This paper presents a spectral community detection method for dynamic graph representations of integrated chemical process networks. Although online community detection can be performed via frequent reclustering, this approach is computationally prohibitive for integrated systems. To address this limitation, the proposed method introduces a preclustering stage and a novel update policy that selectively evaluates a subset of nodes, thereby substantially reducing computational effort. Case studies demonstrate that focusing on strategically selected nodes and using a greedy search algorithm mitigates the locality problem inherent in modularity optimization. The effectiveness of the proposed method is validated through simulations of an output tracking problem in a benzene alkylation process under operational changes. Results indicate that the proposed adaptive graph-based community tracking approach achieves improved computational efficiency and modularity compared to frequent reclustering.
Ebrahimi et al. (Thu,) studied this question.