Flight parameter anomaly data detection based on timing consistency approach

Flight parameter data is one of the objective and scientific bases to describe the flight process. Its error will significantly affect the results of subsequent data processing and analysis. Collecting effective training aircraft flight data is critical for flight safety, pilot training, flight performance analysis, fault diagnosis and maintenance, and data-driven decision making. Because flight parameter data is a typical multidimensional time series data, there are correlations among different dimensions of multidimensional time series data, and multiple dimensions may change cooperatively, resulting in inter-dimensional correlation anomalies. Therefore, anomalies cannot be accurately detected by considering only single-dimensional time series information. This paper proposes a monitoring method of flight parameter anomaly data based on time sequence consistency. Through multiple outlier detection of one-dimensional flight key parameters, one-dimensional outlier data is marked and the time stamp of multi-dimensional flight key parameter outlier data can be obtained after intersection, which can accurately detect abnormal data in flight parameter data files.