基于有限状态机的航天器异常检测专家系统知识获取方法  

Knowledge Acquisition Method for Spacecraft Anomaly Detection Expert System Based on Finite State Machine

作　　者：黄连兵 尹桂松 冻伟东 陈茜 段姝宇 HUANG Lianbing;YIN Guisong;DONG Weidong;CHEN Qian;DUAN Shuyu(Beijing Institute of Spacecraft System Engineering,Beijing 100094,China;National Academy of Innovation Strategy,Beijing 100038,China)

机构地区：[1]北京空间飞行器总体设计部,北京100094 [2]中国科协创新战略研究院,北京100038

出　　处：《宇航学报》2024年第9期1481-1487,共7页Journal of Astronautics

摘　　要：面向航天器异常检测专家系统知识快速、准确获取需求,提出一种基于有限状态机(FSM)原理的“数据驱动+领域知识”融合获取方法。首先基于数据库中遥控指令、遥测参数、正常值范围等结构化数据,建立状态集、输入集、输出集;其次,将设备工况变化看作状态变量,以遥控指令、注入等事件为输入信号,建立遥测变化与状态迁移的映射关系,将“知识生成”问题转化为对状态转移函数的求解过程;最后,利用历史测试数据驱动有限状态机的状态转移求解模型,将得到状态转移函数集,生成可运行的知识。以中国空间站测控分系统遥测为例进行方法实验,实验结果表明,所提方法获取的异常检测知识可解释性强,且具有较强工程可用性,可为后续卫星工程任务检测知识获取提供参考。A“data-driven and domain knowledge”fusion acquisition method based on the principle of finite state machine(FSM)is proposed for the rapid and accurate acquisition of knowledge in spacecraft anomaly detection expert systems.Firstly,based on structured data such as telecommands,telemetry,and normal threshold in the database,establish the state set,input set,and output set.Secondly,considering the changes in equipment operating conditions as state variables and using injection,telecommands and other events as input signals,a mapping relationship between telemetry changes and state transfer is established,which generates expert system knowledge by solving state transition functions.Finally,using historical telemetry data to drive finite state machines for state migration,a set of state transition functions will be gained,which can generates actionable knowledge.Taking the telemetry of TT&C of China Space Station as an example for method experiments,the experimental results show that the anomaly detection knowledge obtained by the proposed method has strong interpretability and engineering applicability,which can provide reference for the acquisition of detection knowledge for subsequent satellite engineering tasks.

关 键 词：异常检测专家系统 数据驱动 知识获取 有限状态机 

分 类 号：V557.2[航空宇航科学与技术—人机与环境工程]