HazelNuts39

For whom it may interest: a crossplot of normal load factor (gee) against AoA at entry into stall from the graph on page 45 of the french report #3 - LF vs AoA at stall entry.

GarageYears

It may be pertinent to the discussion to note that Level D full flight simulators do NOT include sound cues for:

- Ice crystals (hail, yes, one size fits all generally, but smaller ice crystals no)

- Air impact cues from anything other than CAS/Mach related airspeed (i.e. any aircraft velocity that is not horizontal in nature will likely not cause any air impact noise cues.

So an aircraft dropping at 10,000ft/min vertically with at 0 knot CAS would be totally silent in the simulator....

Lonewolf_50

Linktrained:
Yes, descending into a cell is something we are taught not to do in early metro classes.

GerardC: You make some excellent crew points in re "age" not being the end all be all of who is a better choice to have the deck and the con. In regards the assignment of roles and responsibilities, if the Captain tells one pilot "you are in charge" that does not change what sound CRM principles call for. But yes, there is potential for CRM problems.
http://www.pprune.org/tech-log/45687...ml#post6621097

If you are flying the FD, rather than the attitude indicator (is it either or, or is it a blended scan?) that might cause problems if it appears and disappears intermittently.

safetypee: Amen
http://www.pprune.org/tech-log/45687...ml#post6621211

But ISIS doesn't talk to Prims -- am I right about that?

A33Zab

The thrust changes are part of the normal and longitudinal acceleration
experienced by the accelerometers and refelected by the corresponding graphs.

The -normal accell- (Nz) graph is composed of flight control elevator deflections;
THS trim; roll influences; speedbrake and thrust-changes but also environmental disturbences.

For thrust-changes:

If ∆Nz is the difference between the actual Nz and Nz(1g) and the goal is to maintain a 1g flight:
∆Nz = positive above 1g
∆Nz = negative below 1g
At decreasing thrust and a positive ∆Nz this will result in a less ∆Nz to be
compensated for with elevator.
If the ∆Nz is negative (Nz below 1g) this will result in a higher ∆Nz to be compensated for.

Off course it's the other way around with increasing thrust.

Hard to explain when english is not the native, if anyone feels the need to correct and/or make additions he is most welcome.

SaturnV

safetypee, thanks.

GY, perhaps the lack of sound effects in the simulator may explain this somewhat awkward phrasing in the English version,

Italics mine.

takata

Hi HazelNuts39,

Thank you sir, and VERY good work (as usual)
That's exactly what I expected:
https://docs.google.com/leaf?id=0B0C...zZDlm&hl=en_GB
Better posting a smaller representation directly for discussion:

This is showing where the stall/speed alarm system is flawed in relation to indicated airspeed issues:
"Sensed/indicated" airpseed drops below 60 kt, alarm stops because alpha is NCD. At first sight, it seems fair considering the very remote possibility that an aircraft would still be "flying" at such a low airspeed. Also, IAS went below limits of 30 kt "sensed", hence NCD is displayed on main instruments (only ISIS displays IAS down to 0)

While in reality, this aircraft airspeed was never under 100 kt but stayed in the range of 105-175 kt, where, of course it was not "flying" but "stalling". During about 1 minute, 0211:45 - 0212:45, airspeed was not that far from "flying again" values. The reversal of airspeed tendency happened when engines were set to IDLE.

With correctly interpreted informations (and test pilot skills) this aircraft might have been fully recoverable during this whole minute, as late as 0212:45, and maybe later providing enough altitude remained; Even THS setting is not an issue at such a low airspeed. Elevators are powerfull enough to overide it and trim back; this aircraft is longitudinaly very stable and thrust management could have made the difference for pitching down to "flyable" alpha.

curvedsky

Lonewolf_50

HN39 or takata:

Does the graph ( http://www.pprune.org/tech-log/45687...ml#post6621344 ) take into account, or synchronize with, the change in heading to the right over the course of the descent, which would thus vary the airspeed depending upon what winds were at a given level ... or, did you average the winds from what data is known to get the airspeed numbers from ground speed?

GarageYears

Agreed. I'm hoping that some of these 'issues' are recognized by the regulators, leading to enhanced requirements for training devices. The technology is available to provide representative cues for such conditions.

The RAeS ICATEE working group is doing some very good work related to upset training requirements, modeling approaches and similar. I am pretty sure many related to that effort are monitoring this thread.

takata

Hi Lonewolf,
It seems to me that HN39 computed the wind direction relative to heading on this reconstruction, see his explanation:
Change of wind with altitude could be another factor of incertitude, but the end result would provide a good approximation.

HazelNuts39

Yes it does. I chose a wind speed and direction that produced a match between the calculated Mach and the recorded Mach pior to the airplane going to very large AoA, keeping an eye on the wind recorded on the DFDR. One has to keep in mind that the DFDR wind is obtained by comparing IRU- or GPS-based groundspeed to the ADR-derived airspeed, and also that wind speed and direction may change in the interval shown.

airtren

Here is a set of graphs illustrating the association between certain pilot commands, Pitch variation, and Stall Warning OFF/ON/OFF, and the consequences on the confusion in the cockpit.

Unfortunately, the confusion took place during the time window in which the recovery was still possible (35000 ft to approx 6000ft).

Quote:

Originally Posted by airtren

....

The AF 447 pushed the system to its limits, and two problems surfaced:

a) the STALL WARNING stopped during the STALL, with a confusing effect - the a/c was still in STALL.

b) the STALL WARNING started during the transition from STALL to NON-STALL, with a confusing effect - the pilot commands were ND, or lower Thrust, to take the a/c out from the STALL, and the start of the STALL WARNING meant that the commands were wrong, and so they stopped/reverted the commands.

GarageYears

airten:

But you omit two key parameters from your graphs? Airspeed and AoA.... and since these two are fundamental to the generation of the Stall Warning itself I think you are painting a directed picture.

Take a look at the BEA data and it is clear that the Stall Warning basically follows CAS - the rest is coincidental to my eye at least.

Take just one Stall Warning occurrence - the one centered around 2.14.00 - on page 107 of the English version of the 3rd Report.... and exactly coinciding is a period where airspeed becomes greater than 60 knots. There is clearly some hysteresis around the airspeed value since short periods where CAS exceeds 60 knots don't trigger it, but when the value is there for a 'reasonable' period.

I suspect the problem is that excessive AoA leads to invalid CAS due to pitot characteristics - effectively the dynamic pressure is no longer being read since the inlet aperture is no longer 'inline' with the airflow, and the drain opening is exposed... all in all the problem is a vicious circle, driven by the insane AoA value.

Once IN THE STALL many of the systems necessary to understand the situation are compromised.

Solution = avoid STALLING

DJ77

No wonder he didn't understand, he could not know the PF was not trying to do what he said he would do after the zoom climb: descend.

takata

Agreed 100%. I'll add that, beside being told to avoid stalling, pilots should be informed of what would eventually happen (as precisely as possible) to their systems if they missed the first part of the drill.

Lonewolf_50

Thanks, HN. Winds vary (often considerably) from FL 380 to the surface in both direction and velocity ... so
So far so good
Given the lack of quality in air mass sensor input, that was sort of my point in asking the question.

That said, it only makes a difference if one is trying to sort out whether or not, when the airspeed on your chart increased late in the event, there was a chance to resume flying (how close to unstalling the aircraft might have been with some nose down) .... but that's all after the fact, as the inputs required to do so were not made.

hetfield

I'm deeply impressed about all the knowledge and enthuiasm to find out how this accident occurred.

About 1/3 of my career I flew Boeing, 2/3 I drove various Airbusses.

Day after day I'm asking myself, could this happen on a 737, 747, 767, 777.... ?

How far did Airbus go to put three qualified pilots out of the loop?

thx

jimbeetle

airtren:
Be very careful on how you interpret very fine differences on the BEA graphs. I believe the slight dip in the altitude one you have referred to over the past few days is a computer/graphics artifact/anomaly. If you look at all of the other altitude graphs they show a smooth curved without that dip.

xcitation

@takata/airtren (thanks for your recent posts)

Good call.
However how do you identify a real stall from a false stall?
Keep in mind that 6-9 months prior to AF447 incident we had a couple of false indication incidents on A330. Were these events fresh in the minds of the AF447 pilots? If so then it might be understandable to ignore stall warning and nose up!

Exhibit A: QF72 (extract follows...)
I believe it had failed PFD, false overspeed warn, false stall warn.


Exhibit B: QF71

safetypee

GarageYears, re noise simulation - enhanced requirements for training devices.
I recall that the PSM+ICR report identified a similar problem with poor simulation of engine failure sounds / indications. When all of the ‘relevant’ conditions were identified and programmed in simulation, the ability of a pilot to experience them all, would have taken longer than the total simulator sessions in a career.
Crews could end up permanently in the simulator, in ice, approaching the stall, recovering from upsets, etc, etc. All that might be achieved is avoiding the accident that someone else has just had; to avoid the potential for 'the next accident', the training solutions have to be more generic.

Some critical issue w.r.t. ice crystals are knowledge that the conditions exist, and judgement (risk assessment) in avoiding the hazard. Crews must avoid any complacency encouraged by modern technology, e.g. with a better WXR picture the closer that the ‘legal’ Cb misdistance can be flown. Always add a safety margin, ask what if, and have an 'undo' option' (CtrlZ).
Knowledge and judgement are aspects of airmanship – these are the core (or application) of professionalism. These qualities can be taught using a range of example situations, from basic training through refresher training, but perhaps most of all continuously exercised in daily operations with a ‘learning’ debrief – self improvement – striving to be a professional.

SaturnV, I interpreted “… the phenomenon being little known to pilots at the time …” as implying that the A330 engines had not suffered from ice crystal problems thus the safety threat had not been highlighted (A330 engines true/false?).
However, the ice crystal problem could have been known by all A330 crews based on previous pitot incidents with the appropriate dissemination of information – shared experiences / events. Within this I include the EASA decision to accept a delay in modifying pitots (a practicality) on the assumption that a crew could maintain control (in all foreseeable circumstances) for the duration of a ‘short’ encounter and potential loss of speed displays – enabling opportunity for an operator training action check.
The latter conclusion, could be biased by hindsight, but if we can turn that into foresight (foreseeable circumstances) then we might avoid similarly extreme or remote events. Operators should share safety experiences, and then both management and individuals should consider ‘what if’ for a range of scenarios – reinforcing knowledge, providing a basis for situation and risk assessment – a professional approach to operating in the modern world.