Originally Posted by BOAC
I note you have now introduced another event into the stats, namely 'T'.
Not me, I just named it.
I agree wholeheartedly that T is of interest. However, I
don't think it's wise just to assume a correlation with non-X. There are lots of things which don't happen in two-person cockpits which may well happen in one-person cockpits. In a one-person cockpit, a T would likely become a non-X; in a two-person cockpit it is (by definition) not. That said, there are many reasons why T events would be of interest. I'm sure you know most of them but I think it's a good idea to ferret them out explicitly.
First, I want to split T and non-X events, which is what ChristiaanJ could have done to make his point (but I'm glad he didn't, for then he would have deprived me of the opportunity to use the smiley
).
P: PF fails to retard all thrust levers to idle on landing
Q: PM notices, draws attention to it
R. All thrust levers are then retarded
Here are the definitions of non-X and T in terms of P,Q,R.
* non-X = (P & not-Q & not-R)
* T = (P & Q & R)
The factor in common to both T and non-X events is P.
There is also another possible event to consider, another of those "how could anyone possibly do that?" events that we thought/think non-X is. This event is
* V = (P & Q and not-R)
Those who think that V is extremely unlikely are right. But V-like events have already occurred. In the Nagoya accident in 1994, the PF inadvertently triggered the TOGA switch. The PM (the CAP) notified him of that, and told him to turn it off. Which the PF didn't do and the CAP didn't insist. The PF was pushing the column for all he was worth; the AP was trimming against it, to full nose-up trim. When the CAP belatedly took over, the AC balloon-climbed, stalled, and landed tail first.
Let's get back to T events. First I observe that the common factor to both T events and non-X events is P. P is also common to V events, of which there haven't been any.
First, a P event is a necessary causal precursor of a non-X event. Thus there is an opportunity for prophylaxis of non-X events: stop all P's, you necessarily stop all non-X's.
Second, there may be many P events, so one might be able to calculate some useful statistics, unlike with non-X events.
Third, a P event is also an instance of failure of the "Retard" warning. The PF is to retard: heshe failed to do so in a P event despite the warning.
The AI argument I labelled no. (1), produced by Malinge on 9 August, can be countered using this third point. The argument was that the "Taipei" warning was redundant, because the "Retard" warning already advised that thrust should be reduced to idle. If there are relatively many P events in comparison with the 3 non-X events, we would have relatively many instances in which the "Retard" warning has failed, thus suggesting one indeed needs a different means of advising that thrust has not been reduced to idle, contrary to what AI said they concluded with the regulators.
Whether that should be a "Taipei" warning in addition to "Retard", or an altered form of warning altogether, does not follow from this, but it has already been decided by H2F3 Revision 3 so this consideration is moot.
The downside of this may be that if AI realises that P events are likely to weaken their argument, they may be less motivated to collect T-event statistics from their fleet operators until after any Congonhas legal proceedings have finished.
BTW, AI won the civil case associated with the V-like event at Nagoya.
PBL