Cluster analysis of the myriad parameters accessible on flight-data recorders could provide a valuable insight into operational performance, with a view to accident prevention, research from the Massachusetts Institute of Technology (MIT) indicates.
While flight operations monitoring can assess instantaneous values of variables, such as airspeed, or detect occasions when well-defined parameters exceed limits, MIT said underlying problems can remain obscured.
Flight-data recorders offer a rich set of information which could be better exploited, it added. MIT researchers have developed a cluster analysis tool which groups flights which have common patterns.
"Flight data outside the clusters are flagged as abnormal," the institute said, allowing the tool to detect unusual instances without the need to pre-define a specific area of interest.
"The beauty is that you don't have to know ahead of time what 'normal' is," MIT aeronautics professor John Hansman said.
MIT's work involved examining a group of Boeing 777 flights into Abu Dhabi, to illustrate the principle, as well as studying 365 flights over a month operated by an undisclosed carrier, using 777s on various routes.
Take-off and landing data showed anomalies which were "due mostly to crew actions", said Hansman. One flight departed with much lower thrust, another had "erratic pitch behaviour", signalling difficult rotation, while a third was low on approach with an abnormally high flap setting.
MIT, which co-operated on the research with Madrid's Comillas Pontifical University, will present a paper on the technique to the DASC digital avionics conference in Seattle at the end of October.
Source: Flight International
