Research on Auxiliary Detection Technology for Downlink Telemetry Parameter Anomaly Discovery

Abstract

In order to solve the problem of efficient and automatic identification for the anomaly detection of carrier rocket downlink telemetry parameters, an automatic identification method based on the statistical characteristics of historical data is proposed for the slowly varying parameters. Under the condition of time slicing, the method first performs numerical filtering on historical measured data, and then Gaussian Process Regression (GPR) algorithm is used to locally model the segmented historical data. Secondly, the prediction output of the GPR sub-model based on each historical sample data is fused by the dynamic Gauss-Markov estimation algorithm to obtain the prediction value and prediction variance of the target data, and the parameter discrimination interval is constructed based on this. The real-time abnormal alarm is given when the statistical distribution of the measured target data of each sub segment in the discrimination interval exceeds the limit. Finally, after accumulating the complete and effective data segments of the measured parameters of the target, the global distribution is counted, and if it exceeds the limit, the parameter abnormality flag is given. Simulation results show that this method can effectively find abnormal parameters, and has strong ability to suppress random noise and more accurate parameter estimation ability.