@Machinbird:
IIRC the turbulence conditions were described as "MODERATE". Would that really have been enough to roll the aircraft on its own? I may be being somewhat naive in this supposition, but I would think it requires significant sustained force on an object of that kind of mass...

What was the delta ISA?

The period of time in discussion is between 2:12:33 to 2:12:43, around FL210. The ND position was dominant, the average value clearly on positive side of the graph. However, the elevator did not followed the input, it never passed half way from maximum of NU. Therefore, is not due to the "developed" stall condition. But because the Dnz feedback response tends to fall off, due to lift loss and the airframe has a natural tendency to pitch ND. Then elevator is more and more driven NU, due to integration of the increasingly large CL error. Yes, there the pilot was right, but the aircraft response went badly wrong.
The automatic stabilizer trim function will only start moving ND after the elevator (not the sidestick) has moved from a NU to a ND position. I think even 30 seconds are not enough to straighten the THS from "braking chute" position.

I think translation is accurate since it's copy/paste from Appednix 1 of the final report and the french verb monter = climb, nothing else.
I agree, Dubois warned Bonin to go down.

He had a heavy hand or maybe the sidestick is not suitable for the hand of pilot in distress
I look to the roll command input vs the phenomenal NU input, between 11:40 to 12:15, I'm wondering what if he kept the sidestick inadvertently at max NU while fought to bring the wings to horizontal...

Machinbird


Some have discussed keeping hands off completely in the period immediately after the A/P disconnected and the switch to Alt 2b law.

Just a personal opinion, but that course would have a high probability of ending in a spiral dive. There is nothing to level the wings in Alt 2b but the aircraft's natural roll stability (which isn't all that good in most jets).


We don't have to speculate - Fig 64 of the Final Report shows a simulation of the aircraft response to the turbulence present both with and without pilot input.


With no pilot inputs 40 secs after loss of airspeed the aircraft would have ended up at 280 kts CAS, FL350, 3 deg pitch, a shallow climb and bank wandering around 10 deg right, but not divergent.


Just a thought, but there is nothing to say that Alt2b loses the stability augmentation terms (roll and yaw damping) in the laws.

IIRC, AF447 was GE CF6-80E engines - the CF6-80C2/80E engines are impressively tolerant to inlet distortion. Some other big fan engines would not have reacted so gracefully.


As an aside, we propulsion engineers tend to take a somewhat perverse satisfaction when the engines are operating fine right up until impact in an event such as this (the Aero Peru 757 also comes to mind :( ). OTOH, Lauda 767 was our worst nightmare :uhoh:.

tongue in check
 

All the normal planes love to roll, loop, stall, dive, spin, recover, hammerhead, in one word to fly.They are born to fly! Planes, and pilots too! So long nobody limited their pleasure to fly by any limiting/protecting/castrating design, software or system, every well born plane is easy to roll to well born pilot ;)
Going from normal law to alternate 2B law without enough training is more hasardous as a difference appears in rates of pitch and roll stick movements. Regulators have to verify pilots are prepared to that event. It is not a stability issue, but adaptation from pilots when trained to such modification in flight.
Thank you Machinbird and Owain Glyndwr to point the right things both.:)
Regulators must do their work better, faster.

For that kind of mass, the aircraft has that kind of wings and vertical stabilizer...
Remember, the aircraft encountered weather with ice pellets.
See Figure 26 of final report. In only 2 seconds, after AP disconnect, the roll angle passes from 0 to 8.4 while roll input remained neutral. Next 15 seconds, PF continuously corrected the right tendency of the aircraft, 4 times, 2 times at left limit stop.
In the same figure, it is interesting to notice the AOA for "STALL STALL". Robert might saw 270 kts and asked "What is that?"

A pathetic tv reconstruction uses "What is that?" to emphasize the "idiot" scenario

Air Crash Investigation S12E13 - Vanished (Air France Flight 447) - YouTube

Sorry - not buying that - I've never been entirely convinced that the DFDR elevator readout doesn't take the THS position into account. I see OG has popped along - Owain, what's your take on this?

What I can tell you is that in the sim, the THS began rolling forward less than a second after the stick was held ND.

Also, the movement in the time period you state is two "blips" forward - while it trends that way, it's nowhere near enough to make a significant difference.

Plenty of successfully and safely landed emergencies on Airbus FBW types say otherwise.

No - the spring centralisation is very noticeable - you know when you're breaking it out of the central position. A deflection of that magnitude can only be deliberate.

@DozyWannabe
Final report: "The trimmable horizontal stabilizer (THS) began a nose-up movement and moved from 3 to 13 degrees pitch-up in about 1 minute and
remained in the latter position until the end of the flight."
The sidestick was held ND, at least in couple of occasions

Notice the neutral position of sidestick. The NU stop is close to vertical
The NU travel is much shorter than max roll command. If you have a heavy hand pilot in distress, my supposition "holds water"

AIRBUS SIDE-STICK OPERATION - YouTube

No it wasn't. Your "10 seconds" consisted of two short "blips" - one to the full ND position for about 1s, then back to neutral, followed by a "blip" to the half-forward position for less than a second, then back to neutral before resuming the NU input.

Having used one in the simulator, I can tell you that's not true. You're using a video in which the lens is distorting the apparent Z-axis.

This was not accidental - this was a pilot making a "haché" of controlling the aircraft when he didn't (immediately) need to.

DozyWannabe,
You have to zoom a bit in the 15 s interval, see there 2 triangles, in about 10 seconds interval. The area of those triangles is the quantity of ND input.
For Robert, the quantity of ND input is half of area under the graph, due to "smart" dual input

edit: the prolong left roll command came after 2:11:31, not after AP disconnect.

You can't "zoom in" much past 200% without the resolution becoming too coarse to tell much. Also, your second "triangle" has too many "blobs" of red/brown (representing data points) to make a call of that ilk. Robert was making no inputs during "2:12:33 to 2:12:43, around FL210". This is not a software flight control problem, this is a "pilot freaking out and playing whack-a-mole with the controls" problem.

The only time of appreciable "DUAL INPUT" was around 2:13:44 to 2:14:02, with Robert trying to get the nose down and Bonin sporadically pulling up.

At about 02:14:18, Bonin slams the stick back one last time, and hits the sidestick priority switch, locking Robert out. He makes no mention of doing this.

Dozy


Sorry - not buying that - I've never been entirely convinced that the DFDR elevator readout doesn't take the THS position into account. I see OG has popped along - Owain, what's your take on this?

I don't think I have a "take" and I'm not sure I understand what you mean by the readout not taking the THS position into account. The elevator deflection as recorded in the DFDR is measured relative to the THS chordline, so the elevator position relative to the fuselage datum is measured elevator angle plus THS deflection.


Does that answer your point?


Or did you mean the bit about increasing NU elevator?


In which case, although I wouldn't use the same words I tend to agree with Phoenix.


Remember that in both normal and alternate laws S/S movement commands a delta g not an elevator position. Neutral S/S is in effect a command to "keep me at the 'g' I have at this moment", so if the stick remains neutral and the 'g' falls because there isn't enough lift to sustain the g at the time the stick was released the system will apply up elevator in an attempt to recover the selected level. If this persists then the THS, following the integral of elevator deflection, will also move NU.


Coming back down, so far as I can see, one would need to apply a negative g command greater than the actual (less than 1g) situation at that time before the system would actually apply ND elevator. That command would have to be held for some considerable time before the integrated elevator signal caused the THS to start moving ND.

Obviously when it will be talks of Airbus at the trial .. discussions on behavior and the system of THS going to be warm while it operated in accordance with the specifications
The STALL alarm system and behavior will certainly be put on the grill also

They would have to be on a fixed (safe) gain basis if so.

That means that roll stability terms might be a bit twitchy in cruise at altitude and thus might have helped contribute to the initial roll PIO. (sheer speculation at this point)

They would have to be on a fixed (safe) gain basis if so.

That means that roll stability terms might be a bit twitchy in cruise at altitude and thus might have helped contribute to the initial roll PIO. (sheer speculation at this point)


Yeah, they might well be fixed gains, but we don't know if those particular terms are scheduled anyway do we?


OK, it might be a bit less damped in alternate but I suspect that the major change was from stick movement commanding roll rate to stick movement commanding roll acceleration.

Absolutely right. That difference was apparently never appreciated on AF447's flight deck.

The roll oscillation finally stabilizing was likely related to the decrease in roll acceleration available as a result of the substantial deceleration that the aircraft had experienced as was approaching stall speed.

Machinbird


The roll oscillation finally stabilizing was likely related to the decrease in roll acceleration available as a result of the substantial deceleration that the aircraft had experienced as was approaching stall speed.


Not sure about that. I take your point that system gain (roll acceleration/stick movement) was reduced as the airspeed fell, but against that you have to set worsening dutch roll damping and increased roll/sideslip ratio as AOA increases. And I don't think one can exclude Bonin learning how to fly it!

There's a pretty good in depth article about the accident and the culture of pilots in France in the latest issue of Vanity Fair. It's an interesting read.

Vanity Fair ! Not very professionnal reference for p..p..RuNers ! You may find better anywhere :(