Computer Aided Chemical Engineering
Last decades witnessed a significant increase in the production of high-valued products from industrial batch processes. Dynamic (non-stationary) behavior is an intrinsic characteristic of this class of processes. In this work the performance of several 2-and 3-way multivariate statistical methods able to cope with these features in on-line batch process monitoring are compared, along with the use of different control limits and data infilling methods. PCA, combined with leave-one-out control limits and current deviation or zero deviation infilling methodologies, shows the best overall performance. © 2014 Elsevier B.V.
Year of publication: 2014