A Dynamic Process Modeling Method Based on Bipartite Graph and Recursive Monitoring for Catalytic Cracking Unit

Mingliang Cui, Youqing Wang, Jipeng Guo, Tongze Hou, Xin Ma

Abstract: Catalytic cracking unit is a highly dynamic industrial process involving complex chemical reactions, catalyst circulation, temperature and pressure changes, and control system regulation. To address the challenges of dynamic process modeling and monitoring, this study proposes a sparse dynamic matrix estimation method based on a bipartite graph (SDMEM-BG) structure. The proposed SDMEM-BG innovatively constructs a bipartite graph of the dynamic relationships between variables and guides the sparse optimization process of the dynamic matrix through the graph structure, making the identified sparse dynamic matrix more reflective of the intrinsic dynamic relationships of the system and substantially reducing computational complexity. Furthermore, this study derives a recursive update strategy for the dissimilarity monitoring indicator and conducts online monitoring through a local sliding window, thus achieving quick calculation of the monitoring indicators. Experiments are conducted on numerical examples and control valve data from a chemical plant, demonstrating the effectiveness of the proposed SDMEM-BG.

External IDs:dblp:journals/tcst/CuiWGHM25