化工学报 ›› 2021, Vol. 72 ›› Issue (4): 2178-2189.DOI: 10.11949/0438-1157.20200995

• 过程系统工程 • 上一篇    下一篇

基于多尺度核JYMKPLS迁移模型的间歇过程产品质量的在线预测方法

褚菲1,2(),彭闯2,贾润达3,陈韬4,陆宁云5   

  1. 1.地下空间智能控制教育部工程研究中心,江苏 徐州 221116
    2.中国矿业大学信息与控制工程学院,江苏 徐州 221116
    3.东北大学信息科学与工程学院,辽宁 沈阳 110819
    4.萨里大学化学与工艺工程系,英国 吉尔福德
    5.南京航空航天大学自动化学院,江苏 南京 210016
  • 收稿日期:2020-07-23 修回日期:2020-09-17 出版日期:2021-04-05 发布日期:2021-04-05
  • 通讯作者: 褚菲
  • 作者简介:褚菲(1984—),男,博士,副教授,chufeizhufei@sina.com
  • 基金资助:
    国家自然科学基金项目(61973304);江苏省六大人才高峰项目(DZXX-045);江苏省科技计划项目(BK20191339);徐州市科技创新计划项目(KC19055);矿冶过程自动控制技术国家重点实验室开放课题(BGRIMM-KZSKL-2019-10);国家煤加工与洁净化工程技术研究中心开放课题(2018NERCCPP-B03);前沿课题专项项目(2019XKQYMS64)

Online prediction method of batch process product quality based on multi-scale kernel JYMKPLS transfer model

CHU Fei1,2(),PENG Chuang2,JIA Runda3,CHEN Tao4,LU Ningyun5   

  1. 1.Engineering Research Center of Ministry of Education for Intelligent Control of Underground Space,Xuzhou 221116, Jiangsu, China
    2.College of Information and Control Engineering, China University of Mining and Technology, Xuzhou 221116, Jiangsu, China
    3.Industrial Artificial Intelligence and Optimization Institute, Northeastern University, Shenyang 110819, Liaoning, China
    4.Department of Chemical and Process Engineering, University of Surrey, Guildford, United Kingdom
    5.College of Automation, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, Jiangsu, China
  • Received:2020-07-23 Revised:2020-09-17 Online:2021-04-05 Published:2021-04-05
  • Contact: CHU Fei

摘要:

针对过程数据不足,且具有强非线性和多尺度特性的新间歇过程,结合迁移学习方法与多尺度核学习方法的优势,提出了一种基于多尺度核JYMKPLS(Joint-Y multi-scale kernel partial least squares)迁移模型的间歇过程产品质量在线预测方法。该方法首先通过迁移学习利用相似源域的旧过程数据提高新间歇过程建模效率和质量预测的精度。然后,针对间歇过程数据的非线性和多尺度特性问题,引入了多尺度核函数以更好地拟合数据变化的趋势,从而提高模型的预测精度。此外,提出模型在线更新和数据剔除,通过在线持续改善迁移模型对新间歇过程的匹配程度,以消除相似过程间的差异性给迁移学习带来的不利影响,从而不断地提升预测精度。最后,通过仿真验证了所提方法的有效性,结果表明,与传统的数据驱动建模方法相比,本文所提方法能够有效提高建模效率和预测精度。

关键词: 间歇式, 模型, 预测, 多尺度核, 迁移学习

Abstract:

In view of the shortage of process data and the strong nonlinear and multi-scale characteristics of the new batch process, a product quality prediction method based on multi-scale kernel JYMKPLS (Joint-Y multi-scale kernel partial least squares) transfer learning is proposed, which combines the advantages of transfer learning and multi-scale kernel learning. First, the new batch process modeling efficiency and quality prediction accuracy are improved by using the old process data in the similar source domain through transfer learning. Then, in order to solve the problem of non-linear and multi-scale characteristics of the data, multi-scale kernel method is used to better fit the data features, so as to improve the prediction accuracy of the model. In addition, the online update and data elimination of the model are proposed to continuously improve the matching degree of the transfer model to the new batch process, so as to eliminate the adverse effects of the differences between similar processes on the transfer learning, so as to continuously improve the prediction accuracy. Finally, the effectiveness of the proposed method is verified by simulation. The results show that, compared with traditional data-driven modeling methods, the method proposed in this paper can effectively improve modeling efficiency and prediction accuracy.

Key words: batchwise, model, prediction, multi-scale kernel, transfer learning

中图分类号: