BOAC

Mr Idle B - your plea appears to have gone un-noticed. Yes, there are many of us who understand the arrows and hooks and keeping 'rubber side down'. The problem we are facing here is that a particular system of flight (NB no names) appears to engender in some the chance to forget all this and become reliant on the system to look after them. I think you can be assured that many of these have gone away scratching their heads and may now be a little less relaxed about this and hopefully the manufacturers and training systems will also adjust. You will for example, note that the 'drill' taught for stall recovery has been changed following the accident and that I'm sure one large airline will be reviewing the abilities and training of its pilots. Looking at your location I would suggest you have a good chance of being behind someone 'wot knows'.

CONF iture

No.
Because once trimmed the airplane was looking quite comfortable in a stalled status.
But before it started to autotrim :

• It was definitely possible to hold the aircraft in the stall with 3 degrees of nose-up trim and full back stick, but it required effort
• The aircraft wanted to nose down and recover itself
• The nose wanted to come down naturally if I released pressure for even a split-second.

dixit DozyWannabe.

GarageYears

As one who has spent my entire career working on, designing and programming simulators (admittedly my "bit" is the sound and comms bit, just so you get the full picture), any assumption regarding a specific edge-case operation should not be trusted on the sim until verified and checked against aircraft operation OR double-checked against aircraft data (assuming doing it on the plane is either dangerous or really difficult). I routinely find issues on certified and in-service-for-years simulators that were not caught or identified during routine operation and certification testing.

While the manufacturer of any simulator will aim for 100% accuracy, the available data is routinely found lacking, hence my cautionary note above. What may seem surprising to some is the fact that the data package supplied by the airframe manufacturer will vary from one customer to another - I recently worked two 737 platforms for different customers, and received different data from each, for the same system. Not entirely different, but 5% was. Eventually I was able to get the same consistent data, but I was lucky since I had access to two sources.

Aside from those parts of the sim using rehosted aircraft code (which has it's own issues), the device trusted for training is the result of hundreds of thousand of lines of code from a motley collection of software engineers, that typically do not follow anywhere near the same rigid procedures and practices employed when writing aircraft-bound code (costs alone would not permit). This is not to say they are not doing a great job, but trusting the sim blindly, in particular in areas extrapolated beyond the aircraft data or figured out by analysis of airframer supplied data, warrant second and third opinions before assuming the behavior is correct.

- GY

Dani

This is certainly true, but BEA has tested every steering input on AF447 against the real aircraft parameter and has come to the conclusion that they are completly consistent. There was nothing wrong with the way the aircraft behaved like this, it was the steering input that was wrong, i.e. they pulled too long and pushed too short. Simple as that. If they had pushed as long and as hard as they had pulled, the THS trim would have gone certainly back to a normal position.

You are hitting the wrong sack if you want to blame the autotrim. Autotrim is a perfect tool. They shouldn't have pulled into the stall.

RetiredF4

Where does this leave us?

BEA only tells us, that the flight control inputs in the simulator produced comparable outputs to the flight controls and caused comparable flight behaviour. Nothing more, and nothing less.

To read into those words that the aircraft is sanctioned and freed from having contributed in some way to the outcome looks a bit bold. A systemic inbuilt problem like unwanted behaviour of the THS autotrim (i´m not saying that it is one or isn´t one) would show in the aircraft as well as in the simulation. Only malfunctioning systems in comparison to non malfunctioning systems in the simulation would show as difference.

Where does it leave us then?
It only proves, that concerning the flight control system the aircraft had no malfunctions and that another A330 with the same crew (or with a different crew performing the same inputs) at the same place in the same environment would have ended in the drink too. This recognition might cause more headache for a manufacturer than finding the cause in one faulted part.

franzl

Zorin_75

Maybe. But please don't disregard the THS arriving at the stops was preceded by almost a minute of ignoring a stall warning as well as what most would have considered basic airmanship...

Dani and GY were discussing the fidelity of simulators. Nothing more, and nothing less.

OK465

Quote:

 up to the exit from the flight envelope, the airplane’s longitudinal movements were consistent with the position of the flight control surfaces,

(my bold above)

So that although the movements of the control surfaces were consistent with pilot inputs throughout the entire event, there is no direct statement of finding to the effect that after the aircraft exited the flight envelope the aircraft longitudinal movements were still consistent with these pilot inputs even though control surface positions were. Nor does this appear to imply anything further.

I would imagine this is still being investigated.

Dani

this sentence does not mean that they don't know the correctness of flight control surfaces movements, but: You simply cannot predict the flight path in a stall anymore, the aircraft is not behaving "aerodynamically correct" or "like designed". It rather moves in a random pattern, depending on chaotic aerodynamic laws, not predictable and not rationally explainable (unless many more factors would be taken into play, in a much bigger "resolution" of values and mathematical formulae).

DozyWannabe

On the contrary, it did trim up, but it hit a limit of 3 degrees nose up. So we re-ran the test by winding in full nose-up trim manually and even with full nose-up trim the sim pitched down successfully on elevator deflection alone before the trim started rolling forward (although the autotrim did respond very quickly - I made a point of adding it into my scan once we started recovery).

There's a little-known coda to that accident, and that is the story of how the pilots successfully recovered the aircraft from the initial stall, but overcorrected by pulling up too long and caused a second stall that sealed their fate. What the investigators discovered in the sim was that if the pilots had simply let go of the control columns, the A310's protections would have stabilised and righted the aircraft on their own. Sadly, the Russian pilots were not trained on this feature of the A310's design.

I suspect that this is simply an acknowledgement of the fact that they can't replicate the behaviour of the actual aircraft in the stall without stalling the aircraft and measuring it (any test pilots want to volunteer?). By it's very definition, stall is a regime that is out of controlled flight and there are myriad forces acting on the airframe that are not easily modelled.

Machinbird

Dani, I would normally fully agree with your assessments, but the autotrim you know and love in Normal law is not the same autotrim in certain critical respects when in ALT2 law.

Ninety-nine point nine nine nine percent of your flight time or more has been in Normal Law. Just because your regular autotrim has been so sweet doesn't mean she does not have an ugly sister.

Lyman

Zorin75

The THS did not travel to (-) stop. It stopped short, something less than a degree. Neither did it "start" after one minute of SW.

Take a look back at Takata's pic of the Recovered THS. Note the damage to the NU end of the screw. Impact? Possibly. But the damage suggests collapse of the thread collar opposite the implied direction of travel at water impact. Something like an airload, not a water contact.

HazelNuts39

BEA tells us on p.41 of IR#3:
This could be a computer simulation rather than one in a fixed-base or moving-base flight simulator.
That statement presupposes knowledge of the aerodynamic characteristics outside of the envelope where these characteristics have been established by flight test, perhaps extrapolated using wind tunnel data. If these characteristics were known, it would not be particularly difficult to model them (at least for the longitudinal motion), and to use them in a simulation to produce your statement.

Dani

Maybe, I still don't see where autotrim was to blame in AF447's case. It did autotrim what the PF ordered. Autotrim has no artificial intelligence to know what's in a pilot's mind. It stupidly does what it has been told: When you pull for a long time it trims this position to zero force. That's the definition of trim. What's wrong about this function on AF447? If only he would have pushed the stick as long as he pulled it, autotrim would have ordered THS back, this I'm pretty sure.

Lyman

What is the purpose of "zero force"? Pilot feels nothing anyway, and the elevators can sustain and withstand the forces on their own. What the THS adds is lethargy when perhaps an abrupt input is necessary.

It's great for cruise. What is the need in ALTERNATE LAW?

Does the computer get sore muscles?

GarageYears

Zorin_75:
Correct - my point really was that jumping in the next available A320 or A330 simulator and wringing the cr*p out of it into an upset condition (outside of flight data) puts you entirely at the mercy and competency of the individual simulator engineer responsible for that aspect of the simulator performance - so in the case of flight modeling you might happen upon a model built with sensible extrapolations and cleanly interpolated behavior, while other models may simply 'straight-line' the performance once the flight test data runs out. It is unlikely such a flight regime was tested beyond a joy ride or unintended upset from the less capable engineers testing the sim - the last thing on a persons mind at that point is the accuracy of the simulation. The same would be true of any other aspect of the modeling, be it hydraulic systems, electrical or whatever.

So, taking the trim behavior in a simulator, once into an upset condition, with possible invalid airspeed, etc, probably shouldn't be regarded as definitive. I'm one of those engineers and I wouldn't trust it. As noted before, the sound models I produce would NOT include a vertical speed component as a contribution to the aerodynamic noise 'hiss' model. I have not data to model it and would be guessing. I could probably make a pretty good guess and have something representative, but:
(a) I've never been asked for it
and
(b) I'll bet if I ask 3 pilots for a subjective assessment, I'll get at least 4 opinions....

I suspect the Airbus "simulator" referred to in the interim report was an engineering simulator based on flight modeling the aircraft behavior versus control inputs, not what we all know as a full flight sim - perhaps the final report will define this more specifically.

- GY

RetiredF4

no, it trims not for zero force, it trims to maintain or achieve an ordered load factor.

As we know, autotrim not works on behalf of the SS inputs, but on the order of the FCPC, which has an ordered load factor of 1 g (no SS input) or a change (decrease or increase ) of this loadfactor (SS input).

Example (and this is not far away from AF447):
When in ALT2 and the aircraft is in a climbing trajectory (make it shallow or steep, the difference is only time) and the speed decreases (no autothrottle), the FCPC´s will position first the elevators and then the THS in order to maintain 1 g in the climb to compensate for the decreasing speed, even when SS is in neutral / untouched position.

Without preotections in ALT2 THS trim might even drive NU with a SS-ND input, if the increase in descent rate (like in the stall of AF447) decreases the loadfactor more than a small SS ND command would demand and elevators are already full NU.

Therefore machinebird got it right:
franzl

Mr Optimistic

In terms of 'stops', isn't it usual to have software stops somewhat in advance of any mechanical stops. Is there really a discrepancy ?

mm43

Are you implying a suspicion that Auto Trim was misbehaving in ALT2, i.e. the systems functioned in a degraded manner as per the XL Airways / Air NZ A320 accident?
I feel that NOT calling and clearing the ECAM messages and following the UAS QRH procedures introduced so much fog into the situation, that a failure to notice the Trim Wheel turning sits equally in this patch of fog.

In short, the Human Factors were dominate in this accident.

Dani

I don't agree. The whole AI's fbw system is made to maintain or achieve an ordered load factor. Not the trim sub system.

Remember: You give a sidestick input, that's the load factor you order. Now fbw computers calculate the necessary load factor. When you release your side stick (normal law), you maintain the ordered load factor.

What trim does is that it "neutralizes" long term load factor orders, so the flight control surfaces can go to neutral position.

Organfreak

Originally Posted by Organfreak
The Russian pilot who let his son 'steer' that A300 was 'qualified' too, and yet everybody died.
Dozy:
Oh, it was an A310? My mistake.
But I wouldn't call that detail "little-known." Hell, I saw it on Air Emergency on the idiot box just the other night! But, in their defense, they never knew what was wrong in the first place. I doubt they would have trusted (my speculation) the protections at that point. Did the A310 really have that feature?