Stage-based variable sampling period modeling and on-line monitoring strategy for uneven-length batch processes

Batch processes are often characterized by uneven-length durations and multistage characteristics. To reflect the inherent stage nature to improve the performances of process monitoring, simultaneously considering dynamic characteristics within the process variables for some complicated cases, stage-based variable sampling period multi-model dynamic principal component analysis (VSP-MDPCA) modeling and on-line monitoring method is developed in this paper. Batch process is firstly divided into several stages by feature point (FP) extraction method. In the extraction of feature points, wavelet de-nosing is firstly used to prevent noise interference. For each uneven-length stage, process data are sampled with variable sampling period according to reaction intensity. Then dynamic time warping (DTW) algorithm is used to align the trajectory of each stage. DPCA model is built for each stage. The proposed method was used to detect faults in the fed-batch penicillin production. The simulation results clearly demonstrate the advantages of the proposed approach in comparison to MPCA.