In order to solve the problem that the static assessment technology can not be corrected timely and effectively according to the dynamic changes of flight operation risk factors, through the statistics and analysis of over 5000 civil aviation unsafe incidents in the past decade, the dynamic Bayesian network is established for flight operation risk by identifying key risk indicators, and the dynamic Bayesian network parameters are initialized according to the statistic results. Then, a real flight example is simulated and the network parameters are corrected in real time according to the information of the navigation system, and weather forecast data in the risk control system, thus the prediction results of pre-flight and in-flight are obtained. The results show that flight safety level is high in the process of departure and cruise, the risk of landing is significantly fluctuating under the influence of the wind shear of the Airport, and the risk of low risk possibility is reduced to 17.2% and the medium risk is up to 70.9% , which shows that alternate or flight return is recommended for low safe landing likelihood. The strategy is consistent with the actual operation results ,which shows that the scheme is feasible and effective. In order to confirm further, the operation data expands to thunderstorm and snow season in 2017 from a single case, the verification results show that the forecasting program is 80.4% consistent with the actual operations, indicating that the forecasting scheme is reliable.