vilas,
I agree with nearly all of your latest
thoughts about the AF A340 Incident. You make some valid points, including the apparent discrepancy in relation to the overspeed protection. The latter had disengaged the AP in the A340 AIRPROX of 2001, but seemed already overdue to do so on the AF A340 when the backward movement of the F/O's sidestick manually disengaged it. Like you, I find the BEA statement that you quote to be confusing.
Where our respective understandings do still differ, however, is in what may happen
after AoA-mode engages by phase-advance resulting from a
brief up-gust.
It's worth noting here that the extra (3rd) disengagement condition available on the a/c in the AF A340 incident of 2011 caused the first 5 of the 6 reversions to Nz mode, for which the sidestick remained neutral. In the case of the A340 AIRPROX incident, where the 3rd condition had not been incorporated, AoA mode consequently persisted throughout the main event - even after a small push-forward on the sidestick for less than a second - until instantly disengaged by a selection of more than half-forward sidestick.
Quote (my emphasis):
"This is not what happens in AOA protection rather it only can cause pitch down to Valpha prot with stick neutral."
I disagree here, as I tried to explain in my last reply to you. A brief up-gust can cause the phase-advanced value of AoA to exceed alpha-prot briefly. Actual AoA can remain below alpha-prot, and when the up-gust ceases the phase-advanced value will return to the real AoA. Meanwhile, however, AoA mode has engaged. Under the present-day three-condition disengagement logic, AoA mode would disengage after 0.5 sec if both sidesticks were neutral. Under the two-condition disengagement logic in the A340 AIRPROX of 2001, it would remain engaged with neutral sidesticks. So in the latter case the EFCS will have to
increase the AoA to Alpha-Prot, which will probably involve a pitch-up. This is what I think happened in the AIRPROX case, and why I questioned the disengagement criteria.
One of the vulnerabilities of my argument in the A340 AIRPROX case is that,
AFAIK, we don't have any estimates for what would have happened to the pitch (and, therefore, the flightpath) if the AoA mode had reverted to Nz, delivering 1.0 G for the long period that the sidesticks remained neutral.
Quote:
"In any case Pilot’s action should have been to push the nose down and recover the speed. So automation has not played any adverse role in this."
Your first sentence is correct for both events. Your second sentence is arguable in the A340 AIRPROX case, as I've explained above. In the AF A340 case, however, the F/O's early action (pulling the stick) was - like AF447 - what started the zoom-climb. Once he had released the stick, the AoA mode was free to disengage each time the AoA happened to fall below Alpha-Prot for half a second. Each time, Normal mode targeted 1.0G, which on 5 occasions led to re-engagement of AoA mode. As previously noted, the first
re-engagement was too late to prevent an exceedance of Alpha-MAX (perhaps in a gust).
These two events may present conflicting arguments for and against the third disengagement condition. I suspect that AI engineers may have been revisiting AoA-mode engagement and disengagement criteria since 2011.