In order to ensure human-machine safety during training flight, analyzing flight data to monitor and evaluate flight quality has become one of the means to improve safety and training quality. The flight anomaly data analysis method based on Energy-Metrics was used to investigate the safety of training flights. A general method for identifying abnormal flight data during the approach and landing phase of training flights based on energy metrics was proposed. Firstly, the energy metric was used to generate feature vectors for flight data records. Then, with the help of density-based spatial clustering of applications with noise(DBSCAN) clustering and single class support vector machine(SVM) methods, the energy metric metric data was used for flight anomaly detection and recognition. By hiding simulated abnormal flight data in actual flight data, flight data was detected and identified. The results show that the detection success rate of this method is over 95%, and the consistency of abnormal flight data identified using different energy measurement indicators is as high as 98%. It is concluded that this method has strong effectiveness and robustness in identifying abnormal flight data.