An Optimized Support Vector Regression for Identification of In-phase Faults in Control Moment Gyroscope Assembly

摘要

One of the critical components in the satellite attitude control subsystem is the control moment gyroscope. If it fails, the satellite cannot finish its mission. Fault isolation followed by in-time corrective action can help prevent failures. However, it is necessary to know the fault severity for better maintenance planning and prioritize the corrective actions. This way, the more severe faults can be corrected first. Therefore, a data-driven fault identification scheme is proposed in this paper, adopting an optimized support vector regressor to determine the severity of multiple in-phase faults of the satellite control moment gyroscopes. The features are extracted using correlation analysis. A grid search is used to optimize the model's hyperparameters, and the R2-score is adopted to evaluate the model performance. It is shown that the proposed scheme can predict the fault severities with 94.9% accuracy, on average.