A great number of flight parameters are recorded by the Quick Access Recorder (QAR) equipped on civil aircrafts.QAR data is an important criterion for flight safety assessment.Aiming at large-sample and high-dimension features of flight data from QAR,this paper proposes an effective feature extraction algorithm,Deep Belief Network (DBN) algorithm.The DBN algorithm can adaptively extract the features of flight data independent of data-processing technologies and expert experiences.Simulations of different types of flight data sets are carried out.The simulation results show that,compared with the PCA algorithm,the accuracy of classification and identification of features extracted by DBN model is higher.