时段划分的多向主元分析间歇过程监测及故障变量追溯

针对间歇过程的多时段特性,提出一种生产过程操作时段划分方法.该方法利用反映过程特性变化的负载矩阵以及主成份矩阵的变化实现了间歇过程子时段的两步划分.提出了基于加权负载向量夹角余弦的负载矩阵相似性度量以及基于加权奇异值变化的奇异值矩阵相似性度量方法,以更客观的反映负载矩阵以及奇异值矩阵的相似性,进而更准确的判断过程特性的变化.根据同一操作子时段的过程特性,其负载矩阵和奇异值矩阵相似性较大的特点,实现了生产过程的子时段划分.将基于子时段划分的多向主元分析(MPCA)建模应用于三水箱系统的在线监测和故障变量追溯,实验结果验证了该方法的有效性.

Multiway principle component analysis monitoring and fault variable detection based on substage separation for batch processes

According to the multistage characteristics of the batch process, we propose a new stage separation method for the production process. Based on the variation in loading matrices and principal component matrices which reflect the evolvement of the underlying process behavior, a two-step substage separation is proposed. To objectively show the similarity between the loading matrices and the similarity between the principal component matrices, two similarity measurement methods are applied to estimate the variation of the process characteristic with higher accuracy. These two methods are respectively based on the weighted cosine of the angle between loading vectors, and based on the weighted absolute value of the singular value variation. Process substage separation is realized because the loading matrices and the singular value matrices in the same operation substage are with great similarity. Based on the improved stages separation method, the multiway principle component analysis(MPCA) modeling is applied to online monitoring and fault variable detection in a three-tank system. The experimental results verify the effectiveness of the method.