论文部分内容阅读
针对航天器早期故障在闭环系统下难以被检测、数学模型难以精确建立的问题,提出了一种基于数据关联性分析的归纳式系统异常监测(IMS)方法。该方法采用无监督学习的聚类算法,利用具有关联性的参数构建数据向量,通过聚类分析自动建立健康数据向量的族类阈值区间。关联关系的破坏将引起部分参数超出族类阈值区间,使系统的异常程度存在模糊性与随机性。引入云模型评价指标,将闭环系统异常程度的不确定性通过熵与超熵定量表示,从而更加准确地判断闭环系统的异常程度。仿真结果表明:该方法能够建立卫星飞轮闭环系统的族类知识库,并可以根据云模型提供的定性知识有效判断系统的异常程度。
Aiming at the problem that the early failure of the spacecraft can not be detected in the closed-loop system and the mathematical model can not be accurately established, an inductive system anomaly detection (IMS) method based on data relevance analysis is proposed. The method uses unsupervised learning clustering algorithm, constructs data vector using the parameters with relevance, and automatically establishes the family threshold interval of health data vector by cluster analysis. The destruction of the relationship will cause some of the parameters to exceed the threshold of ethnic groups, making the system anomalous degree of ambiguity and randomness. The evaluation index of cloud model is introduced, and the uncertainty of closed-loop system anomaly is quantified by entropy and hyper-entropy to judge the abnormality of closed-loop system more accurately. The simulation results show that this method can establish a family knowledge base of satellite flywheel closed-loop system and can effectively judge the degree of anomaly based on the qualitative knowledge provided by cloud model.