论文部分内容阅读
针对多向主元分析(MPCA)方法用于批过程在线监测时需要对新批次未反应完的数据进行预估,从而易导致误诊断,且需要建模批次的长度相等的缺陷,提出了一种基于变量展开和主元协方差随时间变化的MPCA方法。该方法按变量展开,不需要对新批次未反应完的数据进行预估,而数据之间的动态联系通过时变主元协方差得以保存,并且不需要建模批次的长度相等。将该方法应用于β-甘露聚糖酶发酵批过程实时监测中,并与MPCA、移动窗多向主元分析(MWMPCA)法相比,结果表明该方法克服了MPCA不能处理实时性的问题,避免了MPCA在线应用时预测值的误差,比传统的MPCA,MWMPCA方法更能精确描述过程的故障,准确性和实时性良好。
The method of multi-way principal component analysis (MPCA) is used to estimate the unreacted data of new batches when on-line monitoring in batch process, which leads to misdiagnosis and needs to simulate the defects that the length of batches is equal. An MPCA method based on variable expansion and principal component covariance over time. The method is based on variables and does not need to estimate the data of the new batch that has not been reacted. The dynamic relationship between the data is saved by the time-varying principal component covariance, and the length of the modeling batch does not need to be equal. The method was applied to the real-time monitoring of β-mannanase fermentation process, and compared with MPCA and MWMPCA, the results showed that this method overcomes the problem that MPCA can not handle real-time and avoids Compared with the traditional MPCA and MWMPCA methods, MPCA can accurately describe the fault of the process, and the accuracy and real-time performance are good.