Satellite Attitude Control System Fault Detection Method Based On Telemetry Data

With the rapid development of space technology,the structure and function of satellites have become more complex and diversified.Coupled with the particularity of the space environment,the guarantee of its safety and reliability has become particularly important.Some traditional fault detection methods can no longer meet the increasing demand for satellite support,and data-driven fault detection has become one of the important research directions.In this paper,the attitude control system with the most critical satellite and the highest failure rate is the typical research object.According to its parameter characteristics,the fault detection technology research is carried out based on the telemetry data.The main work and contributions of the thesis are as follows:(1)According to the different manifestations of the telemetry data in the abnormal state of the satellite attitude control system,this paper summarizes two failure modes of deviation fault and sequential fault,and deals with the corresponding data set in a targeted manner to provide a data basis for subsequent fault detection.The possible failure modes of the satellite during orbit have been simulated and analyzed.(2)For biased faults,a weighted isolated forest fault detection model based on the importance of features is proposed.An unsupervised anomaly detection method is used to solve problems such as the difficulty of labeling faults samples and the extreme imbalance of positive and negative samples.The random strategy of node splitting in the original isolated forest algorithm improves the performance of a single decision tree and improves the accuracy and efficiency of the model.(3)For time series faults,a time series fitting model based on LSTM is proposed to learn the long-term dependence of linear and non-linear relationships in satellite telemetry time series,and use an unsupervised threshold setting time series anomaly The detection method is to perform anomaly analysis on the error sequence generated by the LSTM fitting model to achieve sequential fault detection.In summary,based on the particularity of satellite telemetry data,this paper proposes targeted fault classification and detection methods to mine the fault information in satellite telemetry data.Experiments and comparative analysis are carried out in real data sets to verify the accuracy and efficiency of the proposed method.