Application of ECOD algorithm for detection of aircraft unstable approach
Once an unstable approach occurs during one of the approach and landing phases,it may lead to an incident or accident.Therefore,the detection of unstable approaches in terminal areas has attracted a lot of research interest in the field of flight operation status monitoring.Aiming at the detection algorithms for unstable approaches,a data-driven method for unstable approaches detection is proposed using OpenSky's open-source aviation surveillance data set.The unstable approaches detection model is based on the Empirical-Cumulative-Distribution-based Outlier Detection(ECOD)unsupervised anomaly detection algorithm from the perspective of energy management.ECOD computes an outlier score of each data point by aggregating estimated tail probabilities across dimensions,abnormal scores are then obtained for the detection of unstable approaches.The distance of the aircraft from the runway,altitude,ground speed,vertical rate,and total energy are used as features to form a training dataset that includes 5-dimensional features.After dimensionality reduction by PCA,the detection model is obtained by training using the ECOD algorithm.The probability of an unstable approach increases with higher anomaly scores.According to the unstable approach criterion,the labeled data are acquired and the ECOD model is compared with two other popular data-driven models,iForest and HDSCAN.The ECOD model achieves a detection accuracy of 0.73,a breakthrough in the accuracy of unstable approach detection.Aircraft energy state monitoring in conjunction with the ECOD model contributes to energy management and unstable approach monitoring during aircraft approaches and landings based on previously established energy safety boundaries and identified air traffic flow.Go-around data is fed to this model,the results demonstrate the accuracy and efficiency of the proposed model.
万方数据
维普
