文献类型：期刊文献

英文题名：A method for integrated monitoring of process multiple indicators based on quality-aware network

作者：Chen, Zhiyun[1];Cheng, Xin[2,3];Tan, Shuai[1];Qiu, Yuan[2,3];Wang, Jiayi[1];Pan, Zhen[1];Zhang, Zhiya[1]

机构：[1]East China Univ Sci & Technol, Shanghai, Peoples R China;[2]Shanghai Aerosp Elect Technol Inst, Shanghai, Peoples R China;[3]Shanghai Key Lab Collaborat Comp Spatial Heterogen, Shanghai, Peoples R China

年份：2026

外文期刊名：CANADIAN JOURNAL OF CHEMICAL ENGINEERING

收录：;EI(收录号：20262721026189);WOS:【SCI-EXPANDED(收录号:WOS:001806167700001)】；

基金：This research is sponsored by the National Natural Science Foundation of China (62273147), Fundamental and Interdisciplinary Disciplines Breakthrough Plan of the Ministry of Education of China (JYB2025XDXM402), the Foundation of Shanghai Key Laboratory of Collaborative Computing in Spacial Heterogenous Networks (CCSN-2026-08).

语种：英文

外文关键词：fault detection; graph fusion; graph neural networks; multiple quality indicators

摘要：The running status of production process often contains multiple evaluation indicators, including economy, output, energy consumption, and other evaluation indicators. Complex industrial process monitoring needs to consider the correlation and support of information between multiple indicators. This paper proposes a method for integrated monitoring of process multiple indicators based on quality-aware network (QAN). This method introduces a correlation measure to screen the variables related to the indicators, and uses the self-attention mechanism to mine the interconnection relationship between the variables, establishing the 'quality-source graph' of multiple indicators. The method starts from the quality source graph space which is strongly correlated with the index, and uses the structural topology which contains the interconnection relationship between the process variables to realize the division of the index correlation feature space, which improves the fault detection ability of the method for single index correlation. Two quality-source graph fusion methods are proposed to address varying emphasis on multi-index monitoring in actual production: experience setting and machine self-learning. Experiments on Tennessee Eastman process and ammonia synthesis process showed that QAN can fuse the quality source map through experience setting and self-learning mechanism, and realize the comprehensive monitoring of process operation.