摘要
核概率主元分析(kernel probabilistic principal component analysis,KPPCA)能够有效去除过程的非线性。但是KPPCA仅构造了生产过程的静态线性关系,处理具有较强动态特性的实际工业生产过程效果较差。为克服上述缺点,提出一种基于动态KPPCA的过程监测方法,利用核函数将经过压缩的动态增广数据映射到高维空间,然后利用PPCA对满足线性关系的过程变量映射值进行监测。仿真结果表明:该方法监测指标对故障的灵敏度高,误报率和漏检率较小,故障状况与正常状况很明显的分离开来。
Kernel probabilistic principal component analysis (KPPCA) can effectively eliminate nonlinear process features.However,The KPPCA method only constructs linear static relations among the process variables,so it cannot effectively deal with real industrial processes with strong dynamic characteristics.A dynamic KPPCA was developed for process monitoring to eliminate these disadvantages by mapping the compressed data matrix extended by the time series into a high-dimensional space by a kernel function.Then PPC...
出处
《清华大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2008年第S2期1824-1828,共5页
Journal of Tsinghua University(Science and Technology)
基金
教育部新世纪优秀人才支持计划(NCET-05-0485)
国家"八六三"高技术项目(2007AA04Z198)
江南大学创新团队发展计划资助
关键词
动态核概率主元分析
过程监测
非线性
高维空间
dynamic kernel probabilistic principal component analysis (DKPPCA)
process monitoring
nonlinear
high-dimensional space