conclusion suggested by the article's title, and Langewiesche's questionable decision to work for a magazine that many seem to see on the same level as "Vogue" make the article look sillier than it actually is.

rudderrudderrat

Hi _Phoenix_,
BEA asked the same question on Page 210 Final report.

"4.3.7 Ergonomics
The crew did not de-activate the flight directors and did not call out any changes in FMA mode. It is not sure that they noticed the appearances and disappearances of the flight director crossbars. It is likely that the crew did not know of the mode changes when the flight director became active again, reading and assimilating the displays on the FMA in dynamic and stressful conditions not being instinctive or natural. It seems that requiring an action from the crew to re-engage this automatic system would, on the one hand, lead to a consistency with the autopilot and the autothrust, and on the other hand stimulate a check on the modes and the consistency of the commands presented at the time of the re-engagement.

Consequently, the BEA recommends that:
EASA require a review of the re-display and reconnection logic of the flight directors after their disappearance, in particular to review the conditions in which an action by the crew would be necessary to re-engage them; [Recommendation FRAN-2012-047]

Further, even if it is not sure that the crew followed the orders from the flight director while the stall warning was active, the orders from the crossbars were in contradiction with the inputs to make in this situation and thus may have troubled the crew.

Consequently, the BEA recommends that:
EASA require a review of the functional or display logic of the flight director so that it disappears or presents appropriate orders when the stall warning is triggered. [Recommendation FRAN-2012-048]"

neila83

Is it just me or is there a lot in that Vanity Fair Report that wasn't in the official CVR transcript? And a lot that really helps understand the cockpit dynamics and some of what was happening? Some of the conversation/lack of it is slightly alarming, the Captain seems focussed on anything except managing the flight prior to going to the bunk.

roulishollandais

Thank you rudderrudderrat for your acurate post - very far from the Vanity Fair/Langewiesche magma including many false assertions - .

roulishollandais

Reading Langewiesche - supposed expert "human factors" - in Vanity Fair in October 2014 and the Airliners.net forum commenting on the same ... vulgar article ... shows that the forum PPRuNe is still better than that of Airliners.net !
I am appalled that a "specialist" human factors as Langewiesche could write so much against facts. The manipulators write beautiful texts, with 90% true and 10% false.

Owain Glyndwr

Phoenix


Are we looking to a Dutch roll mode or to an adverse yaw damping?


Good question! but impossible to answer definitively from the published DFDR traces.


OTOH:


One cannot roll the aircraft without exciting all three lateral modes, so dutch roll behaviour must be in there somewhere!


IIRC the outer ailerons are locked out at high speeds and with much of the rolling moment coming from spoiler deflection on the downgoing wing I would think it quite difficult to generate any significant adverse yaw.


PS: Looking at the traces again with a critical eye I see there was a permanent small rudder trim offset ( 0.5 deg right rudder) which could have generated a slow spiral divergence to the right.


PPS: I am now reliably informed that this apparent indication of a standing trim offset is a quirk of the A330 design - something to do with dis-similar metals in the linkage and changes of temperature apparently. This would make my speculation a nonstarter, but to maintain thread continuity I'll leave it here.

DozyWannabe

Right, but the rudder position appears to be more-or-less centred correctly. Would the Yaw Damper not compensate? Unfortunately the graphs are imported at a low resolution, so it's harder to determine exactly what the Yaw Damper and TLU were doing exactly (0.5 degrees offset being relatively minor on those graphs).

mm43

@Owain, Dozy

I've been following this current discussion, and remembered something similar in Thread No.8, i.e.

Post #219 - AF447 - Thread No.8

DozyWannabe

@mm43 - Cheers for the reminder!:

However, (and begging your pardon if I'm mistaken) - in that case you were talking about the behaviour after the stall had developed, whereas here the discussion seems to be centred on the initial right roll immediately after AP disconnect - i.e. when the aircraft was still flying.

[EDIT : Sorry - missed this...]
Right, but on the second roll input to the right (02:10:13), he has the stick over a little further for slightly longer than he had it to the left, which could also be another explanation for the right-roll tendency. Additionally, they were in turbulence, which makes detailed correllation of input to response significantly more difficult.

OG's comment regarding the Yaw Trim setting is very interesting, because that may indeed have initially caused a tendency to roll to the right - however (and despite the low resolution of the graph import), it appears that the Yaw Damper might have been compensating towards the left shortly after AP disconnect.

Again - that's a possibility, but by no means a certainty.

It doesn't always, which is why the first memory action in the "safe conduct" Airbus UAS drill:
http://www.bea.aero/docspa/2009/f-cp...nexe.06.en.pdf

is to manually ensure that both AP and FDs are OFF.

Owain Glyndwr

Dozy


Right, but the rudder position appears to be more-or-less centred correctly. Would the Yaw Damper not compensate? Unfortunately the graphs are imported at a low resolution, so it's harder to determine exactly what the Yaw Damper and TLU were doing exactly (0.5 degrees offset being relatively minor on those graphs).


I agree it is very difficult to read the rudder deflection - after a couple of attempts at digitisation I conclude the actual rudder and trim deflection are both around 0.3 deg rather than 0.5.


The yaw damper reacts to rate of yaw which would be VERY low in the early stages of a spiral divergence so it would be unlikely to have any measurable effect on these few seconds after AP disconnect. Prior to that of course the AP was holding the aircraft steady.

PJ2

Dozy;

You've mentioned a previously discussed key point in all this and that is the complete absence of memory items employed to deal with an abnormality. Instead, the PF launches into something on his own, leaving the PNF out of the loop, bewildered and trying to catch up while guessing what was happening.

Had the memory items been called for and actioned as trained and the ECAM discipline executed as trained, the stall and all ancillary effects of the stall would almost certainly not have occurred.

Some may put the crew's response down to "startle effect" but I don't think so and I don't buy it, for that is what training is all about - to avoid or reduce the effect of surprise and momentary anxiety and to maintain cockpit discipline in the face of elevated risk to the aircraft. The point is to stabilize/maintain control of the aircraft, start the drills/checklists, secure the aircraft and decide the next course of action. That is absolutely standard trained/expected behaviour in an abnormality or emergency, (which this was not, until it was turned into one).

I think there are important lessons here including flight control systems, automation dependencies, FBW design, triage of warnings, loss of the stall signal, movement of the THS during an abnormal flight control computer condition and so on, but the price of such learning in exchange for the original set of behaviours which led to this outcome, for whatever reasons such occurred, is obviously extremely and needlessly high.

Machinbird

It has been apparent to me for quite a while that AF447 encountered a roll PIO after the the autopilot dropped out. What I have been surprised at is the lack of understanding by other pilots, particularly some very experienced ones, of just what this means to the pilot who is attempting to fly during a PIO. In fact, from the low key recognition by BEA of what the roll oscillation implied, I suspect that they did not appreciate the impact of a PIO on the pilot flying.

As I have said in the past, I did briefly experience a divergent roll PIO in an early visual simulator, and I do understand the potential impact on the pilot flying and it is devastating in a number of ways. Apparently this experience makes me part of a very small group who had an encounter, so I feel obligated to try to pass on an understanding to others.

Supposing you were hand flying and your aircraft suddenly began to roll continuously in one direction. Would you be able to handle that using CRM?
The best you would probably do is have both pilots applying counter roll yoke/stick inputs. It would de-structure a crew in short order if you did not get it under control.

Having your aircraft roll in an oscillating manner contrary to your control inputs is equally distressing. I have described my experience as like beginning to fall and desperately looking for a hand hold. Does that sound like another day at the office?

PIO only occurs when a pilot is executing direct manual control. Since most of you are flying almost the entire flight on autopilot, you will have fewer potential encounters with PIO than was previously the case in my day.

There have been some PIO incidents with modern aircraft. Some that come to mind are a few roll PIOs on approach in the A320 series and AA587 with the yaw PIO, and of course AF447. They are infrequent and difficult to predict. They are very dependent on pilot control strategy. You can just about bet that Bonin routinely employed the mayonnaise stirring technique when he flew.

PIO incidents greatly increase the probability of accidents. There are a few things that can be done during training to decrease the likelihood of encountering PIO while hand flying.

• Expose pilots to PIO events in simulators. Once in a lifetime should be sufficient.
• Monitor pilots' control techniques during training and promote a minimalist style of flying. Just because the FBW system lets you get away with mayonnaise stirring does not mean that it will not bite you some day as it did with AF447.

wiggy

Leaving aside the timing of bunk rest I think the translations aren't great, but we need someone who is both a French national and a French speaker to really join the dots on this. I live in France so I've got a bit of a handle on the shrugs and winks side of thing ...but that said my French language isn't great. Even so having read the BEA French language CVR transcripts a while back I think the Vanity Fair translations, whilst being correct at a school boy/girl level, are a bit "clunky" and don't really convey the full meaning or emotion.

jcjeant

wiggy
The BEA French language CVR transcripts convey also not the full meaning or emotion
This is not in regard of each sentences .. explanation about the tone used by the speaker
The real sound record of the CVR is the evidence you need for have the full meaning or emotion (provided your mother language is french)
AFAIK

DozyWannabe

Right, but as you know, it doesn't always work that way - and startle effect is now very much a recognised phenomenon. I believe we were discussing the Orly Tarom investigation earlier in this thread in which 80% of pilots tested instinctively pulled up when they received a Stall Warning they weren't expecting.

My view (for what it's worth) is that it's a more nuanced scenario than saying the whole crew were subject to startle effect. I believe Bonin certainly was initially, whereas Robert - while he didn't break into the UAS drill - was definitely trying to be more analytical. I'm going to cross-post something I put in a PM a couple of days ago to illustrate what I see:

In fairness, I don't think this reference was particularly "low-key" - they certainly seem to have considered the effect on the PF (it's worth reading the whole section, but it's a long piece, so I won't quote it all here):
Going back to the discussion of the roll, I was clued in to the existence of Figure 64 in the final report, which is quite illuminating - particularly the 5th (Roll attitude), 7th (Lateral wind) and 8th (Vertical wind) graphs down.

Looking at the bottom two graphs (7 and 8), you can see that the prevailing lateral wind was a crosswind to the right which has a high peak of 35kts and a low peak of 5kts. In general, from 02:10:08 to 02:10:20 the right crosswind is around 30kts. The Vertical wind graph indicates an updraft of around 15-16kts at the point of AP disconnect, which if I'm not mistaken would, in combination with the right crosswind, initiate a tendency to roll right.

Comparing this data with Fig. 26, it appears that Bonin's first left roll input at 02:10:07 coincides with a drop in the prevailing right crosswind to 8kts, which I suspect goes some way to explaining the initial overcontrol. The crosswind subsequently picks up again to 30kts one second later, and remains around there for the following 12 seconds.

What is interesting to me about the 5th graph (Roll Attitude) in Fig. 64 is that there are three plots. One (blue) is the DFDR data, which we've seen in Fig.26. The second (orange) is data from a simulation with both the wind and pilot input reconstructed and, unsurprisingly, this matches the DFDR data closely.

Most interesting to me, though, is the third (pink) plot, which is a simulated reconstruction of the aircraft's response to the winds *without* the pilot input. Now - presuming all the other factors (rudder trim etc.) have been fed into the simulation, this should give an accurate picture of the roll condition had the controls been left alone.

As such, the pink plot indeed indicates a right roll - however, it maxes out at 15 degrees (which is significant, but still less roll than would be considered normal) at 02:10:25. This roll attitude reduces with the reduction in average crosswind and updraft - and by the time the speeds start to come back online at around 02:10:38, the roll has reduced to 7 degrees.

Now - *if* this simulation is accurate, it indicates that there was no immediate risk of a spiral condition that required immediate attention. Obviously there's no way of saying for certain, but to me it indicates that there is at least some degree of accuracy to the "should have left the controls alone" theory.

Sorry for the essay folks! I'm away off with work for the next few days, so if I don't respond it's not because I'm ignoring people.

Machinbird

Dozy, I totally agree with your analysis of Bonin's mental state, except you really do not understand the reason for Bonin's mental state. That was a direct result of the PIO.

Bonin was startled by the AP disconnect, yes, but that was not what sent him into PIO land. His past bad flying habits caught up with him. He did not initially appreciate that he was no longer in Normal law, however he made control inputs that were unexceptional (to him) and had the aircraft react in an entirely unexpected manner. From that point on, he was mentally fighting for his life. That is the effect of PIO on the pilot flying, and you might call that a startling result, but it was not the result of startlement.

If BEA had personnel with PIO experience on the team, they should have made that connection, but that is not the direction they took, ergo, they were not able to properly cover that base IMO. This is not to cast doubts upon their work however. It seems that hands-on PIO experience is very rare.

I am not sure Bonin actually adapted to the new AF447 flying characteristics. It is equally possible that as the aircraft slowed and Q decreased, the aircraft's response began to match his mental model and he was finally able to stabilize the roll oscillation.

jcjeant

AF447 : un accident d'une extraordinaire complexité - Page 2
Even because the Stall alarm problem (- 60 knots no alarm ... + 60 knots alarm)
the pilots reflex is not good
When stall alarm sound .. it's never good to pull (raise the nose)

PT6Driver

Machinebird
Not often I post on here but..
The PIO is somthing of a side issue, because his imediate reaction is to pitch up. The slowing speed will surely then have compounded the roll control problems.
There was a complete and utter breakdown of the methodology in dealing with non normal events. No one called the correct failures, no one allocated tasks, no one had any command of the cockpit far less the aircraft.
PF's initial reaction to pitch is what killed him and everyone else.
We need to look at why he pitched up and why did the pair of them not follow any known procedure for dealing with problems.
If 80% of pilots pitch up when faced with an unexpected stall (witness the Colgan crash) then there's an industry wide problem that needs addressing far more urgently than the roll issue.
(I am not contesting the fact that as the event continued the issue would have worsened his mental state)

AdamFrisch

What the Langewische and Vanity Fair article highlights will not be found in any accident report. It's also what makes it so interesting. And that is that French pilots are predominantly from the upper classes, well educated and from "good" and respected families high up in academia. All the pilots on AF447 were graduates from prestigious schools. Something they share with many Southern European, Asian or Middle Eastern airlines. Rich kids, basically. Rich kids who might be there simply because it was expected of them, or it maintains their social status, not necessarily because they're great pilots or have a passion for flying.

This is not the case in the US, Canada, UK and northern Europe. There's a fundamental difference there. And it is possible that it could have played a big part in the outcome...

jcjeant

PT6Driver
At the early time .. maybe this ......