(华南师范大学 南海校区 计算机工程系, 广东 佛山 528225)
摘 要:针对传统的多向主元分析(multiway principal component analysis,MPCA)批过程监测的缺陷,提出了一种连续更新的改进移动窗多向主元分析(consecutively updated improved moving window MPCA,CUIMWMPCA)方法。该方法采用连续更新的多模型非线性结构代替传统的MPCA固定的单模型线性化结构,一旦通过改进的移动窗多向主元分析(improved moving window MPCA,IMWMPCA)判断出某一新批次过程正常,则模型参考数据库就随之更新。在实时监测新的批过程时,只需利用已收集到的数据信息,并且在线连续地更新模型参考数据库,提高了批过程性能监测的准确性,克服了MPCA不能处理非线性过程和实时性的问题。通过采用CUIMWMPCA与移动窗多向主元分析(moving window MPCA,MWMPCA)方法对青霉素分批补料发酵过程的实时监测,结果表明CUIMWMPCA比MWMPCA更适合于对缓慢变化的批过程进行监测,具有更可靠的监测性能。
CUIMWMPCA with application to fault monitoring of batch process
(Dept. of Computer Engineering, Nanhai Campus, South China Normal University, Foshan Guangdong 528225, China)
Abstract:A consecutively updated improved moving window CUIMWMPCA for online batch processes monitoring was proposed. The key to the new method was that whenever a new batch detected by IMWMPCA successfully remained within the bounds of normal operation, its batch data were added to the historical database of normal data and a new MPCA model was developed based on the revised database. The proposed approach only used the known data for monitoring batch processes and could consecutively update MPCA model. The proposed method was applied to monitoring fedbatch penicillin production, and the results clearly show that the proposed method is a more reliable method for monitoring slowly varying batch processes and could eliminate the many false alarms in comparison to MWMPCA. ......