A Pilot Fatigue Prediction Method Based on Dynamic Bayesian Networks

Pilots of long-haul aircraft face a variety of challenges, including unstable flight environments, confined and narrow cockpit spaces, complex human–machine system operations, multiple tasks, and long-haul flight times. This study analyzed the factors leading to pilot fatigue from four aspects (human, machine, environment, task) and predicted the fatigue risk of long-haul flights using a dynamic Bayesian networks method. First, we identified factors related to fatigue during long-haul flights from four aspects: human, machine, environment, and task, and established an index system containing 20 fatigue risk factors. Second, 10 experts in the field of aviation evaluated these factors within the fatigue risk system to derive the prior probabilities for the dynamic Bayesian networks on pilot fatigue on long-haul flights. Finally, we introduced the Noisy-OR model to derive the conditional probabilities and calculated the posterior probabilities using the dynamic Bayesian networks. We validated the proposed method with a real case study, and the results showed that this method can predict fatigue during long-haul flights.