I've lurked on this forum for years, but for the first time I think that I might have a valuable idea to contribute. (I'm just a PPL, but as this conversation is more about engineering than pilotage...)
In regards to the idea posited by
bob.arctor and commented on by
Mad (Flt) Scientist - if I understand it correctly - to take a model-based approach to determine airspeed, I have this to suggest. Instead of building a complete flight model, or using a new sensor technology (eg. covering the skin of the aircraft with piezoelectric devices), why not just create an database with prior good data?
Presuming that you've had at least a few minutes of reliable indication off of a pitot source, you now have
some reasonable data on aircraft performance. Put another way, on any given flight, you know what airspeeds have resulted with recent combinations of AoA, power, vertical speed, and temperature. That should be sufficient to backdrive at least an approximate aircraft mass. If you wanted to get fancy you could incorporate flight control positions as well, but I'm primarily thinking about level, unaccelerated flight.
If airspeed data then become unreliable due to any factor - sensor failure, icing, impact damage, whatever - you should already have a reasonably close estimation of present aircraft mass (perhaps supplemented with data from a fuel totalizer) and, if I'm not mistaken, should then be able to calculate airspeed given present AoA, power, vertical speed, and temperature. This system should also make it possible to resolve disagreement between multiple pitot / static systems.
If each airframe had a onboard database of prior combinations of the same types of data, then it should be possible to deal with the failure of one sensor, whether it was an AoA vane, pitot tube / static port, thermometer, or whatnot. This strikes me as something that would be quite valuable for pilots, especially because it would eliminate a lot of guesswork as to what sensor or system was providing bad data. It such a system was incorporated into flight logic, it might have prevented recent upsets due to sensor or software failures. (And no, I'm not just referring to AF447, I'm also thinking of the Malaysia Airlines 777 upset over Perth in 2005.)
Obviously, for a system such as I've described to be useful, a single sensor failure should not make two variables unknown. For example, you'd want vertical speed to come from a gyroscopic source, not from the static system, since static port blockage could simultaneously render both airspeed and vertical speed unreliable.
Also, it's worth noting that I'm not sure how good the temporal resolution of this sort of system would be, but I'm reasonably certain that it would be good enough to avoid overspeed / stall during cruise and to maintain Va on approach. If its response were made fast enough it might also be able to take care of cruise gust load alleviation.
Perhaps someone's already implemented such as system? Thoughts and comments would be appreciated.
Cheers,
Alex