论文部分内容阅读
针对卫星姿态控制系统的故障预测问题,给出了模糊基函数网络(FBFN)与自回归模型(AR)相结合的故障预测方法,并提出了预测置信因子的概念,对故障预测的准确性进行评价.首先利用卫星正常运行时的姿态数据训练FBFN,将训练好的FBFN作为卫星姿控系统的标准输出模型;然后把卫星实时姿态数据与FBFN输出数据之间的差值作为残差,利用AR模型对残差序列进行建模,进而对未来的残差进行预测;最后依据预测残差的统计分布给出了故障发生概率,利用故障预测置信因子来描述预测步长不同时故障预测结果的可信性.
Aiming at the problem of fault prediction in satellite attitude control system, a fault prediction method based on fuzzy basis function network (FBFN) and autoregressive model (AR) is proposed. The concept of prediction confidence factor is proposed and the accuracy of fault prediction is evaluated Firstly, the FBFN is trained by the attitude data of the satellite during normal operation and the trained FBFN is used as the standard output model of the satellite attitude and attitude control system. Then, the difference between the real-time attitude data of the satellite and the FBFN output data is taken as the residual. Finally, based on the statistical distribution of prediction residual, the probability of failure is given, and the prediction result of failure prediction with different prediction steps can be described by using the confidence factor of fault prediction Credibility.