Fault Detection and Prediction Method of Satellite Senor In-orbit Data Based on SVM

Satellite sensor is one of the most important devices for spacecraft flight control, which is used to gather the in-orbit flight control data. The timeliness and accuracy of the sensor fault detection determined the success or failure of the mission. The traditional data monitoring strategy is triggering the alarm by setting the upper and lower threshold value of the telemetry data. This paper tries to achieve a fault detection and prediction method of satellite sensor through the process of classifier training using the normal and abnormal in-orbit simulated data based on SVM. During the training and detection of the sample data, multi-parameter has been adopted, the two-class classification and the one-class classification algorithm has been compared. According to the results, the method based on SVM was demonstrated effectively.