Bonin pulled the A330 into a fully developed stall
 

DozyWB - you make some good points and I agree both Airbus and Boeing have gone some way to educate the flying public on correct AoA reduction techniques. Flight testing FAR/JAR25 demands AoA reduction then 20% above stall speed before adding power.

Low alt, approach to stall min height loss recovery was stressed by FAA and CAA examiners for PPL and Commercial pilots - it instilled the wrong message at the foundational stage of flight training. In aircraft with underslung engines or tractor propellors adding thrust adds to the pitch up moment (not so much at high altitude grant you). Slowing down also activates aft pitch trim (AF447 A330 had maximum aft tailplane trim) which makes recovery even more difficult as the Thomsonfly Bournemouth stall episode revealed: Air Accidents Investigation: 3/2009 G-THOF

The Airbus get out of jail free option (Windshear, Terrain, Woop Woop pull up, etc) was regularly emphasized for LST/Initial Type Rating training - essentially (in normal/non degraded mode) it allows you to get to the light buffet (Cl max) with maximum thrust and no stall. It requires Full Aft stick and TOGA. What did Bonin consistently do in the confusion? The BEA have a video of the flight recorders that has not been on general release - it shows the side stick positions on both sides - the FO also applied aft inputs (his seat was found wound aft at the sea bed).

Bonin was spooked long before the fatal events - read his reaction to the smell of ozone. But his techniques and both pilot's ignoring a stall warner speak of poor training and poor understanding of the situation their reactions had created.

Have you discovered how much gliding Bonin had done and how long it had been since he stalled a glider? The BEA AF447 investigator at a recent RAeS Upset recovery lecture did not consider his glider training to have been very frequent or recent to events.

Apparently the junior crew had practiced unreliable airspeed some time before AF447 and the SIM route was from Rio with drifting ADIRUs.

Cheers! :ok:

OK, but was this "option" ever trained at high altitude? It not only looks inappropriate for high-altitude ops, but also seems to be at odds with what the manual states in that situation (i.e. 5 degrees nose up, thrust lock)...

It also doesn't explain Bonin's apparent (and inappropriate) concern about overspeed which followed shortly after, causing him to attempt to deploy speedbrakes. I agree that his reverting to the procedure you describe is a possibility, but the rest of it looks to me like he was throwing almost random elements into the mix (likely due to panic). Nor does it explain why he pulled up so soon after AP disconnect.

True, but as I said, 5 other pilots (most of them captains and some of them military veterans) have in the past become similarly spooked and made exactly the same mistake. I refer back to the tests done by the BEA in which 80% of the pilots they studied immediately pulled up in response to an unexpected stall warning.

I think I have something with the qualifications and when he did them kicking around somewhere, but the point I was trying to make was that while training was certainly deficient - it doesn't necessarily follow that Bonin "didn't know how to fly", as the media seem to have suggested.

Caricature or the future ???
In one month .. start from nothing ... too
https://www.youtube.com/watch?v=OcrpR1YI1qc

Seems like perhaps only 0.0001% of airline pilots have ever experienced any form of PIO and most do not understand the implications of the AF447 roll PIO encounter right after the autopilot dropped out.

All that wing rocking that Bonin did had the effect of destroying his trust in the reliability of his aircraft's control system so that later when the pitch control system did not behave as he expected (upon stall encounter), he assumed that that had failed too.

Folks have been minimizing the roll oscillation. Statements such as,"it wasn't that bad", "the roll rates were not that high", "he recovered roll control within 30 seconds" seem to predominate. I maintain that those with that view simply do not understand the nature of the PIO beast.

From my very limited exposure to a roll PIO in an older simulator, two PIO oscillation cycles is sufficient to grab your complete attention, and 30 seconds of fighting a PIO oscillation is a lifetime. The rest of you are going to have to take this on faith unless you also have had some sort of a PIO encounter. (And if you do have such experience, lets hear from you.)

Bonin's roll PIO was started by his initial excessively large roll input. The roll PIO was continued by an incorrect control strategy (Trying to get ahead of the roll oscillation). The reason he fell into this trap starts with lack of training in a simulator in roll direct at altitude (ALT2B mode). The other ~ 35 aircraft that successfully survived loss of airspeeds apparently did not encounter a roll PIO.

That Bonin used to fly gliders has very little bearing on the AF447 outcome. When the chips were down, he could not transition smoothly to manual control under night instrument conditions. If he had been gentle with the controls and had just kept on truckin, this would have been a non-event.

As Gums would say, "That's my story and I'm sticking to it.":)

I'm with both of you on that one.

PIO is a very surprising event, and the problem lies in recognising it from the beginning. If not recognized the trained corrective action (to counter the unintended roll) agrevates the problem.

The same applies for the stall situation. Without recognizing an imminent stall situation prior stall warning by analyzing the energy situation of the aircraft (climb- gaining altitude - loosing speed) the stall warning might have been judged as invalid and caused by the failing speed indication. We all had faulty warning lights before, and first thing we do when one shows up or "Betty Bitch" is shouting is validating most of those warnings before acting. Excluding the stall from the equation it comes down to the basic trained behaviour: The aircraft goes down, therefore increase pitch to make it go up again.
That's what they did, and what others with stick shakers did.

@Machinbird, RetiredF4:

You're right - however if I recall correctly from the mega-thread and the report's DFDR graphs, the PIO was on its way to being stabilised, and was almost there at the point the aircraft stalled (apex of the zoom climb) - suggesting that the PF was at least getting a feel for the roll aspect. Unfortunately the consistent backpressure (intentional or not) while doing so meant that stabilising the roll became something of a moot point as the aircraft went into a full stall.

No, my point was to agree with him about teaching the correct process. His point on "min alt loss stall teaching" can have some unintended outcomes in terms of what the pilot has as his objective and actions during stall recovery. The key is to teach the correct stall recovery technique, which will take care of losing altitude (minimize it) by unstalling the aircraft and returning (and maintaining) a flying AoA. IT is in how you teach stall prevention and recovery that the lesson is correctly or incorrectly learned. Teach the correct process.

As to "lower the nose" there's a more technically correct idea behind that regarding "reducing the AoA' though in most stalls reducing pitch attitude will help with the AoA problem one is dealing with.

Sorry to bite at you there, I could have been less terse.

@ Machinbird:
Amen, Deacon.

PIO's
 

Be careful taking advantage of an old dinosaur that can't take any of you up and demonstrate some rather severe aerodynamic conditions, and then how to avoid them or recover.

BTW, Roulli had a great discussion on the USAF tanker crash due to "dutch roll". It seems to have a fair amount of PIO present once the yaw and roll got out of phase.

Retired has one great point about being aware of your energy state, and then your response to an abnormal event. Being low and slow requires a different technique than when you have a warning or loss of a flight parameter ( think airspeed) being high and fast. I can guarantee all that on approach with all the drag devices out and the plane not handling as crisply as when at 400 knots and clean, that my initial reaction for a stall indication was relax back pressure or even push forward a bit. For roll excursions I would relax pressure on the stick and watch for a second or two.

'bird had a great video on the other thread about PIO, and that sucker lost it in about 2 seconds due to his very high "q".

In all fairness to Doze, his idea of the "startle factor" seems to have a basis. But sheesh. We were trained to not get hyper-startled and do things until we figured out what was going wrong.

later from the peanut gallery...

Aye, but let's not forget that the A330 is a much more modern design and its more advanced aerofoil and yaw damper designs make it much more forgiving.

Of course, but what I've been trying to get at is that the training as it stands might not be enough. Specifically I'm concerned about the "inexperience" or "magenta line" narrative overshadowing the fact that we've had several accidents now where even very experienced pilots have been spooked by a sudden abnormality and proceeded to respond in such a way that their aircraft was lost. Startle effect isn't a new phenomenon, but it is nevertheless somewhat poorly-understood on the line. To be frank, I'd be inclined to consider it a type of incapacitation that needs to not only be trained against on an individual level, but also trained in terms of recognition in one's colleagues and consequent assumption of control until the colleague calms down.

Reading the KC-135 mishap report shows that nothing happened in that Dutch Roll PIO that can't be done with a new A330, any modern or new plane. Dutch roll is the most basic PIO (Pilot induced oscillation). The loss of rudder in flight followed by the total destruction of the plane in three parts and explosion in flight followed by CFIT and fire shows the potentiel danger of any oscillation. The modern data recorder and conversation showed exactly that any modern and new plane cannot resist better to resonance developing very quickly threw the pilot's pedals and feet out of the flight envelop.:ugh:
(Figure 6 page 11/54, figure 9 page 14/54 , Conclusion of Brigadier General Steven J. Arquiette page 48/54 of the 1534 pages report)

Interesting development, but that journo wants a quiet word had, because this bit:

is unsubstantiated nonsense. The autopilot disconnected because of the discrepancy in airspeed, that much is true, but I don't think it lowered the nose. As I recall the aircraft ended up slightly nose-down at disconnect because of a bump of turbulence.

Rubbish is rubbish
 

Yeah, Doze, et al, whoever wrote the article doesn't seem to understand aero, flight control systems, autopilot implementation, and the beat goes on.

Sheesh. After so many years, one would have a better idea of how the plane worked and basic aero, ya think?

- All of us here have a decent understanding of the basic and reversion control laws for the jet. However, I can't find details on the AP implementation.

I raise this issue due to my Viper experience( early years) when we lost a troop who was on AP and slowly descended into the Salt Lake. Unknown to us at the time, the AP had an AOA limit that was basically half the "normal" limit - think 13 degrees versus 27 degrees. So with a mission abort at heavy weight, the troop turned back and was BZ changing channels, IFF and such while slowly descending in a gray sky with low ceiling and over a lake that was smooth as glass that day. We single seat folks used everything we could, contrary to popular depictions of us.

- With an AP disconnect and a reversion to the next lower law, the "startle factor" is in play, as some here have opined. And then maybe not realizing the system was in a reversion law with less "protection", hence the "pull back" syndrome. After all, "you can't stall this jet, huh?"

- Maybe the AP control laws depended more upon airspeed than AoA? But my understanding is Otto disconnected immediately.

Good grief, this accident will be the poster child of system design, human factors and sensor faults.

Dozy, have you considered that the one size fits all roll direct control gain became more appropriate as the aircraft decelerated?

As for startlement effects, the initial strong control input that started the roll PIO could have been from startlement, or possibly from annoyance at having to take control. In any case, it was an unfortunate response to a situation that merely required minimal control inputs.

I can't find the use of "standby gains" in the pubs I have recieved from you guys. Only the "direct" mode resembles what we used to fly with pure mechanical connections ( no hydraulics).

Seems the "alt" laws still use rate commands, and I can see this if the rate sensors are still active, and no reason to think they would be influenced by pitot-static problems, ya think. We had a Viper fly for about 10 minutes or so after a large pelican smashed the radome and the AoA sensors - no airspeed and no AoA, only inertial inputs, absolute pressure and "standby gains". The damage apparently got to the flight control computers, which eventually failed. The pilot was flying on instruments the whole time, as the blood covered the forward part of the canopy.

So I go with Doze that the roll PIO was pretty much dealt with by the time the jet ran outta energy and defeated the AoA protections ( same as Viper deep stall entry).

I also go with 'bird that the "startle factor" should not have been the primary cause of later control inputs. There was even a call about "alternate" law early on, wasn't there?

Finally, seems to me that the jet's control surfaces would have been trimmed pretty well when the pitots froze and AP disconnected. In short, let the jet go where it is trimmed before doing anything rash and getting into the roll PIO we are theorizing. BTW, I can't seem to find when the pitot system came back on line. If it did, did the input to the FLCS have to be enabled by the crew?

One of the things I have often wondered about was the sidestick control, left hand verses right hand. I would think normally the Captain or the pilot in control would be seated in the left seat and using his left hand on the sidestick. However, in this instance, the pilot in control was seated in the right seat and would be using his right hand on the sidestick. Considering the abrupt A/P disconnect and reversion to alternate law, plus the startle factor, might this seating arrangement have played a part in first the roll control situation and then the subsequent continual pitch up problem, e.g., lack of sidestick control sensitivity because of not using the usual (dominate) control hand and sidestick? Would it have an influence or make a difference? Just asking…

Dinosaur sir,
You will need to re-read how Alt 1 and Alt 2 differ.
Your forgetter is working too well.:}

Hi Turbine D, Assuming Bonin was right-handed, then he would have been using his stronger arm to fly the aircraft. This made it easier for him to jerk the aircraft around initially.
A number of Airbus pilots have remarked in the past that it would have taken a lot of work to make the lateral control inputs that were made due to the viscous damper in the lateral channel. I suspect that his arm was burning with fatigue after the first 30 seconds and that his ability to make small corrections was then compromised as a result.

As Bonin was copilot .. he was used to control the plane from the right seat (so right hand) .. his usal place
Robert had taken the seat (left) of Dubois (captain) ....
So .. no influence there .... AFAIK

It's possible, but given the situation I'm inclined to give the PF the benefit of the doubt and think that he was indeed bringing the roll under control.

Or no control inputs at all! The A330 is a very stable platform - and whatever one may think of Learmount, I'm inclined to agree with his opinion that with no further inputs the aircraft would have simply ridden out the turbulence as best it could, and after 30 or 40 seconds the ice would have melted from the pitot tubes and the speed indicators come back online.

I'm basing my "startle effect" call on previous similar incidents where pilots made an erroneous initial call on the problem (in some cases because the warning systems were giving misleading signals). Those incidents are Birgenair 301, ColganAir, NWA6231 and West Caribbean 708. In the Birgenair case, the Captain as PF seemed to fixate on the first warning he received, which was an erroneous overspeed. In the case of West Caribbean, the Captain thought he was dealing with a dual engine flameout. In both of those cases, the F/Os correctly diagnosed a stall and inappropriate attitude, but the Captain did not respond.

The other reason I suspect the AF447 pitch input to have been inadvertent is because the two F/Os had been discussing not being able to exceed their present altitude for safety reasons only a few minutes before. At AP disconnect they had well over 30,000ft to play with and they knew that to climb would increase risk. To initiate a climb would therefore be completely illogical, and I suspect this is (at least in part) why the PNF was initially so incredulous as to what seemed to be happening.

There was an "Alternate Law" call from the PNF to which the PF did not respond - we therefore don't even know if he heard it (which - if he didn't - to my mind supports the "startle" theory)

You can see the speeds coming back online in the DFDR traces at about 02:10:35 - at that point the aircraft is established in the climb at around 37,000ft.

No, there was no need for the crew to perform an action to bring them back online.

Did they? Must confess I don't remember so well. As far as I know the primary damping mechanism in the sidestick is actually spring-driven, and the simulator version I had a chance to get to grips with was actually very responsive and not too difficult to deflect once you had it going. The springs do centre the SS, and there is a small degree of 'breakout' force, but once deflected it offers enough resistance to let you know it's deflected, but not enough to cause physical issues.

UAS
 

Origin of that text ???

I suspect it's a Winnerhofer "original". The usual contraction for unreliable airspeed is UAS as far as I know!

That said, the points are fair...

@Dozy
I don't think it is a Winnerhofer original.
Use of "URA" is from somebody who didn't read the BEA report
I don't agree with that complicated "philosophy". Pilots don't fly philosophy, not enough time to "do nothing":*

Alt 1 and Alt 2
 

Salute!

O.K., 'bird. I thot the jet was in Alt 1. So Alt 2 is a different beast WRT roll commands and aircraft reactions.

It still bugs me that I have to figure out exactly what law I am operating with.

Seems to me that "stby gains" ( gear up or gear down values) and basic gee commands/protections and AoA inputs should have saved the day. Then there are all the inertial/rate sensors that don't need a single atmospheric input.

When I talk about the various modes, I am talking about limits on max bank angle, max pitch, max gee, ,max AoA, etc. Then there are the control surface rates that depend upon "Q" - the "gains".

Oh well, the AF447 tragedy shall remain the "poster child" concerning airmanship and aircraft software implementaion for at least a few more years.

This surprises me, because you've certainly been active enough on the thread of doom to know otherwise!

The difference is fairly academic really. I think it boils down mainly to the aircraft being slightly more sensitive in roll - other than that it's fairly transparent.

We covered this in the mega-thread though. The "soft" protections would have stabilised the aircraft until well into the sequence *had the PF just let go of the sidestick*.

The assumption in the EFCS design is that if there has been a failure sufficient to degrade from Normal Law, then the human pilots are given priority over the systems on the grounds that they are likely to know more than the systems do about the status of the aircraft. Such priority is exercised by being able to override the soft protections via control inputs.

@Winnerhofer - OK, we know AF has had a few issues in recent times, but I'm struggling to understand what you're trying to say by providing us that link.

@Winnerhofer
I was seeing that you replaced the unknown "URA"s by the "UAS"s (i.e. BEA and AF SOP). Much better !

But that text is not usable by pilots in flight and if you teach pilots on ground with that text about UAS they will start to write a book waiting 30seconds (Eternity) until the pitots ice is melted (perhaps never in case of failure) and systems fit again. You have to pilot your bird with pitch and power, that is "doing nothing" but as Machinbird said it "without keeping your hands in your pockets". Gums suggested the good delay for observation, appropriate decision, and starting correction without precipitation : 1 or 2 seconds...

That delay of 30 seconds appeared with A320 in some office. I was amazed since February 1988 that Pierre Baud accepted it.

SOP must be simple, but understanding the SOP during the time you do it must be simple too, as often a succession of factors and results happen in an emergency.

After some months gums pointed alt1 instead alt2b and came back to the C* definition, with limitation and rate limitations from fondamental pilot controlled parameters - pitch, bank, gee, power and their variations with help of reliable information like intertial data. More ordinary pilots cannot do more complicated than gums or Machinbird. Accurate pedagogy is a condition of survival in flight.

I don't think it was intended for that, I think it was an attempt at a general summary.

Right, I don't think anyone's suggesting anything different. I think what does need to be pointed out, though, is that outside of a significant mechanical failure or sudden change in the weather conditions being transited, the pitch and power at the point of AP disconnect will usually be more-or-less correct for the conditions, and one should only change it immediately if a significant problem is developing.

Sorry, need some more detail on this - what are you referring to?

As a rule, you don't really need to worry about systems functions at that level. The aircraft will still go where you point it.

Perhaps you can help me to find again that document in a ICAO magazine from january or february 1988, where Pierre Beaud was describing the strange innovations of the A320 : Push the red button 30s after the systems decide you have to learn piloting again, addition of both inputs, A320 doesn't stall. Etc. I no more remember the tittle and exact reference, and since ICAO does no more sell their documentation in Neuily/Seine I found it too difficult to find anything.where you point your HUD ! Free falling I learned "Body goes where eyes are looking"

Dozy, you continue to astound me with your blithe statements on subjects you have little practical knowledge of. That "academic" difference just might have been the difference between AF447 making in into the barn that day and taking that big swim. That sensitivity difference permitted initiation of a roll PIO from what I can see.

I am gonna go with 'bird and gang up on the Doze! Heh heh.

There is one helluva difference between the "direct" roll commands in Alt 2 and the roll implementation in Alt 1 ( thanks, 'bird..... I forgot the mode -- exxxxcuuuuseee me!)). My documents show something that looks like "stby gains" for Alt 2 according to configuration, but also states the control surface deflection is directly related to stick delection - full stick gives you full deflection and the "gains" move the surfaces at an appropriate rate ( just like the Viper did over ten years before, except our roll was still a body rate command and not an actual control surface deflection).

So @ Doze: Easier to get into a PIO in Alt 2 than Alt 1 or "normal". Doesn't excuse the "startle" factor we have discussed, if that's what happened, but going from a rate command with all the protections and such to a "direct" mode is not trivial. Feedback to the pilot is basically "seat of the pants" and your inner ear gyros. Not sure what the pounds per degree of stick deflection are, but the Viper was basically 16 - 17 pounds full pressure/rate command for roll ( my leading edge flap failure had me holding about 15 - 16 pounds for the 15 minutes I was airborne just to get zero roll rate in actuality).

Still appreciate your inputs, Doze, and wish I coulda been there withya in the sim and trying to recover the jet. I thot someone else tried a "recovery" from the "top" of the climb and entry to a deeply salled condition.

Just got back home after driving through WW1 battle areas, so I am going to stick my head above the parapet and duck back quickly!

I'm not at all convinced that it is fair to describe the lateral oscillations immediately following the AP drop out as being a PIO, although I can see why people should describe them as such. PIOs, technically speaking, have control inputs out of phase with the necessary corrective actions and the bank angle excursions should be increasing, not falling.

I think the point being missed in the discussions of 'sensitivity' and 'startle' is that there was a fundamental change in the piloting dynamics when the FCS reverted to direct law. IIRC in normal and Alt laws a sidestick movement commands a roll rate, but in direct law sidestick commands roll acceleration as in most aircraft.

Now I am not a pilot, but what I learned in studies of several aircraft is that to get stability the necessary control techniques are very different in the two cases.

What I see in AF447 is a pilot struggling to come to terms with an unfamiliar control dynamic; OK, it is one he had seen before throughout his early career and training, but not one which he had used recently. The (damped) oscillations result from his re-learning process and personally, I think he made a pretty good fist of it. I can't see them as a fault on his part, which describing the motion as a PIO somehow implies.

Sounds like a trip thru memory lane, OG

I don't think the "direct" roll implementaion is an "acceleration" command of the control surfaces, but actual control surface position. The "rate" of movement seems to be dependent upon the "gains" that the FCS uses according to aircraft configuration. This was same for the primitive FBW law we had in the Viper - "x" dynamic pressure according to gear up or gear down.

The "direct" command of aileron position is vastly different than commanding a roll rate or whatever. So at stick being neutral, the ailerons should move to whatever position the "neutral" position of the stick is using according to last trim condition.

Our force on the Viper stick was pretty simple, and just relaxing on the stick with zero force resulted in the jet rolling according to the last trimmed roll rate ( usually zero). You can see this on the Thunderbird solo pilots that do rolls to inverted passing in front of the crowd. The sucker flat stops rolling and no overshoot.

Just my experience and observations, but the 'bus FCS seems to implement the Viper laws except for the max bank angle.

Hi gums,

Guess I should have been a little more precise. Direct law does produce aileron movement proportional to stick movement as you say; but aileron movement produces a roll acceleration, which is why I wrote that stick movement commands a roll acceleration. That develops eventually into a roll rate which depends on aileron deflection, airspeed and the natural roll damping derivative. That means the characteristics are speed dependent whereas in the other laws you get the same roll rate for a given stick movement at any airspeed.

Putting the stick back to neutral in direct law will kill the roll rate not leave it at a commanded value as it does in normal/alternate. That is same as non-FBW aircraft and differs from your description of the Viper operations.

You remember right, however in this case the aircraft did not degrade to Direct Law, but to Alternate 2 (the equivalent of Alternate [no speed stability] on the A320 IIRC). In Alternate 1, roll is still rate-commanded, but in Alternate 2 the roll commands become direct. In Alt 2 the pitch commands are still rate-commanded, unlike Direct Law.

While I agree the dynamics definitely change in Alt 2 versus Normal Law, there are two caveats - namely that there was no immediate need to make significant roll inputs upon AP disconnect (and thus risk potential PIO), and that as the PF had no training in high-altitude manual handling at all it is conceivable that he may still have initially overcontrolled even if the roll aspect was still rate-driven.

EDIT : Though, as pointed out, I am not an aviation professional, I do however submit that this should at least be considered a matter of common sense - namely that if a malfunction (which will initially be of an unknown nature) has occurred, then it would be advisable to treat the flight controls very gingerly at first, because you don't know what works and what doesn't. I believe the phrase used in the mega-thread to describe the inputs made by the PF was "stirring the mayonnaise" - and this is something to be avoided absolutely unless there's an obvious and life-threatening reason to do so.

Indeed, most of the incidents and accidents I can think of in which the outcome was at least relatively successful have involved the crews concerned doing just that - "feeling out" the aircraft to see how it responds, what works and what doesn't. Examples include UA232, QF72, BA009 and NWA85 - not to mention the other 30-odd incidents of Thales AA-related UAS problems!

Yeah I knew that Dozy, but talking of a lateral oscillation the pitch mode is irrelevant and for all practical purposes my statement that the roll degraded to a direct law is correct even if you label it Alt2

Of course - I was just being irritatingly pedantic (as usual) for the benefit of thread-watchers who were new to the discussion. :ok:

Generally speaking, I think it's important not to get too hung up on the technical aspects of control laws, as some people's eyes tend to glaze over at the mention of them - while the specifics are definitely of interest to a technical audience, I think the best way of explaining them is in terms of more traditional setups.

Specifically, in real terms, all the control laws and their underlying mechanisms actually do is provide maximum controllability in the event of systems or mechanical failures - the technology means that this re-configuration is automatic, but ultimately it's no more than the equivalent of what a flight engineer would do in the event of a e.g. a dicky hydraulic pump or a stuck control cable on an older type.

The common thread remains, however, that the presence of a malfunction requires a gentle touch with the aircraft until the consequences of the malfunction on the aircraft's handling are understood and compensated for by the crew.

Welcome back from the fronts Owain.
The definitions of PIO have evolved over time and encompass a broader spectrum of oscillations.
The modern definition of PIO is Pilot Involved Oscillation and gets away from blame and fault concepts. Without the pilot in the control loop, the oscillation does not occur.

You can bet your money that Bonin was not rocking the wings for the fun of it. From his control inputs, I can conclude that he desperately wanted it to stop. Unfortunately, he chose a bad strategy to stop the oscillation and thus exacerbated his problem. If you look at the oscillation cycles, you will be able to see that there were a number of low points in the oscillation amplitudes followed by an increase in amplitude in the opposite phase to the immediately prior oscillation. To me, this is an indication that he attempted to get ahead of the cycle of oscillation.

To him, it must have seemed that he was a little too late with his control input, so he accelerated his control inputs, i.e. went faster. I think that you can see this pattern in aircraft behavior and in control behavior if you look at the timing on a cycle to cycle basis.

Too bad he didn't have prior experience in flying in Alt2 Law. He would have learned that all he had to do was stop making control inputs.

The development and implementation of Bonin's control strategy must have heavily dominated his thinking and almost certainly explains a lot of the otherwise inexplicably bad flying he demonstrated.

As pilots we learn to put our aircraft under accurate control. When we have an aircraft that is not behaving as we command, your whole attention is drawn to the problem, and it is difficult to leave sufficient attention for other important requirements of flying. The one very short roll PIO I experienced in a simulator was traumatic enough to leave a lasting impression. It compared to the feeling you get when you trip and begin to fall, and you are looking desperately for a hand hold. In my case, I was desperately looking around for a visual attitude reference that did not have a time lag in its display. I am relaying this so that the reader can get some idea of how compelling a PIO problem can become.

Originally posted by Machinbird

Which confirms my fear that I am thoroughly out of date on some things!

That is my point really. Control of the oscillations is all about getting the phasing of the control inputs right. Roll rate is 90 deg out of phase with roll acceleration, so if he was accustomed to a roll rate response to stick he could have been very confused by an aircraft that took a little longer to develop the response he was anticipating. That would be entirely consistent with your remarks

Sometimes getting control of the oscillations just requires the pilot to get out of the control loop, but this is not a one size fits all solution.

I can think of at least one PIO event where this course of action had immediate fatal results.(Sageburner).

Henri Poincaré feared to say that the solar system was not exactly periodic and could be unstable (in the traditional use of "PIO" divergence or non convergent oscillation).We know that periodic dynamic is governed both by phase and frequency, but traditionally lack of math tools focused on the only phase control which does not involve energy use at the contrary of frequence modification.
Another well known oscillation in airliners is passengers' hyperventilation. Frequense of breath is increased by different factors disturbing the CO2 and O2 regulation. Trying to correct it with phase has no chance of success.

Sometimes periodic movements are not more or less sinusoidal oscillations, I.e. chalk squeaking on the blackboard when pushed and doing rapid successions of jumps.

Progressively theory of chaos taught us that difficult art of controlling oscillations.
Multi variable problems are more difficult to describe in short words.

"Pilot" in PIO may be Autopilot too, as we see it in the KC-135R dutch roll report when the captain tried to stop it with A/P. The notion of "fault" is of no use in science. We are concerned by cause.

The human is a variable gain device when it comes to aircraft control systems.
For some type of input, you can get either a large or a small output and that output can have further modification as to phase angle or even anticipation of a signal.

The autopilot is a machine that is constrained by its construction and design to create a certain type of output for a particular input. Normally an oscillation involving an autopilot is not considered to be a PIO. As I understand the KC-135R accident however, the presence of human feet on the rudder pedals caused additional inputs to the system that resulted in an increasing Dutch Roll oscillation. If my understanding is correct, then that would be a PIO under the newer PIO definitions.

@Mac / Chris

IMHO the most perplexing of all the holes in the swiss cheese is the total absence of discussion given to the stall warnings by all 3 pilots. Was it a Titanic belief in the un-stallable nature of the airbus that the warning must be a glitch not even worthy of discussion?
Mac - if the patients heart stops during surgery and you hear an alarm I would hope that it gets someones attention.