Putting flight patterns under the microscope
Identifying patterns from aircraft accidents and near misses could unlock vital clues to help improve flight safety, according to a James Cook University researcher.
PhD candidate Aoheng Ma is poring over historical flight data produced by the Automatic Dependent Surveillance–Broadcast (ADS-B) system, on the lookout for anomalies in the flight paths of aircraft.
Mr Ma is using an artificial intelligence-based deep learning method, known as Generative Adversarial Networks (GANs), to detect both short-term and long-term patterns and anomalies from historical flight data sourced from the OpenSky Network.
“The aviation sector has grown quickly over the past decades, and abnormal flight behaviours causing both fatal and non-fatal aircraft accidents remain a real concern,” Mr Ma said.
In 2022 alone, there were 11 aircraft accidents around the world resulting in 224 fatalities.
“We’re hoping that by using several different GAN-based sub-networks we can build up a good anomaly detection framework for this data, and improve on the existing methods,” Mr Ma said.
Mr Ma’s approach to detecting anomalies will focus on five different flight phases – take-off, climb, cruise, descent and landing.
“We’ll be able to spot anomalies from one or a group of flights, based on radar data,” he said.
“As the technology evolves in the future, we’d hope to develop an online model that is capable of detecting flight anomalies early, so the pilot or ground staff can be warned that the aircraft is veering off course.”
Mr Ma said he will initially limit his research to analysing flight data from one geographical area but hopes to grow that footprint in the future.
The project is expected to take three to four years.
Media enquiries: firstname.lastname@example.org