Dynamic Assessment of Cabin Crew Psychological Safety Margin Using a BP Neural Network with Self-Supervised Learning

The assurance of aviation safety and service quality is highly dependent on the psychological state of flight crew members. As a core determinant of safe decision-making and behavioral resilience, mental health necessitates the development of a scientific and dynamic evaluation system. This study centers on the coupling mechanism between mental health and safety margin, and proposes a dynamic evaluation model for the psychological safety margin of cabin crew members based on a backpropagation (BP) neural network. Experimental results show that the BP neural network achieves a coefficient of determination of R[Formula: see text] between the predicted and actual safety margins, indicating an excellent level of predictive accuracy. The model successfully establishes a real-time assessment framework that links psychological health with operational safety margin. By enabling the dynamic coupling of crew mental state and safety performance, this BP-based approach introduces a novel paradigm for proactive risk warning in aviation safety management. By leveraging quantitative tools to uncover the black box between psychological health and safety behavior, this approach facilitates a shift in aviation safety management from post-incident accountability to proactive prevention. It offers an advanced risk management paradigm for high-reliability human-operated environments.