摘要
为了全面便捷地检测出有效的潜在的设备异常,基于设备大量有价值的工况数据,对设备异常检测方法进行研究,提出一种面向设备群体的工况数据异常检测方法。通过对设备状态监测数据潜在异常特征进行分析挖掘,识别设备运行中的异常状态和用户异常行为,提高产品服务质量。给出了设备异常检测原理,对数据预处理、特征提取、正常特征空间分布建模、异常检测和人工验证过程进行了详细阐述。通过对5 000多万条真实监测数据进行异常检测实验,验证了所提方法的可行性和有效性。
To detect effective potential equipment abnormal comprehensively, the detecting method based on amount of valuable equipment condition monitoring data was researched, and a device group-oriented method for abnormal floor data detecting was proposed. Through analyzing and mining the potential abnormal features of equipment con dition monitoring data, the abnormal state and the user abnormal behavior in equipment operation were identified, and the service quality of products was improved. The principle of abnormal detecting was presented, and the processes of data preprocessing, feature extraction, spatial distribution of normal features modeling, abnormal de tecting and artificial validation were elaborated in detail. More than 50 million real monitoring data were detected in anomaly detecting experiment, and the feasibility and effectiveness of the method were validated.
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2013年第12期2993-3001,共9页
Computer Integrated Manufacturing Systems
基金
国家863计划资助项目(2012AA040911
2012BAF12B23)~~
关键词
设备群体
工况数据
特征空间
分布建模
异常检测
device group
floor data
feature space
distribution modeling
abnormal detecting
