文献类型：期刊文献

英文题名：110th Anniversary: An Overview on Learning-Based Model Predictive Control for Batch Processes

作者：Lu, Jingyi[1];Cao, Zhixing[2,4];Zhao, Chunhui[3];Gao, Furong[1]

机构：[1]Hong Kong Univ Sci & Technol, Dept Chem & Biol Engn, Kowloon, Hong Kong, Peoples R China;[2]East China Univ Sci & Technol, Key Lab Adv Control & Optimizat Chem Proc, Minist Educ, Shanghai 200237, Peoples R China;[3]Zhejiang Univ, Sch Control Sci & Engn, Hangzhou 310027, Zhejiang, Peoples R China;[4]Tongji Univ, Shanghai Inst Intelligent Sci & Technol, Shanghai 200092, Peoples R China

年份：2019

卷号：58

期号：37

起止页码：17164

外文期刊名：INDUSTRIAL & ENGINEERING CHEMISTRY RESEARCH

收录：;EI(收录号：20193907480509);WOS:【SCI-EXPANDED(收录号:WOS:000487179800005)】；

基金：Z.C. gratefully acknowledges careful proofreading by James Holehouse. F.G. acknowledges support from the National Natural Science Foundation of China (Project No. 61433005) and Hong Kong Research Grant Council (Grant No. 16207717).

语种：英文

外文关键词：Learning systems - Batch data processing

摘要：Batch processes repeatedly execute a given set of tasks over a finite duration, whose versatility and ability to adapt to rapidly changing markets make it prevalent in a multitude of industrial fields, particularly in the era of "smart manufacturing". Nevertheless, frequent switching and wide-ranging facility operations incur significant nonlinearity and time variability in process dynamics, both of which constitute remarkable challenges toward the regulation of batch processes. Among the various regulatory schemes, the integration of model predictive control and iterative learning schemes stands out, because of its inheritance of the merits of both: (i) ease of handling physical constraints and (ii) utilizing the repetitive operation pattern to adjust control input, process variables, and reference to improve control performance, consequently enhancing product quality. This review intends to account the recent technical advancements during the past two decades, from the perspective of the three different levels of learning mechanisms: control input, model parameter, and tracking reference. We conclude by providing insights into future research.