Unsupervised flight phase recognition with flight data clustering based on GMM

Currently, with the rapid development of the aviation industry, researchers are paying more attention to the improvement of aviation safety. Aviation safety mainly includes flight safety, aviation ground safety, and air defense safety. In terms of flight safety, the analysis of large amounts of flight data has gradually become a useful tool for timely detection of potential dangers at various stages of flight. As a result, flight data analysis has been one of the hot topics in aviation. However, due to the complexity of the aircraft operating conditions throughout the aircraft, if the data is analyzed at the entire flight phase, it is very difficult and time consuming to identify the problematic fight phase. Therefore, flight phase recognition for civil aircraft is implemented in this study. A flight phase recognition method based on Gaussian Mixture Model (GMM) is proposed in this work, which is the important foundation for timely detecting the abnormal event and improving the system safety and reliability. Firstly, the FDR data are preprocessed by spline interpolation and normalization, and then a GMM-based flight phase clustering is realized. In addition, a set of evaluation method is developed to evaluate the quality of flight phase recognition result. Finally, the effectiveness of the method is verified by using real FDR data from NASA's open database.