Transfer learning for nonlinear batch process operation optimization. (May 2021)

• Record Type: Journal Article
• Authors: Chu, Fei; Wang, Jiachen; Zhao, Xu; Zhang, Shuning; Chen, Tao; Jia, Runda; Xiong, Gang
• Is Part Of: Journal of process control. Volume 101(2021)
• Page Start: 11
• Page End: 23
• Publication Date: 2021-05
• Subjects: Transfer learning -- Nonlinear batch process -- Insufficient data -- Operation optimization -- Data selection; Process control -- Periodicals; Fabrication -- Contrôle -- Périodiques; Process control; Periodicals; Electronic journals; 660.281
• DOI: 10.1016/j.jprocont.2021.03.002
• Languages: English
• ISSNs: 0959-1524

Abstract

Abstract: This paper concerns with the JY-KPLS model based transfer learning for the operation optimization of nonlinear batch processes. Due to problems of data insufficiency and uncertainties in a new nonlinear batch process that has just been put into production, the model-(new) process mismatch is usually inevitable, which is also the main reason for the poor performance of the batch process. To solve this problem, this paper first adopts the JY-KPLS model to capture the behavior of the nonlinear batch process, and takes full advantage of the information in similar batch processes to assist the modeling and operation optimization of a new process. Then, a data selection based batch-to-batch optimization control strategy is proposed in this paper to reduce the adverse effects of this mismatch on the operation of the new batch process. Finally, the feasibility of the proposed method is demonstrated by simulations. Highlights: The optimization of the batch process with strong nonlinearity and scarce process data is realized. JY-KPLS transfer model is applied to nonlinear batch process operation optimization. NCO mismatch problem is analyzed and reduced by model updating and data selection. The algorithm framework of JY-KPLS model based batch-to-batch optimization is given.