Fault prognosis in control system for satellite attitudes based on fuzzy basis function networks and autoregression model
卫星姿控系统基于模糊基函数网络与自回归模型的故障预测

A new method based on fuzzy basis function networks(FBFN) and autoregression(AR) model is proposed for predicting faults in the control system for satellite attitudes. Firstly, normal satellite attitude data are used to train FBFN which is used as the standard model of the control system for satellite attitudes. Secondly, the real-time attitude residual errors are obtained by subtracting the FBFN output from the real-time data of satellite attitudes. Thirdly, the time series of the residual errors is used to build an AR model. Therefore, the faults in the control system for satellite attitudes are predicted by using the AR model, and the failure probability is given according to the statistical distribution of the prediction errors of the AR model. Finally, the con dence factor is determined which shows the con dence measure of the fault prognosis.