Similarity Search for Flight Data

Abstract

Similarity search can be used in search of data necessary for model training, and is therefore, an important part of information mining. Since flight data presents itself as a typical time series, for aircraft and its subsystems in similar operating modes or flight conditions, there should be series sets with similar change trends. By means of searching data series with similar features, similarity search provides technological support for flight data research and “mines out” information of empirical values and their underlying rules, thus laying down a solid basis for flight data reconstruction and monitoring of aircraft conditions. Due to the insufficiency of the traditional method of point-to-point comparison, this chapter will elaborate on unary similarity search of flight data in relation to slope distance, angle distance, and curvature distance. In this chapter, the method of multivariable-series-oriented similarity search will also be introduced. This method, verified by real flight data, makes use of variable step length curve binning and QR decomposition of incidence matrix.