Abstract: Batch processes are widely used in fine chemical and biochemical industries. The objective of batch processes is to produce high value-added products of high-quality. In order to reduce the variations of the product quality, multivariate statistical process control methods based on Multi-way Principal Component Analysis (MPCA) are used for on-line batch process monitoring. However, the MPCA and MPLS, based on the decomposition of relative covariance matrix, are strongly affected by outlying observations. In this paper, a PSO-based robust MPCA is proposed. How to construct the robust normal operating condition (NOC) model and robust control limits are discussed in detail. It is evaluated by monitoring an industrial streptomycin fermentation process and compared with the conventional MPCA. The results show that the robust MPCA is resistant to possible outliers and thus robust.

摘要： 间歇过程广泛应用于精细化工产品、生物化工产品等高附加值产品的制备.为提高间歇生产的可重复性，提高批次之间产品的一致性，多向主元分析法(MPCA)广泛应用于间歇生产过程的监控.针对MPCA统计监控模型容易受到建模数据中离群点影响的不足，提出了一种基于微粒群优化算法(PSO)的鲁棒MPCA分析方法，并进一步给出了相应鲁棒监控统计量的计算方法.对于链霉素发酵过程的监控表明，相对于普通MPCA，鲁棒MPCA在建模数据中存在离群点时仍能够给出正确的统计监控模型，从而有效减少了建模过程对数据的要求.