Anomaly location method for QAR data based on principal component analysis hierarchical clustering

Abstract

The existing QAR (Quick Access Recorder) data anomaly detection algorithm can detect abnormal flights, but it cannot effectively locate the abnormal QAR data of abnormal flights. In order to address the issue, QAR data of abnormal flights are smoothed and pre-processed based on k-medoids algorithm. Pre-processed QAR data are transformed into a one-dimensional time series represented by angle cosine, and a subsequence feature matrix is generated from the angle cosine sequence by sliding window mechanism. Then, the dimension of the matrix is reduced based on principal component analysis, and the top-down hierarchical clustering is performed on row vectors of the reduced matrix according to the amount of information within each column. The anomaly nodes are detected according to the number of vectors contained in the clustering tree nodes. The abnormal data segments of QAR data of abnormal flights are located by the positions of vectors generated in the cosine angle sequence in the abnormal nodes. The experimental results on real flight data sets show that the proposed method can locate not only the known exceedance events, but also the abnormal fragments besides the monitoring items.