jcjeant

As reminder (concern October 7 2008 incident A330 airbus type) the ATSB report released some days ago is available here:
http://www.atsb.gov.au/media/3532398/ao2008070.pdf
Seem's computers are the culprits

RR_NDB

Hi,

Incorrect data from a sensor measuring airspeed, altitude, air pressure, temperature and the flying angle was fed to the computers controlling the flight, investigators found.

Garbage in, Garbage out...

PS

More and more we are dependent on complex Systems. A lot of skill is necessary to act timely and precisely when things go south.

PS2

From page 215 of the ATSB Safety Report:

It is very likely that...Despite extensive testing...the exact origins of the failure mode could not be determined.

Data spikes

Old Carthusian

Machinbird
My apologies for the misunderstanding
My comment was not meant to attribute a judgement on the PF or other flight crew's personal characteristics but on the flight deck performance. For this I hold Air France culpable. Their policies and culture and lack of training set up this flight crew ill-equipped for the situation they faced and the poor individuals were out of their depth. I rather regard this as a tragedy for not only the passengers but for the flight crew as well. They too were let down.

Machinbird

OC, No apology needed since you have clarified your intent.

I agree with your assessment of their training readiness for the events of that night.
They were not ready, that is plain.

If you will observe the roll PIO video that Organfreak posted. Once the initial overcontrol input was made, there was no way to promptly stop the oscillation. In fact what finally appears to have stopped the oscillation was likely intervention by the other crew member, probably by freezing the yoke in roll, but it is also possible that PF changed his strategy. It is hard to promptly undo this type of overcontrol mistake. That energy has to be dissipated.

I think something comparable happened on AF447. Those big spoilers popping up alternately are able to generate a lot of roll momentum quickly. But on the FBW 'Bus, PNF cannot freeze the stick, he must take the initiative and take control.

RR_NDB

That energy has to be dissipated.

An important reason to refrain from applying high amplitude stimuli.

Reading Learmonth ATSB report it comes to my mind a question:

Possible oscillation (MB), Possible Data Spike (difficult to trace as per ATSB), WRG fault (as per ACARS/Svarin emphasis), RHS not recorded (as per BEA), Persistent NU from PF (at RHS), "combined" may led to an unexplained (AF447) case in the end?

DozyWannabe

Alternatively, the final touchdown resulted in a natural switch from lateral movements to pitch - keeping the nose down by pushing forward. I know I'm a critic of the "No pilot would ever..." statement, but the fact is that continung lateral input while pushing forward would be incredibly unnatural.

Organfreak

I am told that what actually saved the landing was Ms. Bertha Pound, a purchaser of two tickets for herself, was standing to reach the overhead bin and suddenly fell down against the window, righting the aircraft.

roulishollandais

Thank You MachinBird to come on the PIO .

I elaborated in 1992 a method to get out from a DUTCH ROLL, and demonstrated it five times ago, on a MD83 simulator (Finnair, Helsinki, dec 1992...).
Nobody understood how I did that (counting seconds precisly with my free-fall experience, and resolving two differential equations by head computing the "response" I had to fly, and did it with full success.
That is much more that what a pilot learns to fly. But a dutch roll who is not finished near of the ground is extremely dangerous, and I attribute some of the past crashes (I did not find them in the Aistotel data base) to this.
With my method it has been possible to get out the dutch roll, in less then 30 seconds, with loss of height less then thousand feet, it is then possible,falling in the dutch roll at the outer marker, to go around and come back safely. It is quite possible to do that without to have the mind overloaded with the new task. The captain who tried to come out the similar dutch roll, had 6000 hrs on the plane, and going on the back right and left, struggled and lost 11 000 FT to bring the wings level and stable .

The dutch roll is a severe PIO in MachinBird/Aristotel classification.

I agree totally with JCJEANT that without enough formation, the first thing is "HANDS UP". But a minimum training must be given to any professionnal pilot to be able to idendify the problem (is it a stall, a spin, a dutch roll, a deepstall, a lazy eight, a dynamic looping, aso.. Only next is it possible to learn a better method that Ŧ hands up ŧ to correct the situation . This method does not concern a unexperienced pilot landing a C152 (Aristotel !).

I agree with MachinBird that it is not enough to "just fly", even for a very good pilot. Flying must be adapted to the dynamic situation. I already posted about the fact that dutch roll is not a problem of ge/speed, but of what doing what at which moment, how long, aso.


How does the dutch roll happen ? (I shall not explain here the total theory of coming out of a dutch roll, only main ideas). It is a frequency resonance (f2=n.f1 ; n=1,2,3,4,5,etc.) between the normal action of the pilot (is he volounteer if nobody learned him that ?!!!) and the natural roll stability of the aircraft.

a- Some aircrafts have this aerodynamic problem, the manufacturer has to correct, it comes sometime on little details.

b - The yaw damper can fail (leak of hydraulic liquid on the electronic box, modifying the electric characteristics, or mechanical failure inside the yaw damper) (failure or sabotage)

c - Modification of the aircraft characteristics due to bad, or insuffisant maintenance.

d - Modification of the aircraft characteristics due to bad flying, without respects of the speed limitation (for example >250 KTS under FL 100 for some aircrafts) repeated day after day...

e - Bad certification, without fully verification that ALL tests have been done, with positive proofs of acceptability for all them

f - To engage the p/a may be enough to start it too (classical or FBW aircraft), and disengage the p/a may be enough in a FBW aircraft. The use of the a/p On or OFF must be done ready to surprise.

g - For ANY AIRCRAFT, so stable he is, hard turbulence can start an oscillation, and the PILOT achives that catastrophic.

h - Last but not least, for ANY AIRCRAFT, ICING SUDDENLY, or ICE GOING AWAY SUDDENLY, may significantly modify the aerodynamic caracteristics and begin the furious mouvement with or without the help of the pilot !



After 1992 I tried to explain that to my pilot colleagues, but NOBODY in my french aeronautical community including administration of civil aviation wanted to listen anithing about that, rejecting strangely mathematics...
.
I wrote to the french BEA, and got no response
From 199 to 1998 nobody asked me how I had done, or to teach my method….and i turned my back , desesperating about flight safety.
In 1998, I got informed that a friend of Michel Baroin died in 1987 as passenger with him in his Learjet crash (not in the Aristotel data base). I had been his flight instructor and what I did read from the crash made me suspect a dutch roll finishing in a deepstall. I decided to start again to speak about coming out from dutch roll. I wrote a book… but found no editor .
Worse, I phoned with the Learjet Instructor in Genčve (CH) ... and he explained that they (he ?) had found their own "method" to try to come out from the funest dutch roll : "Faire des cisaillements sur le palonier" (sic) (Quickly shearing pedals) !!!
This extremely wrong and dangerous "method" has been used by the copilot on the flight AA587 in the wake turbulence over the Queens : THE RUDDER BROKE (Nov 12. 2001).
The NTSB started after some time a public enquiry, and found terrifying facts, about flying, and maintenance of airliners and cracks in the fin. Some newspapers wrote about that, from New York and London : I traduced them from english to french, and the second times I showed them, I got death threats... Aerodynamic is really dangerous !

Reading the Aristotel document I am effraid to discover we are still not ready to see the end of the PIO or PAO : Why do they mix helicopter and fixed wing aircraft ? Equations are different ; solutions also. Beautiful categorisation from 1 to 8, but it is just a normalisation. Why do they mix pitch and roll PIO ? Solutions are different. Coming with the phase is a bad entry : without frequency resonance the phases will be different at the next cycle (and no PIO). Aristotel says the first element for an aircraft pilot coupling (APC) is a unfavorable vehicle dynamics : that is not true. A stable aircraft like A330 seems to be, in some specific environment may get in a APC. The Ŧ mental mismatch of the pilot ŧ ? generaly the pilot flies same manner as he has learned to fly WELL, but this manner is not adapted to the situation. Finaly the close loop wood be for Aristotel the third key element : When we fly an aircraft we are always in a CLOSE LOOP. Pilots have to learn that and understand that. Aerobatics was much better to learn that, than 200 hours on C177RG or Trinidad ! Aristotel uses four categories of APCs : 1-2 linear and quasilinear ; 3.transitions in the aerodynamic configurations 4. Elastic structural modes or biodynamical couplings : But oscillation is not linear… Air is not elastic … And fly training is not a bioreaction but a Pavlov repetition ; stop with psychologising human factor : the main human factor in dynamic systems, is that a human is essentially a Ŧ time constant ŧ !


As the AF447 fin was found by the Brezilian in one piece I already imagined the AF447 fin has broken (see my first pprune's post) like the AA587's one, and I am wainting due arguments from the BEA that it did not happened.


However I often looked BEA#3' DFDR graphs, and oscillations : I cannot find the dutch roll regularity, but that does not mean it was not : We still are unknowing the detailed Airbus software design (management of interruptions, time constants, damping coefficients, system architectur, size of piles, Bode diagrams, aso). (We also do not know the exact description of the flight laws : industrial secret ! We learned here that even AB simulators ,used to qualify the pilots, do not use flight laws but discrete description of the aircraft coming from test flights. But in our endless question about the crew conversation we find no word of wrong heading, despite it changed... It is impossible having 10, 100, 1000 or 10000 hours on the logbook to say not a word of this change . The BEA'silence, once more does us doubtus : Was is a pilot in the aircraft ? If the tail broked, due to a crak or misflying, with or without dutch roll, had they an explosive depressurization ? I was wondering too the oscillations in what seems to be an unstable deeptstall : strange, but possible if the fin has already started to fly away.



I am really very very very happy to read MB today ! Now we are doing aeronautics

grity

what did you think is the frequency for the natural roll stability for a330 at this high and speed?

RR_NDB

Now we are doing aeronautics

A rich and long lived thread, highly motivating many of us.

Question: When (probably) VS separated? There are differences to the AA587 fractured VS as you did see in the released pictures.

When you have an insight* on something most are unprepared for, you pay the price. I "saw this movie" some times.

(*) I elaborated in 1992 a method to get out from a DUTCH ROLL, and demonstrated it five times ago, on a MD83 simulator (Finnair, Helsinki, dec 1992...).
Nobody understood how I did that (counting seconds precisly with my free-fall experience, and resolving two differential equations by head computing the "response" I had to fly, and did it with full success.

DozyWannabe

Not exactly - the vertical stabiliser separated inflight... *only* after being subjected to stresses beyond the Ultimate Design Load of the structure.

Neither cracks nor the repair to that A300's empennage had anything to do with the separation, which happened considerably beyond the Ultimate Design Load of the vertical stabiliser.

Using the rudder is not in and of itself dangerous, but rapid stop-to stop movement in a sideslip at high speed is, especially in an aircraft with wing-mounted engines and a large rudder to compensate. The problem with the AA case was that the First Officer was attempting a technique designed for the DC-9 series (small rudder, fuselage-mounted engines) on an A300 (large rudder, wing/pod-mounted engines). In the A320 sim using the rudder was better for ironing out the roll, but that was only performed after the speed had decayed past a certain point.



Forgive me for being a little sceptical, but from whom were these received?

On impact with the ocean surface. Unlike AA587, the AF447 vertical stabiliser had a large chunk of fuselage attached to it.

jcjeant

Indeed .. These cracks are not (according to the report) a contributory factor
However the NTSB revealed by examination of other aircraft in the same series .. cracks that existed and had not been detected by maintenance services .. as their means of investigation were not suitable
It is clear that this finding by the NTSB and the development of new means of investigation has certainly allowed to prevent accidents from occurring in the future

Methink that the point of "roulishollandais" was about "wrong method" and this can be dangerous .. as it was show by the A300 case (obviously the pilot used the rudder control in a way that it has exceeded the certification of resistance of the materials concerned .. but if he had not used this method .. the accident would not have happen)

Indeed .. again ...
Nevertheless this does not prove that the vertical stabilizer separated from the plane in contact with water or when it was in flight
Other elements of the survey suggest a near certainty that the vertical stabilizer separated from the fuselage on impact

Machinbird

Dutch roll is a different oscillation than PIO is. Dutch roll is one of the natural oscillation modes of an aircraft in flight. PIO is an oscillation that results from a pilot's attempts to control an aircraft.

To help understand the difference, take a look at this Wiki post Dutch roll - Wikipedia, the free encyclopedia, Then compare with the Aristotel definition. This is the reason Aristotel does not describe Dutch roll as a PIO characteristic, and why the Aristotel database does not have accidents listed that result solely from Dutch roll.

I've seen no evidence in the record that AF447 suffered from Dutch Roll. The A330 aircraft has an independent BYDU unit to provide yaw damping. Normally yaw damping alone is sufficient to prevent Dutch roll, although one jet I flew had "stability augmentation" on all 3 axes (and flew the same way a drunk walks when it was switched off.)

RetiredF4

Iīm with you on all you wrote concerning PIO and itīs possible influence. I had my share of it once during airrefueling on a KC-135 in pitch dark night just prior contact flying from the backseat during instructor training.

And i can confirm your last sentence ......."and flew the same way a drunk walks when it was switched off" .......as we probably think about the very same aircraft.

Some reference to the Backup Yaw Damper Unit
:

No "turn coordination" and hence no sideslip compensation most probably was contributing to the aparent problems in lateral control in the phase from AP dropout/ALT2 Law to stall entry and later on in the stalled situation. The FDR on Page 106 of BEA IR3 shows the graphs for Sideslip, and on page 112 itīs together with aileron input.

HazelNuts39

RetiredF4,
Would he really need turn coördination when he is not turning?

RetiredF4

The aircraft stays flyable without it, but that is not the question here.

The PF was used to it, we can assume that he never flew without it in a real aircraft. Any aileron input causes increasing drag on one wing hence sideslip, and any bankangle causes change of lift. It is a fine system that you donīt have to even think about those changes because the aircraft systems conmpensate for it on their own.

Without this compensation the lateral SS inputs causes sideslip and any bankangle change causes change of lift. Thatīs why you use some rudder and some nose up input on old aircraft without those systems when turning.

Therefore it should feel different to the normal everyday smooth riding especially with those big inputs made by the PF without autopilot and in roll direct (Alt 2).

We humans tend to feel uneasy with an input, if the desired result is not what we exspect. Look at car drivers on slippery roads, if unexpierienced they slam on the brakes if the car picks up a slight drift and agrevate the situation.

gums

yeah, Retired, Mach, Wolf, Smilin', et al.

I can't figure out how much aileron-rudder-interconnect (ARI) there is in the various "alternate" laws from looking at my skimpy 'bus manuals. Maybe Retired or A33 can enlighten us?

The Normal law appears to have not only a great ARI function for rolling into a turn, and also has an appropriate pitch input to make a level turn if stick pitch input is neutral.

Same question about the "damper" functions. I thot the 'bus retained the dampers regardless of mode, even "direct". This would resemble what we fighter pukes flew with since the 60's. And it definitely helped mitigate PIO's and other bad things.

Viper dampers NEVER cut out, but ARI cut out when one main gear WOW switch was compressed, hence no "wing low" landing, you landed in a crab. Otherwise, the rudder position that had been manually commanded for the slip went off the wall upon touchdown ( and we tested this for ourselves, heh heh). You don't need air data or inertial system data for the function, just some rate gyros embedded in the flight control computers via hardware/firmware or independent rate/gee sensors feeding the computer(s). In other words, we had no need to have a "basic" computer to filter and use all the external inputs, like the ADRIU the 'bus seems to rely upon.

I am not sure we saw but a glimpse of PIO with respect to the roll commands, so won't join that opinion. Same for inadvertant pitch inputs, as looks like they were intentional because "the aircraft won't let you stall" mentality.

DozyWannabe

Absolutely - maybe I should have clarified by prefixing "among other things", chief among which is the fact that the flight path is consistent in every simulation the BEA has run with the vertical stabiliser remaining attached until impact and final breakup.

But sometimes it's more fun to make the "weak tail" conspiracy buffs go and spend a few hours trawling Google Images to understand for themselves...

Machinbird

And it is still probably more complicated than that with rapid reversing control inputs. If the response time of ailerons and spoilers are not perfectly matched while transitioning, you might be off the planned schedule and creating additional yaw.

Normally, with a brief interval of mismatch, this is no big deal. However, when the controls are continually transitioning as in the case of a roll PIO, the percentage of time that this potential imbalance applies is greater and may be the source of yaw inputs to the airframe.

Having flown jet aircraft with aileron only roll control, spoiler only roll control, and combination aileron spoiler-aileron roll control, I prefer the combination approach, but spoilers can create cruise inefficiency with a heavy wing. For that reason, spoilers often only move with larger roll inputs, and smaller inputs are handled with the ailerons only. I do not know how the A330 gears the relationship between ailerons and spoilers.

roulishollandais

I don't forget you. ... Christmas !!