摘要
为了在故障样本稀缺、故障模式不完备的情况下监控涡轮泵状态,并剔除传感器失效故障造成的虚警,提出用于涡轮泵状态监控及传感器故障识别的单类支持向量机新异类检测方法。该方法以正常状态为目标类构建单类支持向量机检测器,用于检测涡轮泵是否出现异常;以传感器故障为目标类构建单类支持向量机检测器,用于判断检测到的异常是否属于传感器故障。对涡轮泵试车数据的分析结果表明了该方法的有效性。
In order to monitor the condition of a liquid rocket engine turbopump in the case of lacking fault samples and prior knowledge about fault modes, and to eliminate the false alarms caused by sensor faults, a novelty detection method based on one-class support vector machine （OCSVM） is introduced. An OCSVM is trained on the basis of normal samples, and is used as turbopump condition monitor. Another OCSVM is trained on the basis of sensor fault samples and is used as sensor fault detector. Turbopump condition monitor detects whether novel events may occur, while sensor fault detector identifies whether novel events detected are caused by sensor faults. The validity of this turbopump condition monitoring and sensor fault recognition method is verified with historical test data.
出处
《国防科技大学学报》
EI
CAS
CSCD
北大核心
2010年第2期119-123,共5页
Journal of National University of Defense Technology
基金
国家自然科学基金资助项目(50675219)
湖南省杰出青年科学基金资助项目(08JJ1088)
关键词
状态监控
新异类检测
单类支持向量机
液体火箭发动机
涡轮泵
传感器故障
condition monitoring
novelty detection
one-class support vector machine
liquid rocket engine
turbopump
sensor fault
