Denouement
Denouement
As in most such accidents, the final revelations prove to be the disclosure of a previously un-encountered but (in the aftermath) easily explicable phenomenon - and an ensuing chain of complicating and (pilot) confounding circumstances..... some of which are related to design and envelope test-flying deficiencies (but more on that later). Is it a computer-created chain of events? Judge for yourself, from the discussion below, what part the automation (or even weather) played.
After autopilot disconnect due to pitot ice-up and ADR disagree, and some ensuing pilot surprise "disconnect" (from priorities), max power was applied at or near the aircraft's cruise ceiling, the underslung engines provided a pitch-up moment that caused the aircraft to quickly climb into a stall in what's colloquially called "coffin corner". Unfortunately and simultaneously, the autotrim caused the trimmable horizontal stabilizer (THS) to motor (quite unnoticeably) to its maximum "nose-up" sustaining position. This combination of power and "back-trimmed" trim-state created a configuration that was to thereafter (i.e. during their high-rate descent) sustain a deep-stall condition that could only be entered via such a powered speed-to-height "inadvertently zoomed" scenario. Why did they add power? It's a natural pilot response to do something similar when the airspeed winds back off the clock (even though in most such circumstances it only serves to "upset the applecart" of stability).
To explain, the translatable inertia of an aircraft at height has always amazed me. When you zoom climb (even quite shallowly) at very high altitude, you are translating, into a rapidly expiring upwards vector, the very considerable energy that's there solely courtesy of its TAS - (i.e. at cruise height the True Air Speed - i.e the actual speed through the air - is roughly double the indicated airspeed ). When that kinetic energy excess is exhausted (the height increment becoming potential energy), the out of kilter aerodynamics then must dictate what follows. The resulting flight-path down from a stall in that rarefied atmosphere is a "locked-in" function of attitude and power and residual trim-state. The trim state they ended up with was the deadly legacy of the auto-trimming THS rapidly reaching max "nose-up"... as the speed decayed in the thrust-induced zoom. The power of that tilted THS slab was now enough to keep that A330 pitched up into the stall, particularly whilst under engine power. The deep-stall condition is quite a stable flight regime - and that has been known since the first BAC-111 accident almost 50 years ago (only a tail-chute deployment could have saved it). Back prior to automation it was thought to be a phenomenon only associated with T-tailed aircraft (wing-wash swamping the all-flying tail and negating its control). But since AF447 we've discovered that the A330's THS has the power to emulate this lethal configuration... but with deceptively smooth passivity.
Lacking an AoA attention-getting read-out or aural/visual alert, the pilot can only "go for" an ADI pitch attitude (and one that approximates the level flight attitude would seem reasonable to most - and it's what's been generally, but quite mistakenly, advocated here). Unfortunately in a deep-stall condition the relative airflow is not from "ahead" (but from "well below" and ahead), so maintaining (or just abiding) a notional 5 degrees nose-up "cruising" pitch is to fatally elect to live within that deep-stall's boundaries. Deep-stall is a non-alarming condition because of the lack of tail-plane buffeting. In a normal stall the empennage is being "bathed" in wing-created boundary layer separation turbulence - and it's an airframe and control feedback situation that pilots everywhere can recognize as indicative of a stalled condition. Without it you are plummeting down in an eerily smooth and silent (but non-apparent) steep trajectory.
Of course having a stall warning that only momentarily cautions an initial approach to a stall is a design that (for its pilot's alerting function) actually conceals the deep stall. Once embedded in such a stall, the aural alert ceases ("how can I be stalled, there's no stall warning?"). Any stick forward initiative (i.e. angle of attack reduction into the lower AoA numbers) only cooks off (quite perversely) that aural stall warning and increases the perplexed pilot's misunderstanding and non-recognition of his predicament ("now we're stalling with stick forward, how can this be?"). He's never seen this profile on any simulator ride. Even the Airbus test-pilots just "haven't been there" beyond the flight envelope, to test this "post departure" phase. Whatever assumptions about it were ever made? I'd suggest absolutely none. It was an area that, like max achievable Mach and IAS, is best theorized about (only). Much of what is specified for air-testing is about safe boundaries (beyond which there be dragons best left undisturbed). Airbus fails to shine torches (or even look) into dark corners of their beastly envelopes and automation perturbations.
It's no surprise to me that, once locked into their deep-stall smooth-flying regime, the AF447 pilots failed to recognize the nature of their predicament. They momentarily acknowledged their high-rate descent but due their 3-way interaction, their confusion, the smoothness of flight, the silence of low IAS, the distraction of myriad alerts, the darkness of night and lack of any other visual cues, all available cues failed to trigger awareness of any imminent catastrophe. Even the PF's side-stick grip and applied input was concealed from the captain's view. If they'd had a distinctive AoA alarm, experience of it and a laid down memory recall procedure for affirmative action (i.e. positively lower the nose until the stall-approach warning recurs (then ceases) or at least to 20 degrees nose-down), well we'd not now be agonizing over their needless deaths. It's a recognition "consciousness" trigger that's generally lacking here in our automation. As pilots of automated airplanes, we need a climactic bathos (or Eureka moment) whenever we're required to spring into an alert and cognitive state. Other interim pilot-level "fixes" (such as not TOGA'ing the power and prioritizing the lowering of the nose and using manual trim) have already been addressed in Airbus pilot bulletins.
I'd be interested in any rebuttals..... technical or otherwise. Do I think that any fixes implemented will be effective? Partially but not wholly is my suspicion. No Airbus or Air France or BEA bod is going to want to acknowledge the totality of the AF447 conundrum. AF447 will reside in the Pilot Error Hall of Infamy forever. But to allow this, without demur, would be a calumny against the profession of the professional aviator. And that's why I'm penning this. Put this out for the public and journalists to see and understand. That's how you might avert the way I see the final report going.