In order to investigate the dynamic propagation process of flight operation safety incidents, a layered propagation network model was used to investigate incident evolution. Triggering events, propagating events, and top-level events were extracted from accident reports and statistical reports using the event tree method. Subsequently, the unidirectional risk network was constructed. The initial node layer, probability propagation layer, capacity-load propagation layer, and system dynamics propagation layer were defined based on propagation characteristics. Specific propagation methods for each layer were developed. The probability propagation layer quantifies trigger probabilities using node metrics. The capacity-load propagation layer quantifies load distribution through node capacity functions. The system dynamics propagation layer simulates unsafe events through time accumulation. Network analyses from multiple perspectives, including cascading failure scale, risk evolution processes, and event trigger frequencies, were conducted. Event control measures for flight operations were proposed based on simulation results. Analysis of the Yichun air crash indicates that reducing load transmission efficiency of connected edges extends the risk evolution process by 23%, and blocking connected edges prevents the accident occurrence. For the Shandong Airlines accident, reducing initial node loads extends the risk evolution process by 40%, and controlling edge states extends the risk evolution process by 46%, thus preventing accident symptoms. The results show that this method accurately characterizes the evolution process of unsafe events during flight operations, with an average simulation time of approximately 3 minutes per unit time.