摘要
针对多向主元分析法(MPCA)在线监控传统方法的预测能力差而常常导致误报、漏报的实际情况,基于广义相关系数法提出一种在线监控方法。从监控模型库中找出与待测多元轨迹之间广义相关系数最大的批次,将该批次监控时间点后的数据作为待测多元轨迹的预测值,得出完整的批次数据,并用多向主元分析法进行监控。用青霉素发酵过程仿真器(Pensimv2.0)进行仿真,结果表明,基于广义相关系数法的在线监控方法与其他传统方法相比,其结果最接近实际过程离线,并减少了由于其他预测方法带来的误差。
The problem of poor prediction of the traditional online monitoring approaches based on multiw＇ay principle component analysis （MPCA）is discussed. Based on generalized correlation coefficients（GCC）, an online monitoring method is presented. The batch with the largest GCC to the being tested multivariate trajectory in historical model library is taken out for monitoring. The part of the batch data after monitoring time point is regarded as the complement of predictive values of being tested multivariate trajectory. After receiving integrated batch data, MPCA is used to monitor the process. A penicillin fermentation simulation process （ Pensimv 2.0） is used to testify the algorithm. The results show that, the monitoring with GCC prediction is similar to the real offline process and the errors from other prediction approaches are reduced.
出处
《控制工程》
CSCD
北大核心
2009年第1期113-116,共4页
Control Engineering of China
基金
辽宁省科学技术基金资助项目(20052023)
辽宁省教育厅基金资助项目(2005320)
关键词
广义相关系数法
多向主元分析法
在线监控
青霉素发酵过程
generalized correlation coefficients
muhiway principle component analysis
online monitoring
penicillin fermentationprocess
