文献类型：期刊文献

英文题名：A high-accuracy deep learning framework for digital twin model development of actual chemical processes

作者：Li, Yue[1,2];Li, Zhongmei[3];Ren, Jingzheng[4];Du, Wenli[5];Shen, Weifeng[1,5]

机构：[1]Chongqing Univ, Sch Chem & Chem Engn, Chongqing 400044, Peoples R China;[2]Guangxi Minzu Univ, Sch Chem & Chem Engn, Nanning 530006, Peoples R China;[3]East China Univ Sci & Technol, Key Lab Smart Mfg Energy Chem Proc, Minist Educ, Shanghai 200237, Peoples R China;[4]Hong Kong Polytech Univ, Dept Ind & Syst Engn, Hong Kong, Peoples R China;[5]East China Univ Sci & Technol, State Key Lab Ind Control Technol, Minist Educ, Shanghai 200237, Peoples R China

年份：2025

卷号：159

外文期刊名：ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE

收录：;EI(收录号：20252918810993);WOS:【SCI-EXPANDED(收录号:WOS:001554893000003)】；

基金：We acknowledge the financial support provided by the National Natural Science Foundation for Excellent Young Scientists of China (No. 22122802) ; the National Natural Science Foundation of China (No. 22278044) ; the Chongqing Science Fund for Distinguished Young Scholars (No. CSTB2022NSCQ-JQX0021) ; the Chongqing Innovation Support Key Program for Returned Overseas Chinese Scholars (cx2023002) ; and the Open Research Project of the State Key Laboratory of Industrial Control Technology, China (Grant No. ICT2024B01) .

语种：英文

外文关键词：Digital twin modeling; Deep learning; Chemical process model; Intelligent factory; Interpretable deep learning framework

摘要：The development of intelligent chemical industry urgently calls for high-accuracy digital twin modeling methods of actual chemical processes. However, although intelligent technologies such as deep learning are utilized, digital twin modeling is challenged by the complexity and nonlinearity of actual chemical processes, leading to unexpected deviations and poor generalization. To provide a valuable modeling methodology in this field, a novel deep learning framework based on chemical process mechanisms is developed. According to the properties of different neural networks, the framework is designed in a hierarchical way to learn the underlying process mechanisms and deal with the nonlinearity of actual chemical processes. Thus, a digital twin model of the actual chemical process can be developed, which is high-accuracy and good-generalization. Based on the sensor data from the distributed control system of an actual industrial distillation process, the proposed deep learning digital twin modeling framework is verified by thoroughgoing and careful analyses, including the modeling test, ablation experiment and residue distribution analysis of model results. Compared with wide-spread baseline models, the new framework decreases the predicting mean absolute error by 45.6 % and increases the coefficient of determination by 13.8 % on average. This intelligent framework reveals desirable application potentials that it can create a high-fidelity digital twin of actual chemical processes. Such digital representation acts as a basic model, which can support advanced manufacturing tasks such as real-time optimization by detailed digital inference. This work provides an intelligent digital twin modeling methodology in chemical engineering and facilitates the development of intelligent chemical industry.