GarageYears

Something to ponder?

The thread over the past few days seems to be fairly fixated on autotrim behavior and stick "feel" (or the lack of it, particularly w.r.t. trim). However I think (personal opinion, I'll take the flack) that in any control task, humans are amazingly adept at learning and understanding whatever set of controls we are given (subject to sufficient training/use/learning). What I see here is a collection of folk that have various backgrounds (F4, F16, Boeing this or that, and more) and I think it is fair to say each specific aircraft had unique "quirks". I see comments here related to the lack of trim response on the Airbus sidestick (unlike the yoke crew where trim unloads the force needed to maintain an attitude), but, 'hello!' the F-16 is no different... and in fact I'd state the F-16 is a good analog to the Airbus flight control system, except, more extremely, the F-16 sidestick moves almost not at all - I think the travel is something like an eighth of an inch (Gums?) total. Control is achieved not by waving the stick around but by applying pressure to it. Quite different. Similarly many of us will have played with various flight simulator games... while I suspect many here are gasping in horror at the mention of such 'toys' the point is even with significantly compromised controls (compared to the aircraft) many gamers become extremely adept at controlling the 'aircraft' and can fly the thing pretty much like the plane. What we do is we LEARN. Humans are good at this.

What's my point? The point is we can throw mud at the Airbus control system, but like or not, thousands of Airbus aircraft are trucking around the worlds airways with millions of passengers and they are not falling out the sky any more often that those from other manufacturers. The point is, the crews of those aircraft learn how they work and understand the behavior of trim, just the same way they learn the response rate to a stick input, and so on. Just because that control system is different to Boeing or Embraer or whatever, doesn't mean it is not as good. Clearly it is, since those planes are certified and make millions of flights daily.

You can argue as much as you like that changing this or that is the hail Mary for the aircraft type, but that just doesn't hold water. Does ANYONE seriously think Airbus is going to change the sidestick to provide force feedback for example? The stick currently is force 'loaded' (spring loaded I believe) and I don't really see the need to change that. But others have previously argued otherwise.

So, why'd AF447 fall out the sky, you might ask? Back to something I wrote earlier - "In any control task, humans are amazingly adept at learning and understanding whatever set of controls we are given (subject to sufficient training/use/learning)". This is what went wrong, not the aircraft design. For whatever reason (pitot icing), the automation dropped out, and as designed, handed the aircraft back to the most sophisticated computing devices on the aircraft - the humans. Unfortunately the most important safety device on the aircraft got it wrong. Unfortunately they got it wrong over and over. Zoom-climb (wrong), pulled into a stall (wrong), continued pull (wrong), TO/GA (wrong - but I'll pass on that since it seems this might have been a trained response), lack of CRM (wrong), and so on. Lot's of wrong here.

But what was really "wrong"? Training. Training. Training. With sufficient training the crews response clearly would/should have been different - very much so. Would it be different today - I suspect emphatically "yes". Because this accident has highlighted so many issues, any Airbus pilot worth his paycheck should have been following the accident investigation and learned a massive amount - mostly what not to do, but I'm sure a lot about the Airbus control systems. The key word in the previous sentence = LEARNED.

How did Gums learn to fly the F-16? Hands on the stick and throw the thing around the sky (some in simulators I hope... job security for me!), but NOT hundreds (or thousands) of hours tooling across the oceans on autopilot.... I see so many references to pilots with 11,000 hours (or whatever) and every time my immediate thought is "subtract ALL autopilot hours off that and what do you REALLY have???". Low hundreds perhaps, if we're lucky...

In my book this is the problem. The time when the crew were needed the most, was exactly when the automation took a timeout - and that corresponds exactly to the skill-set the crew has least ability.

So what am I advocating? More hand-flying on the line? Perhaps not - the autopilot is there for a good reason and does a great job. Simply holding the stick for hours transiting the Atlantic doesn't teach much. We need to get pilots into aerobatic aircraft and get their basic flying skills on the edge of the envelope up to snuff - recover from stalls, spins, etc. Then get those same crew into the training devices (simulators, flat-panel trainers, etc) and fail the automation and get their hands on the stick. Create situations where the control laws degrade. Learn Alt Law. A lot of sim time is spent training one engine out situations on takeoff, etc - are the current training scenarios really relevant today - how often do engines fail? I believe that LOC is the single most significant cause of accidents today - not engines failing/catching fire/etc.

Don't get me wrong - if there are simple changes to the systems within the Airbus cockpit that will help, they surely should be implemented (I suspect the stall warning inhibit below 60 knots should be revisited). But the single most significant change is reprogramming the HUMANS who get to sit in the very front seats.

- GY

airtren

BOAC

Mbird - can you expand on please? I am not sure I understand what you are saying and have a feeling I might not want to understand it either.

Gums - - you need to define marginal here, as where they were cruising on the carpet graph is where we all fly - day in day out, and is fine and safe..

gums

Salute!

Thanks for the nice words, Garage.

We had no simulator back in 1979. Didn't have one for another three years, and it was a POS.

We had a very simple FBW system with few "laws" or "protections" compared to the Airbus. Our "laws" were based upon maneuver limits and not "autopilot" functions like max bank angle, restricted pitch attitude, etc. Our "limits" were there to provide max performance while reducing the odds of a ham-fisted pilot getting into trouble.

Our autopilot was extremely limited in its authority. The GD flight control wizards didn't want something getting in the way of their system.

I see the reverse in the Airbus.

I do not want to see airline pilots "experimenting" with the limits of their jets, especially with 200 SLF's in the back. But I would hope that they would occasionally fly some planes that have classic stall characteristics and learn to cope with unusual conditions.

I had just arrived at Hill when we got the word that our neat jet, which could not stall or spin or....., could enter a "deep stall" due to the basic FBW design and the aft c.g. we used. So we saw the films and the test pilot interviews and we were PREPARED for when it happened to us!

We had little, if any, "feel" for mach buffet due to the wing design. So the Airbus crews had a leg up on us. But they also had a small envelope that allows for little error WRT mach and AoA. They had to be "better" than us in that regard.

My feeling is that better training will be the cure, and not a massive overhaul of the FBW system

airtren

I did pick up on..... At a quick reading, just didn't have time to confirm in a post. Am not sure how that counts in the flow of things, as those that need read them, and understand them are not those that are in agreement with you, like I learned unequivocally recently. Your contributions in the last several pages were masterful, but that bold last short sentence is a perfect engineering description, that I am most envy of. Thank you.
.

CONF iture

Automation did not drop enough : Leave the trim alone and we have a different game to play.
Then include it in the wrongs ... Why not ?

airtren

I think I grasp your delineation.

It may be just semantics, but "delaying", may practically mean "preventing it", when it's part of a complex set of causes of a final fatal outcome.

AlphaZuluRomeo

GarageYears,

I totally agree with your last post. As for myself, I concentrate on "small" technical problems (autotrim, stall warning) because I don't feel qualified to discuss -more important- training issues. Besides, here is the tech log.
The risk is of concentrating "only" on those relatively minor issues, and take them as an excuse to continue to "save costs" on the training...

Lonewolf_50

Yep.

It's been an expensive "lessons learned," hasn't it, this crash of AF 447?

It has cost dearly AF, 228 people and their families, and who knows how many others.

In its wake, it is likely that a good many AB flying folk (and organizations) have delved deeply into their machines and know them better now than they did on May 30, 2009.

To fly your aircraft you have to know your aircraft.

What was it we used to say in the Navy?

NATOPS is written in blood.

Looks like a few things have not changed yet ...

Lyman

From the git, AB has been marketing complexity as dependability, and "Ease of Operation" as simplicity.

Both are utter lies, and errant horse----. Training needs to start at the Line, and I don't mean the AIR Line. The Production LINE.

To require pampered pilots to inherit a wild thang in the worst of circumstances, without some history of mitigation/preparation, is manslaughter, imo.

My opinion, and we'll see what France thinks.

Wolf, I appreciate your experience, but flying the line is not Blue Water.
Nor should it be. These babies are not Ace Sixkiller, nor should they be.

They are Lambs for slaughter, 447 case in point.

Machinbird

AZR, I apologize for including a sentence you did not author. It seems the wonders of the Windows operating system require great care to avoid posting stray cat and dog comments in conjunction with those you intend.
If I clarify my definition of a barrier as an impediment that can be overcome with some degree of difficulty, I think you will see that a we are still in agreement regarding the import of the nearly full nose up trim.


No, I don't think they are going to change that anytime soon. My point in posting the lack of feedback in the stick is merely to explain the "open loop" control condition the PF faced once the aircraft stalled. There was no convenient feedback path for him to know what kind of control displacement he had actually requested. This was primarily due to his extremely rapid mayonnaise stirring control inputs. If he had made and held a control input, then the surface would eventually catch up to his demand the way it did in the nose up direction. The rapidity of his control inputs well exceeded the ability of the control surfaces to respond.

Quote:
Once the aircraft is stalled, the control inputs do not have much influence on the aircraft, and sometimes act in a contrary fashion.

BOAC, I am probably not telling you anything you don't already know. Once you stall you will likely find that your control inputs may have a reversed effect in the case of roll (due to adverse yaw), or that the surfaces are relatively ineffective and the aircraft's motions due to vortex shedding and cross channel aerodynamic coupling mask the effects of your control inputs. The elevator inputs are masked by the THS input and likely a post stall phugoid like effect results (a nose bobble). To have an influence, you need to make a control input and hold it or in the case of roll, you need to use the rudder and lay off the ailerons.

Clandestino

I might be wrong but that's not the way I see it. Yaw damper works as expected and only starts rudder-wagging when aeroplane is stalled and at low forward speed so chances are yaw damper was not compromised mechanically but rather by inefficiency of fuselage-blanketed rudder. Is there something I'm missing?

No can do. What you wrote is widespread and utterly wrong understanding of way Airbus FBW works in pitch in normal and alternate laws. Sidestick neutral iz not 1G it is 0G. Yes, you have read it correctly: sidestick neutral is zero gee. "You are pulling our legs!", you probably think. After all, whole lot of Airbus publications, FCOMs included, clearly state: "With the side stick at neutral, wings level, the system maintains 1g corrected for pitch attitude". It is so and it is true. So where is the catch?

Catch is that the sentence I've quoted is often understood to be the description of the principle on which Airbus FBW operates. It is not. It is the description of end result.

Sidestick command does not order G in absolute terms. It adds G demand to already measured, therefore if hit by updraft giving you 1.3G, pull on the stick that would give you 1.1 absolute from straight and level will now result in 1.4 pitch up. Push giving 0.9 would now be 1.2. Same goes for coordinated turn induced acceleration. In other words, stick G command is superimposed on measured level on normal acceleration. Why would anyone make so complicated flight controls system, Because...

...and that's exactly what you get with such setting. You might be commanding G instead of elevator movement but command sense is strictly conventional: stick down - nose down, stick up - nose up, lest gods of aerodynamics decide you have trespassed over AoAcrit and take away the lift from your wings, that is.

What our esteemed PPRuNe colleague has described as his own experience is what you would get if G command were absolute. In real life it is possible to achieve such a net result only with: severe malfunction of inertial reference, severe malfunction of flight control systems or severe turbulence. Until the time our honourable PPRuNe colleague decides to quit his incommunicado status and shed some more light on his story, I'll file it under "unreliable".

You, and I'm using "you" here in strictly plural sense, have gone wrong when you started believing very good sounding but flawed theories that resonated with your prejudices. Don't worry, it's basic human limitation.

No. We were lured into spending shed-loads of money on CPL training, believing that once we graduate, large pay for short work hours and lot of time in downroute hotels with attractive, young, available and free-minded hosties await us. To this end we have learnt all JAA thinks we need to know about aerodynamics, which can be summed up as: BBDEA - AACED - DAADC - ADEBA. Alas, there was no pot of gold at the end of the rainbow so lucky few of us slave away on pay that leaves one on half ration, after paying off the installment of training loan. You know what is the worst about it? I am not being as sarcastic as you presume I am.

Comfort can be always taken by scrolling the page all the way down and reading the big red script. Works like Prozac, if not better.

You have summed it up brilliantly: any A330 with same malfunction and same control inputs would do the same. Therefore, no mechanical or electronic surprises were present. Why would that cause more headache for manufacturer, beats me.

AF447 was the first A330 that achieved such a high AoA - basically it went into uncharted territory. That's why sim BEA's sim assessments stops short of going into extreme AoAs. You might theorize and test scale models ind wind tunnel until the cows come home, there's no replacement for testing the real thing to know whether all the theory translates smoothly into practice. Why no test were done at 40° AoA with real aeroplane? Dangerous. Expensive. Unnecessary.

I realize you were talking hypothetically. To set the record clear who might not understand: such a test will never be made.

There's always first time: Tarom A310, near Orly on 24 SEP 1994

I don't think I really need to repost pictures of A330 and Viper. One is passenger aeroplane, other is designed for combat. Every time my name came up on flight order involving A320, it was to move passenger & goods from A to B, if safely possible. Never was I ordered to strafe, bombard or intercept anything when strapped to A320 seat, which incidentally did not have rocket below seat pan or parachute packed in the headrest. Therefore, it is pretty safe to assume that design & certification criteria of two aforementioned superb machines (each in her own court) diverge wildly. Airbus logic is all about passenger transport. To repeat the lesson: stick free Airbus is flight path stable, not 1G chasing. As for G and AoA protections, principle is the same on F-16 and Airbus: full pull back in normal law will give you 2.5 G till AoA max is achieved.

Correct but pretty irrelevant to AF447. There was nose up moment from elevator. There was nose up moment from THS. There was nose up moment from underslung engines at high trust, yet the nose was mushing around 10°ANU. If the moment counteracting those wasn't pitch down of stalled wing, I really have no explanation what it could be.

Protections are lost because there is no simple way to compute whether airspeed or AoA is wrong. Stall warning remains, as the detection of aeroplane's energy state and associated decisions now is unloaded on intelligent entity, which must determine whether warning is true or false.

Correct.

That would be very damning, if found true. Proper way to perform instrument flying in civil aeroplane is by visual reference to instruments, not to column/stick position. Taught from day one of IR training. That's why no one made a fuss about non-backdriven sticks on Airbus. At least no one not anonymous.

If they relied more on their hearing and sight, and less on their feelings, supposing they relied on anything at all, the outcome could have easily been different.

Way around. It was not trained because of low probability but because it was believed that proper training in: aeroplane energy management, dealing with approach to stall and good aeroplane's stalling characteristic (for the cases where crew really needed time to gather their wits) would make occurence of extreme AoA stall in a passenger transport aeroplane so unlikely as to be unworthy of consideration. Both pilots of two man crew getting so confused to do almost everything wrong was beyond scope.

It does not. Stick forces notwithstanding, Airbus is pretty classic about airplane-pilot coupling. First time I've heard the phrase "stick-stirring", it was not related to Airbus but Let L-13 Blanik.


While I might agree, I'd advise caution not to slip into conjecture. First 3 reports are heavy on technical side while HF side is seemingly neglected, which comes as no surprise to me as it is much more difficult and time consuming part of the investigation. Human are not machines, which sometimes comes as mixed blessing. While aeroplane doesn't care what time of day it is, human beings tend to perform better at 10:00 AM than 04:00.

Machinbird

Quote:
Originally Posted by Machinbird
The feedback channel he had used all his flying career was no longer available. He was operating open loop.

Not really. This is really a Human Factors type evaluation of what the PF had to work with while using his mayonnaise stirring stick technique, and only applies while in a stall. He is not supposed to be in a stall, but it appears that mayonnaise stirring is a particularly bad control technique while in a stall. That might be a lesson we can take from this.

I'll bet there are quite a few engineers that do not agree.

Quote:
Originally Posted by Machinbird
A crew not recognizing a stall! That just should not happen. Particularly with the amount of time they had at their disposal to recover.

Quote:
Originally Posted by GarageYears
But what was really "wrong"? Training. Training. Training

Even the BEA seems to hold the viewpoint espoused by Garage Years and myself:
.

I've been trying to pin down what specific training in my background convinces me that I would recognize the stall in the same situation that the AF447 crew faced.
The two factors are:

1. Having read D. P. Davies description of the deceptive nature of the stall in a relatively level attitude.
2. Having spent time maneuvering swept wing aircraft to their performance limits. (This has already saved my posterior on more than one occasion. Once was in a puddle jumper.)

rudderrudderrat

Hi Machinbird,
I was wondering the same thing myself. The Airbus Flight Training Study Guide (2003) had the following recommendation when in ALT LAW and approaching the stall: (see page 13) http://www.737ng.co.uk/a320training.pdf
"Eventually, the master warning and aural warnings will activate (crickets and “STALL, STALL” ). Recover at the stall warning by selecting TOGA thrust, maintain a pitch attitude for level flight and accelerate through VLS."

Unfortunately, AF 447 crew discovered that it doesn't work at FL 350.
(They probably had it demonstrated at 5,000 ft in their sim conversion course.)

HazelNuts39

Thank you for your elucidating explanation/correction. I'm amazed how one continues to learn new 'secrets' of the control system after so much time. What you write makes sense to me, but leaves me with a couple of questions:
I've read somewhere that full back side stick corresponds to a demand of 2.5 g. Suppose you are at 1.3 g with neutral side stick, then pull the stick to the back stop. What do you get?
Then we have this description in mm43's post #366: "With STICK FREE in turbulence, small deviations do occur on the flight path but with a tendancy of the A/C to regain a steady condion". Does that functionality only exist with stick free?
Agreed. I've edited the last sentence of my post.

Clandestino

That technique, while patently wrong, would not be lethal if center of the stirring movements were set around neutral or moved forward as ADIs have shown pitch increasing. Problem is that average input was heavy nose-up. That's not ham-fistedness. That's confusion.

They even did not try to set attitude for level flight. How could they discover it wasn't working?

Big part of being pilot is to be able to tell the difference between 5000 ft and FL330. Consequent to that is to know how aeroplane behaves at different altitudes or at least which altitude dependent procedure to apply.

G protection comes into play. 2.5 G slats retracted, 2G slats extended. As for how exactly G demand is related to stick position: me knows not. It felt linear and progressive in the area I have ever needed to use, which was certainly never above half travel, and that was good enough for me.

With stick in pitch neutral; yes. If you move it out of neutral you command flight path change so nose might be bobbing but overall it will move in the commanded direction.

Now you reminded me I have to further qualify my statement that Airbus is not particularly prone to aircraft-pilot coupling; it is valid in still air. In turbulence, tendency of the airplane to self-correct pitch and bank disturbances may (and too often does) lead to pilot induced oscillation. Statement you quoted is mighty correct and is followed by very good advice to avoid large interventions with stick.

infrequentflyer789

At what point did they try maintaining pitch attitude for level flight ? They (or at least PF) only tried pulling up. And they knew it. As someone said earlier on thread: PF responded, "But I have been climbing for some while...."

infrequentflyer789

First:

If this and other LOCs were down to pilots who could handle a conventional aircraft perfectly well in the same circumstances but were constrained/prevented/confused by the airbus flight control laws, then I'd agree - ditch Alt laws and drop straight to direct. Give the crew the conventional aircraft if/when things start going wrong.

...but I don't think the above is a correct assumption. There's a subset of pilots (probably including those who care enough to follow accident threads on here) who would handle direct law just fine (or better), but is that the majority when I also see comment after comment along the lines of Machinbird's "compare what we learned about actual aircraft handling compared to what is presently being taught in the puppy mills, it is night and day".

Second:

I don't think stopping autotrim would have affected this accident. I know you've argued that had the nose gone down they'd have diagnosed the stall, but I'm not so sure. When the a/c did pitch down in stall what was the reaction ? Pull-up, hard. If you're already prepared and briefed for stall in the sim, the nose drop is going to be obvious, but if you think the a/c isn't responding right (which looks like an issue in roll at least even before the stall) and you're pulling back and the a/c suddenly drops the nose, what will you do ?

Third:

Changing something one way to "fix" one accident may make things worse in other cases and end up killing more people. We should look at all the LOC incident history not just this one. I've done some looking, though not in any way a systematic survey, in what I've read the common factors were:

Airbus: no. FBW: no. Sidestick: no. Trim / AutoTrim: oh yes. Time after time. Not every incident, but probably a majority - and much more common factor than the others.

But what is the typical problem with trim, what's its MO when it kills ? Looks to me like it's trim-up before stall, autos drop out, trim not managed by crew then contributing to the upset and/or preventing the recovery. Exactly the opposite of 447 - which looks like the odd one out.

chrisN

IF789, in the sample of accidents you looked at, as well as finding a common factor with autotrim, do you see a common factor in what has been identified as the problem by more than one expert above – to quote, training, training, training - ?

Chris N

infrequentflyer789

Maybe not, but I can't provide the reassurance asked for. I raised similar concerns myself back here and triggered a well written response from PJ2 here

Could be a lot of reasons for that. PJ2's comments referenced above; increased auto-everything; SPOs and regulation (RVSM) that all but prohibit hand flying; and Glass.

That last one's a huge difference for the non-pilot looking in (and maybe for the trainee starting out...). Look in at the old style hundreds of steam gauges and row after row of switches and be awed, and fearful - you better learn and understand every one of those, this is complicated. Look instead these days at three computer screens a handful of switches and (maybe) a joystick - well, how hard can it be ?

Yet the fundamentals of flight haven't changed, the machine underneath is as complex as ever, if not more so, and the margin for error doing a few hundred knots in an aluminium can at 30k ft is still just as small...