Securing air transportation safety through identifying pilot's risky VFR flying behaviours: An EEG-based neurophysiological modelling using machine learning algorithms

• Published in: Reliability engineering & system safety 2023-10, Vol.238, p.109449, Article 109449
• Main Authors: Li, Qinbiao, Ng, Kam K.H., Yiu, Cho Yin, Yuan, Xin, So, Chun Kiu, Ho, Chun Chung
• Format: Article
• Language: English
• Subjects: Air transportation; Classification; EEG; Pilots; Risk flight behaviour; Visual flight rules
• ISSN: 0951-8320; 1879-0836
• DOI: 10.1016/j.ress.2023.109449

•The effects of weather and flight routes on pilots’ unsafe behaviour are revealed.•The moderation of these impacts caused by flight phases is investigated.•The neurophysiological modelling was built to recognising risky flight behaviours.•The optimal classifier's accuracy for identifying unsafe behaviour is 80.8%. Most general aviation accidents result from pilot errors while flying under visual flight rules (VFR), many of which are linked to situation awareness prone to meteorological and situational effects. This study reveals the impacts of weather and flight route factors contributing to unsafe flight behaviours on pilots’ performance (flight operation & neurophysiology). Importantly, another purpose is to identify risky flight behaviours by executing EEG-based neurophysiological modelling for securing air transportation safety. A two-by-two (adverse/clear weather; experienced/inexperienced flight route) within-subject experiment was conducted. Inexperienced flight routes have an explicit impact on contributing to unsafe flight operations except during departure and descent. However, the neurophysiological activity in power spectrum density and cooperative correlation showed that meteorology deterioration may require more mental efforts and have an implicit impact on maintaining safe flight awareness, making it challenging to correct risky flight operations due to mental resource exhaustion possibly. For the pilot's risky behaviour identification, neurophysiological modelling using EEG data was established, and a leave-one-subject-out cross-validation was used to evaluate the modelling feasibility, with the highest accuracy of 80.8%. This work could serve as the foundation for developing a real-time risky flight behaviour prediction modelling that could support safety-enhancing air transportation via artificial intelligence.