一种新的多阶段改进MPCA间歇过程监视方法的研究

针对间歇生产过程存在的多阶段问题,提出了基于数据动态特性CPV(1)(cumulative percent variance of the first principal component)指标进行模糊聚类实现多阶段软划分的方法,解决了传统分段方式对间歇过程进行硬划分的缺陷,使得过程多阶段划分更加准确。在此基础上建立多阶段具有时变主元协方差的改进MPCA(multiway principal component analysis)模型进行间歇过程的监视。将此方法应用于青霉素发酵过程,验证了该方法的可靠度和有效性。

A New Method of Improved MPCA for Multi-Stage Batch Process Monitoring

For multi-stage problems in batch production process,a method of implementing the multi-stage fuzzy clustering based on the data dynamic characteristics of CPV(1)(cumulative percent variance of the first principal component) is proposed to realize multistage soft division.The defects of the traditional way of hard division of batch process are solved,and the process of multi-stage is divided more accurate.The improved MPCA (multiway principal component analysis) model based on multiple stages with time-varying principal component covariance is built for batch process monitoring.This method is applied to a penicillin fermentation process,the reliability and effectiveness of the proposed method are verified.