In order to make full use of flight big data to realize its data value, a flight safety assessment model based on multi-source information fusion is constructed. The data of all time, space, all directions and all elements in the flight process are gathered and integrated, and the logical architecture of multi-source heterogeneous flight big data cloud platform is designed; the flight safety assessment model based on multi-source information fusion is established, and the deep mining analysis and fusion of flight big data are realized through 3-layers information fusion, and the risk sources of flight training are identified, the risk level is determined, and the analysis results are summarized by visualization. The model can provide decision support for flight training management, flight safety assessment and risk early warning, and ensure that the decision-making has rules and data to follow, so as to improve the level of flight training safety management.