On-line Monitoring of Batch Processes Using Additive Kernel Partial Least Square

Abstract

In this paper, the additive kernel partial least square (AKPLS) is proposed for on-line monitoring of batch processes. The proposed AKPLS is a special case of kernel partial least square method which inherits advantages of partial least square and considers the nonlinear relationships among monitoring variables. Moreover, the monitoring statistics can be estimated only with incomplete data because the additive kernel can be decomposed into the sum of kernels in different time slices, which is suitable for the online-monitoring application of batch processes. Finally, the effectiveness of proposed AKPLS is verified by experiments on the fed-batch penicillin fermentation process.