Here is what an A330 captain said about the crash:
The key ingredient most everyone seems to be overlooking: The flight control laws of an Airbus. An Airbus has flight envelope protections that cannot be overridden by the pilot. This is almost always a good thing because the airplane won’t allow the pilot to overspeed, stall, overbank or overload the airplane. In the peculiar case of [Air France Flight] 447, the airspeed reading was inaccurate because the pitot tubes were blocked — a very rare occurrence in a jet — almost never happens.
But when it does happen, the airspeed then acts like an altimeter: When the airplane climbs, the indicated airspeed increases, and when the airplane descends, the indicated airspeed decreases. My best guess for AF447 is that the airplane was climbing, most likely due to turbulence; I believe they were in a thunderstorm. From a pilot’s perspective, this is a bad place to be. It’s rough and difficult to read instruments. Autopilot disengages due to turbulence or ice on the airframe or pitot tubes. The airplane is climbing, and the pilot is wondering what the **** is going on. Then, as the airplane climbs, with the false readings still indicating increased airspeed, at high altitude the margin between cruise airspeed and overspeed becomes very small, so the airplane overspeeds — or so it “thinks,” due to the false reading. And it’s at this point, provided all of this is what really happened, that they’re ******.
Think back to the Airbus flight envelope protections I mentioned, and the fact that the pilot can’t override them. The airplane computers “think” the aircraft is overspeeding and therefore continue to increase the airplane’s angle of attack. That only makes it climb steeper, thus perpetuating this cycle of increasing indicated airspeed and increasing angle of attack. This continues until the airplane is at a ridiculously high nose-up attitude and stalls, regardless of pilot inputs.
This is why we really need to wait for a full analysis, so that investigators can figure out what the pilot inputs were and whether or not they were consistent with what the flight control surfaces were doing. In other words, were the pilots ******* up the control deflections, or were the Airbus flight computers ******* up the control deflections? Because the airplane eventually stalled, I can only surmise that it was the computers ******* up, because when the computers do their jobs correctly, they increase angle of attack in this situation — again, regardless of pilot input.
So the airplane stalled. One plausible theory is what I just described. (There are other scenarios in which the flight control laws are degraded and the airplane can stall under certain circumstances, but that’s a whole other set of seriously complex stuff. Who knows though? Maybe that’s actually what happened.) Even if this guess doesn’t explain precisely what took place, it constitutes a design flaw in the Airbus that needs to be fixed in that the flight envelope protections need to be disabled if they’re receiving inaccurate information.
Source:
An Airbus Captain s Take on the Air France Disaster