Mike-Bracknell: I agree that you'd have to do input checking, but even this would probably have positive safety consequences. For example, if the pitot gets plugged and your airspeed goes from M.80 to M.27 in less than a second, you know that something's gone wrong, and it's overwhelmingly likely to be related to the pitot / static system if there's little or no recorded vertical acceleration. (Again, this presupposes that you're getting VS from a gyro or accelerometer and not your static source!) I don't know quite what you mean about loss of control of the senses: are you concerned that pilots would lose the ability to determine which systems had failed?
bob.arctor: A synthetic IAS / TAS system would require a purpose-built real-time OS that was far more reliable and rigorously tested than a general purpose operating system like Linux or most versions of BSD, including the Mac OS. Thankfully, such RTOSes already exist. Additionally, if a synthetic IAS system were to be implemented, it would have to function with the limited computational resources that are available on an aircraft. That's why I suggest that using a database with known-good data would be a better approach than some sort of real-time modeling. Any system that requires heavy analysis (say, finite element analysis) is almost certainly inappropriate for in-flight use, at least right now. I can imagine that in ten or twenty years we'll have the computational horsepower to alter airfoils in-flight to achieve whatever results are desired.
Given recent rumor and information about similar problems with A330s at TAM, NWA, Qantas, and Air Caraibes, it's tempting to focus on pitot/static system-specific problems.
I would suggest that the bigger issue is how to keep a computer-controlled aircraft flying safely with persistent bad data from all of the sensors of a single type. Since we cannot reasonably build aircraft with three different designs of pitot tube (for example) from three different vendors, we have to prevent an unanticipated design flaw in that one pitot tube design from delivering fatally bad data. There’s no reason for that bad data to cause a loss of control, never mind a bunch of bent metal and lost lives. This goes for
all sensor subsystems: accelerometers, AoA vanes, engine performance monitors (FADECs), etc., and not just pitot / static systems.
Cheers,
Alex