An adaptive distributed architecture for state estimation and control of integrated process networks during operational transitions

This paper presents an adaptive framework for distributed state estimation and control of integrated chemical processes undergoing operational transitions. A spectral community detection method designed for dynamic graph representation of these processes is employed for system decomposition. To overcome the computational challenges of frequent reclustering, a novel policy is introduced that selectively updates specific nodes during each cluster adjustment. By limiting the number of nodes and using a greedy search algorithm, the proposed approach mitigates locality issues often encountered in modularity optimization. The effectiveness of this method is demonstrated through its application to output tracking in a benzene alkylation process during three distinct operational conditions. The results indicate that the dynamic graph-based community detection technique improves efficiency compared to frequent reclustering.