A cloud–edge collaborative hierarchical diagnosis framework for key performance indicator-related faults in manufacturing industries

作     者：Zhang, Xueyi Ma, Liang Peng, Kaixiang Zhang, Chuanfang Shahid, Muhammad Asfandyar Wang, Yangfan 

作者机构：Key Laboratory of Knowledge Automation for Industrial Processes of Ministry of Education School of Automation and Electrical Engineering University of Science and Technology Beijing Beijing 100083 China Key Laboratory of Metallurgical Industry Safety Risk Prevention and Control of Ministry of Emergency Management University of Science and Technology Beijing Beijing 100083 China 

出 版 物：《Journal of Process Control》 (J Process Control)

年 卷 期：2025年第152卷

学科分类：0711[理学-系统科学] 07[理学] 0817[工学-化学工程与技术] 08[工学] 0807[工学-动力工程及工程热物理] 070105[理学-运筹学与控制论] 0802[工学-机械工程] 081101[工学-控制理论与控制工程] 0701[理学-数学] 0811[工学-控制科学与工程] 071101[理学-系统理论] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：National Natural Science Foundation of China  NNSF  (U21A20483  62273031  62373040  62303041) 

主　　题：Cloud–edge collaborative Fault diagnosis Hierarchical cooperation and interaction Key performance indicator Manufacturing industries 

摘      要：In the context of intensifying global market competition and the accelerated advancement of industrial intelligence powered by the Industrial Internet of Things, manufacturing enterprises face pressing challenges in achieving sustainable development through quality and efficiency enhancement. Effective key performance indicators (KPIs) related fault diagnosis plays a crucial role in ensuring product quality stability and efficient production within modern manufacturing industries. However, manufacturing industries are characterized by numerous production sub-processes, hierarchical cooperation and interaction, and complex spatio-temporal features, making the implementation of comprehensive KPI-related fault diagnosis methods challenging. To overcome these challenges and fully leverage the hierarchical and multi-scale nature of manufacturing systems, an innovative hierarchical KPI-related fault diagnosis framework based on cloud–edge collaboration is proposed in this paper. First, a hierarchical information enhancement method utilizing dual-scale slow feature analysis and minimal gated units is developed to handle the multi-scale nature of system levels. Second, graph attention networks are combined with minimal gated units to capture the spatio-temporal dynamics across all levels, and KPI constraints are incorporated to fully extract the KPI-related spatio-temporal features. In addition, bottom-up propagation and top-down updation strategies are designed to facilitate information interaction between levels. Building on this, a cloud–edge collaborative architecture is developed, with specific tasks assigned each side. Finally, the framework is applied to a collaborative prototype system for cloud–edge- device in the hot rolling process, and its effectiveness and applicability are thoroughly evaluated. © 2025