AI flight-safety system wins award

A system that uses artificial intelligence to improve flight safety has been awarded a prize from the Technology Strategy Board.

A research collaboration between Portsmouth University and Flight Data Services received the award for Best Knowledge Transfer Partnership (KTP) in the South East and will now be entered for a national award.

The organisations collaborated under a three-year KTP, the government’s scheme that facilitates academic institutions working with businesses. Together they developed a computer program that uses artificial intelligence to analyse data recorded in an aircraft’s black box.

The program highlights tiny aberrations in the recorded data after every flight that would not usually be identified. It flags up abnormalities that fall outside the airline’s standard safety parameters, which it can investigate and take remedial action if necessary before safety is compromised.

The program was developed in response to the industry’s need for a more comprehensive and accurate system of flight-data monitoring. The industry requires that airlines monitor data from all passenger flights over 27 tonnes. This includes aircraft ranging from 10-seat corporate jets to commercial jets seating up to 850 passengers, such as the Airbus A380.

Currently, flight-data monitoring is a semi-automated process carried out on a flight-by flight-basis using a set of predefined safety criteria that check for known problems. However, every flight generates masses of data from dozens of instruments and hundreds of information feeds, requiring hours of labour-intensive scrutiny by skilled observers.

Dr David Brown, head of Portsmouth University’s Institute of Industrial Research (IIR), said: ‘This intelligent software will do the same job in a fraction of the time and will identify data that would never have been detected by a human being. It literally looks for the needle in the proverbial haystack.’

The system works by comparing flights against each other and looks for similarities within apparently random sequences of data. Those that are most similar are grouped together to identify recurring patterns and anomalies during a flight that would previously have gone undetected.

Flight Data Services plans to build the program into a data-analysis suite, incorporating the technology into the services it provides to its 50 customer airlines worldwide.

The success has lead to a second KTP venture between the university and Flight Data Services, which will analyse the black-box recorder itself.