Machinbird

Using BEA roll data for the period 2:10:07.5 through 2:10:45, the following graph describing the amplitude of the roll oscillation was prepared:


The amplitude is the total roll motion in degrees following a reversal of the roll rate. It is obvious that he was beginning to get a handle on his roll control problem when the aircraft stalled and created a new set of control problems.

HazelNuts39

Machinbird,

Is it?

grity

@machinbird


PIO
1 unintenditional YES
2 PF stears exact 180 deg out of phase
3 extension of the airplane’s operational usage into an area for which it was not intended, or following a failure, and is not the fault of the pilot
4 IMO .......software element upstream of a control effector.....???
5 divergent with time, after 2:10:07
6 number of cycles of oscillation ??? more than 5 ???
7 The most common frequency is in the range for pilot closed-loop control.... here it was 1/3Hz
8 1/3 Hz is not high frequency it was not "mild PIO"
9 roll ratchet...... >10 deg
10 Peak-to-peak angular rates are usually greater than ±10 degrees/sec, and control forces greater than ±10 lb, though rate limiting can attenuate the former and result in large increases in the latter.8 “Severe” PIO requires immediate changes to the airplane, and if it occurs in developmental testing the flight test program should be postponed or redirected until the corrections are made. YES we wait

Old Carthusian

This article seems to define the phenomenon as being anything that authors wish it to be. The evidence is shakey and in some places totally worthless and someone has clearly wasted money commissioning the article. Definitely part of the shoal.

Machinbird

I was expecting this type response. Typical hairy-chested "just fly the aircraft" comment of someone who has never encountered PIO because his aircraft experience involves highly damped flying machines.

My first knowledge of PIO was a news clip explaining how the Navy had lost a Sea Dart aircraft during a high speed flyby.
Later, while assigned to a Test and Evaluation facility, one of the engineers led me to the original full accident report on the Sageburner F-4 PIO accident.

I've personally never encountered PIO inflight, but have experienced roll PIO while flying an early visual simulator with a ~1/3 second visual display computational lag. Disconcerting to say the least. I had to go back to the flight instrument display to regain control. (Note: These were actual instruments-not on the visual display). Small motions did not trigger the PIO, but rapid-high gain lateral motions did. Until you experience the effect, you would have trouble believing it could happen. The onset is sudden. When it does happen, you tend to blame yourself because you are "in control."

The paper I referenced earlier has very little in new concepts. The new part involves the "pilot's mental model of the aircraft" wording. It does an excellent job of bringing together available information on the subject of PIO. The AF447 initial "roll problem" seems to fit the 10 PIO characteristics listed in the paper, but I'll let BEA make the determination, They have better versions of the data.

And for the conspiracy theorists, if BEA does not comment on roll PIO in any manner, either to affirm or deny, then you will have a red flag indicator.

Machinbird

Actually it is very close to 180 out most of the time until he gets it under control but it is converging, not diverging
If you look at the reversal intervals, PF starts out reaching his maximum control input shortly after reaching his roll max angle for a particular oscillation, and he is able to gradually gain on the oscillation so that he is actually leading the oscillation, at which point the amplitude starts to die down, only to have the oscillation flip phases on him. and go through another (smaller) cycle. At least, that is the way I read it.

ChristiaanJ

Machinbird,
Thanks for the link!
A bit laborious, as articles go.... but it covered the subject quite well for this ancient flight control engineer (Rtd.), complete with Bode plots and suchlike, which tell the story faster and clearer to those who can read them, rather than long verbal descriptions.

And for those who've never experienced PIO : 'hands up' for those who've never experienced DIO either (driver-induced oscillations, speed or direction) after hitting a couple of pot-holes or speed bumps.....
Been there, done that (luckily on local country roads, without disaster ensuing).

Sure, it's not the same. But it gives you an insight into how a vehicle you are confident of being in control of (near perfectly) at the time, can suddenly acquire a mind of its own....

Old Carthusian

Machinbird
I have stated earlier that I see merit in your suggestion but that I see it on a mechanistic level - an explanation of how the phenomenon worked but not a casual factor. This is my main objection to the paper you linked. It was too vague and too inclusive. The phenomenon wasn't adequately defined (it could be anything seemed to be the message) and it seemed to want to attribute an overly deterministic approach to a pilots actions. In other words the pilot was not responsible for his actions and a helpless victim of PIO. One is forced into the question why have pilots then? Are they truly unable to take decisions and judge a situation? I tend to believe that a pilot has responsibility for his/her actions and that a statement like 'is not the fault of the pilot' (Conclusion No. 3) is just as dangerous and misleading as 'is the fault of the pilot'.

We end up returning to the human and cultural factors of training, CRM, SOPs, chain of command. You mentioned you experienced PIO in the simulator and how you solved it by using your instruments. This seems to have been lacking from the PF - as I stated before know your machine and trust your instruments. It's a training and cockpit discipline issue and the paper seems to make no mention of this. You yourself mentioned training and I would add practice - skills atrophy without use. Once again one is forced back to the conclusion that this was a poor crew who took a perfectly flyable aircraft and put it into a non flyable state. Whether they were a one off or indicative of a broader trend is open to debate. Of course the crew were poorly served by their employer - Air France; and the Air France safety audit indicates serious cultural issues developing in their pilot corps. One hopes that Air France are addressing these as a matter of urgency.

If BEA do not reference PIO as a factor it won't be because they have an agenda but because they don't consider it germaine. That too is a legitimate judgment.

Machinbird

OC, lets try the following concept on for size. Don't worry how the situation occurred, just how the pilot handles it once discovered:

As you lift off the ground, the aircraft rolls the wrong way in response to your roll inputs. If the pilot is not able to override his natural tendency to correct a roll to port by moving the stick to the starboard side of the cockpit, is it then his fault that the aircraft crashes? Shouldn't he be able to rewire his brain on a moments notice to reverse the response? That is exactly the situation a pilot facing PIO encounters. His learned responses don't quite fit the situation any longer. If you encountered this type of situation, would you also be able to remember your aerobatic training to put the stick forward while inverted to keep from losing altitude? This is similar to the kind of control problem the AF447 PF may have faced (albeit a bit more extreme.)

The roll behavior after the autopilot dropped is extremely anomalous. If BEA does not address why it occurred, then they would be putting their heads in the sand. There would be a lot of people asking hard questions then.

This is the paragraph that you are objecting to. PIO really is a design problem and can be tailored out of the control system if known. If there was not a test point that stimulated the PIO tendency in that corner of the envelope, it may not have been uncovered during the flight test program and subsequent line flying. PIO is very dependent on pilot technique. For example, if a pilot is attempting to damp out a rate of motion, his control strategy is different than if he is attempting to hold an attitude.

Saying that PIO is not the pilot's fault is not the same as saying that PIO is not the pilot's problem. As you well know, when airborne, there is no one available to save you from your predicaments but you. By saying it is not the pilot's fault, the paper indicates that is up to the airframe maker and certifying authority to evaluate the problem and ensure necessary corrective action.

Old Carthusian

Machinbird
Sounds rather 'Camelish' (the said biplane used to pull to the right due to the torque of its Bentley engine). Are you saying that the control response is tardy and that the pilot's reactions are too quick? It has all the flavour of panic and forgetting how the machine actually flies. Those who have commented on flying the Airbus family particularly the A330 comment that it is a very stable aircraft in all aspects of its flight envelope. Therefore for your theory to work as a causal factor there would have to be a hidden design flaw which was not spotted in flight testing. I doubt this and fall back on the explanation that the PF (for some reason) lost it and overreacted terribly. The aircraft is going to respond to the pilot inputs and these are what caused the rolling - not the aircraft. A thorough rereading of the CVR transcript might help here. There is a disturbing lack of professionalism manifest throughout the sequence of events and the rolling is just a symptom of this.
What is missing from Conclusion No. 3 is the word 'can' because PIO can also result from a panicking pilot 'mayonnaise stirring' without the aircraft having any blame (as it were). However, I picked that conclusion out because it was an excellent illustration of the sloppy language used in the paper. What the PF needed to do in the AF447 situation was wait and analyse then act. He panicked and flopped around hopelessly and his colleague who could have taken command (as he seems to have understood the situation better) did nothing. No CRM, cockpit discipline, SOPs, chain of command. I keep on repeating this sorry to say but these factors are there for a reason - so that pilots do not get into messes when faced with unfamiliar situations. Simply a training and cultural issue - if you have the time take a look at Pan Am (late '60s early 70's) and Korean Airlines (1970s to early 2000s) for just how nasty these things can get.

Machinbird

OC,
Consider what percentage of the A330 flight time that has been accomplished has been in Normal Law, Now consider what percentage of that has been flown on Autopilot.

Then consider what percentage has been in Alt 1 Law and in Alt 1 Law with Autopilot.

Finally, consider what percentage has been in Alt 2 Law (Autopilot cannot engage)

Assuming manual control, does the aircraft feel the same way in Alt 1 Law as it does in Normal Law? I believe the answer is yes, just the protections operate differently.

Assuming manual control, does it feel the same way in Alt 2 Law as it does in Alt 1 Law? No. Roll response is different.

Other than the test flight program. I would be surprised if the aircraft has accumulated 10 hours of flight time in Alt 2 Law during its entire line flying lifetime.
But those that may know better should feel to correct any misperception on my part.

If I am correct, the bulk of A330 flight experience would not be relevant to AF447's loss of control in Alt 2 Law.

Old Carthusian

Machinbird
If you will forgive me the use of my favourite metaphor; that is a particularly juicy red herring. Just because most of the flying may be done in Normal Law doesn't mean that no flying is done in ALT2. It possibly is more than 10 hours but I don't know. However, be that as it may, a properly trained and conversant crew should have no trouble using the aircraft in ALT2 law. Those who have tried to fly it in such a state and who have posted say that it flies well. It was designed by Airbus to fly well - you can be sure of that. It would never have been certified if it was an absolute pig to fly without protections. Once again it's a training issue which should have been addressed in the simulator. If it wasn't then again it's a training issue. Pause, diagnose (using your instruments) then fly. But then I am afraid that this crew would have flown anything into the ground.

Machinbird

OC, I might share your faith in the reliability of the certification process except I have seen examples of the subtle ways things can sneak through, only to be found later, usually under adverse circumstances.
There is a reason that the referenced paper has this quote
PIO can be hard to reliably predict and prevent considering the various different control strategies employed by the pilots flying a particular type of aircraft. (But at least the autopilot should follow its programming in a reliable and predictable manner.)

Training can help pilots avoid unsafe control strategies, and can teach them what to do if they recognize a PIO condition. Has anyone received such training?

PIO recovery:

For example, in B type aircraft, the PNF can assist in breaking a PIO by helping hold the control wheel steady, but in newer A type aircraft, this is not possible.

In either type aircraft, PF can back out of the oscillation by loosening his grip on the controls, reducing the amplitude of his control inputs, and even releasing the controls completely until a PIO dampens out.

On either type of aircraft, exchanging roles between PF and PNF may change control strategy sufficiently to break a PIO. (This would have likely have been the best way to help in the AF447 situation).

Meanwhile I think we can collectively make some good use of your pet herring. I'll bet he's tasty.

Old Carthusian

“The design process of the airplane has matured, flight control systems have evolved, criteria and analysis techniques are available, yet PIO persists.”

Certainly it does but that is because it is pilot induced. It sounds a bit facetious - I am not trying to be so but point out that this is not nowadays a design issue so much as a human issue. There is nothing I can see in the accident which can be attributed to the aircraft. A different approach to training is probably the answer and proper CRM but the design oriented approach will not work in most cases because of the human factor. There will always be someone who finds a way to crash an aircraft in a totally unexpected way. In the end it still comes down to the man at the front.

jcjeant

You can train crew as many as you want
You can teach CRM as many as you want
Even if the crew of the AF447 was super trained to cope with all situations you can think about ... it was useless in their case .. cause they never recognized in which situation they were ...
They never understand what was happening .. so they were unable to apply any procedure .. learned or not ...
They were like children playing with buttons and levers .. and hopping something nice will happen ...

Machinbird

OC, as long as you hold that belief, you will have the PIO cause and PIO effect reversed. But you know that and in the big scheme of things it will probably make no difference- unless you encounter PIO yourself.

The training factors are significant in the failure of AF447's crew to recognize and declare a stall, but if they encountered roll PIO during the transition to ALT2 Law, that could be the reason that they did not/could not fly pitch and power successfully and lost track of their energy.

Remember this paragraph from the Aristotel report?
According to my decaying math skills, initial peak roll rate was on the order of 10 degrees per second for the first 4 oscillation half cycles but if someone would like to look this aspect over more closely, I would appreciate it. I can see evidence in the roll position record that PF's attention was heavily devoted to this problem. (Changing his strategy & adaptation.)

Old Carthusian

Machinbird

'throughout the flight, the movements of the elevators and the THS were consistent with the pilot’s input'
BEA Third Interim Report

It is important to recall that the pilot was the one who was making the inputs. This is where the Aristotel report falls down and badly in that it assumes that these inputs are a result of factors that can be designed out rather than factors that rest in human entities. One can almost say that it is written to the requirements of a pilots union. Rereading the 3rd Interim report we have a clueless PF and a PNF who doesn't feel he has the authority to intervene even though he seems to understand what is happening (as jcjeant succinctly notes). The more important document here is the Air France safety audit. It identified cultural issues that would lead to such a series of events. I'm not going to discount PIO entirely but it doesn't fit the circumstances of the accident and would not rate as a causal factor.

Machinbird

In other words, the motions of the control surfaces were consistent with the pilot's control inputs (i.e. no unexpected control system inputs)......but that says nothing of the dynamics of what the control inputs did to the aircraft. From the same report's data, it also appears to have set up a roll oscillation.

OC, What you do not seem to realize is that the Aristotel paper builds on a long history of professional thought on the causes of PIO. This is not a report that just "appeared out of the blue".

That is not an assumption that PIO can be designed out of a particular aircraft's control system-you can take that as a fact, but first it has to be recognized so that it can be addressed. Pilots who blame themselves for all aircraft handling problems are one of the reasons not all PIO problems are addressed.

I will concur that the crew of AF447 seemed to be clueless about a large body of fundamental aviation knowledge that night, but I suspect recognition of a PIO context for their initial manual control performance will explain how they lost control. It may even help explain why they failed to regain control, although the aircraft was then in a stall and no longer subject to PIO. But this area is one for the human factors experts.

grity

180 deg is the way you move on the swing set if you will hold the swing-amplitude, if you will stopp you try to move in 90 deg (or 270 ???)




if I put together HN´s vertical wind velocity and the roll altitude, then both cange with the same frequency of 1/3 Hz

we do not know much about the turbulent air in this case, but it is truly not one uniform mass which moves up and down, it might be that it was one vortex with the dimension d +/- 800m...... if they flew cross through it, the vertical wind velocity is explicable...

but what will be if they flew alongside of some vortex with this dimension and this frequency? changing different AoAs at both wings with 1/3 Hz..... VIO

the PF did not beginn with this oscillation, with the stick he just helped to hold the roll-amplitude.... (and forgot to control the pitch....)

DozyWannabe

I don't know if "forgot" is the right word. No matter how incorrect the pitch inputs were they were very positive, and way in excess of what should be expected at that altitude.

Even with protections out, the airframe itself is very stable - with 20/20 hindsight, the correct response should have been to understand what the turbulence was doing and very gingerly make small corrections once that was understood and a "feel" for the situation was gained. Instead it looks very much like every control input was reactive to each individual bump - possibly resulting in PIO in the lateral axis, but in my opinion there's not enough evidence to say one way or the other.

The main problem is that the response was very instinctive in nature, but not being trained in high-altitude handling the risk was that the response itself was likely to be inappropriate and excessive because the only manual handling the PF was used to was at low altitudes and commeasurately more extreme in terms of sidestick deflection.

As far as I'm concerned though, this is all by the by. The main problem was organisational in nature - Air France operating a type which was known to have potential issues requiring manual handling at cruise level (until fleet-wide repairs were effected), yet nevertheless rostering crews in such a manner that if such an event happened, the flight deck could potentially be manned by a crew without the required experience to recover correctly. That - based on the information we have so far - is the primary reason that control was lost and the aircraft crashed. The details of how that loss of control developed, while useful to understand, are at best a set of contributory factors.