2 ms·

Thank you, that was a good summary. From the incident report (PDF) it seems the system chooses either the median or arithmetic mean AOA values. Could the incid

Thank you, that was a good summary. From the incident report (PDF) it seems the system chooses either the median or arithmetic mean AOA values. Could the incident have been averted with more robust filtering algorithms?

bdonlan 15y ago
Median suffers from the very real (ie, seen in a test flight) problem where one side of the plane can give bad data to two ADIRUs, so that's out. Supposedly there's a new system being deployed via a software update (see 7.1.1) that does more detailed monitoring of each channel individually - looking for oscillations, etc. They also got rid of this 1.2-second monitoring period. The report doesn't go into much detail about this though. Keep in mind, we're not talking about analog noise here. We're talking about one type of data being labelled as another - but only part of the time. Analog filtering techniques might not be helpful, depending on how often it switches.