The chemical industry is the pillar and basic industry of my country's national economy. It is the pioneer field of supply-side structural reform of the manufacturing industry and the main battlefield for green development. However, there is still a gap between the safety of my country's chemical industry production processes and the international advanced level. An important reason is the lack of chemical process fault diagnosis technology to ensure the safe and efficient operation of the production process. Factor-level chemical processes are generally large-scale complex processes coupled with multi-phase and multi-physical fields, and mechanism modeling is difficult. With the development of data collection technology, massive data containing process information has been accumulated in industrial databases. Therefore, how to timely and accurately realize fault detection, fault location, and fault identification based on massive process data has become a basic common problem that needs to be solved urgently in chemical process fault diagnosis. Aiming at the characteristics of factory-level chemical processes involving many variables and operating units, and complex and diverse process characteristics, this project conducts in-depth research on the concise and efficient characterization of data-driven operating states of factory-level chemical processes based on massive process data, and innovatively proposes data-driven factory-level chemical processes. Basic theory of fault diagnosis for chemical processes. The main findings are: (1) Systematically analyzed the impact of high-dimensional variable structures in factory-level chemical processes on process fault diagnoses, innovatively proposed data-driven subsystem segmentation based on the characteristics of variable structures and block-based fault diagnosis methods for complex process characteristics, forming a basic theory of data-driven subsystem segmentation and block-based fault diagnosis;(2) Systematically analyze the impact of redundant features on fault diagnosability in classic multivariate statistical fault diagnosis, demonstrate the necessity of information enrichment and feature selection, and innovatively propose a fault diagnosis method of instant feature optimization and fault information enhancement, and build a data-driven fault information enrichment to strengthen the basic theory of fault diagnosis;(3) The necessity of integrating fault information for data dimension reduction and model streamlining is systematically analyzed, and innovatively proposed variable optimization distributed fault monitoring for fault information supervision, and a visual fault identification method for feature extraction and dimension reduction combined with topology preservation mapping. Basic theory of factory-level process fault diagnosis that integrates fault data information. This project has published 45 SCI journal papers for applied basic research, including papers published in IEEE TIE., AIChE J.、and J. Process Control. and other internationally renowned (JCR Q1, Q2) publications in engineering and technology fields such as control engineering and chemical system engineering published 37 articles. He cited 8 representative papers in SCI 188 times, making a total of 310 citations, and 1 highly cited paper in ESI. Related papers were awarded by the internationally renowned fault diagnosis expert and S. Professor Joe Qin (IEEE/IFAC/AIChE Fellow), Steven X. Professor Ding, Professor Zhou Donghua, Director of the Fault Diagnosis Expert Committee of the China Automation Society, Professor Haibo He, Editor-in-Chief of IEEE TNNLS, Professor Biao Huang, Academician of the Canadian Academy of Engineering, Professor Sheng Chen, Academician of the British Academy of Engineering, Professor Chai Tianyou and Professor Qian Feng, Academicians of the China Academy of Engineering, etc. Positive quotes and positive comments. The theoretical results of this project are applied to the quality monitoring and fault diagnosis of industrial units such as Tongcheng Xincai phenolic resin production process of listed companies and the continuous casting process of China Baowu Group, forming a series of factory-level industrial process fault monitoring and diagnosis systems to ensure safe and efficient operation of the production process. 3 national invention patents were disclosed and 1 computer software copyright was registered. Cultivate 5 doctoral students and 10 master's students.