A Distributed Monitoring Framework for Large-Scale Industrial Processes Based on Differential Grouping and...

Modern large-scale industrial processes, characterized by high dimensionality, strong nonlinearity, and complex spatiotemporal couplings, pose significant challenges to traditional monitoring methods. To address these challenges, this article proposes a novel distributed monitoring framework: differential grouping and hierarchical cooperative monitoring (DG-HCM). First, a differential grouping (DG) strategy leverages deep feature interactions to adaptively decompose the process into functionally cohesive sub-blocks. Second, a hierarchical cooperative monitoring system is constructed upon this decomposition: a multiscale convolutional autoencoder (MSCAE) captures intrablock dynamics; canonical correlation analysis (CCA) with a spatiotemporal context monitors interblock collaborations; and a global statistic integrating all CCA canonical scores assesses system-wide coordination. Finally, all local statistics are fused into a unified decision index via Bayesian inference, while a hierarchical contribution analysis enables precise traceability from faulty sub-blocks to root-cause variables. Extensive validations on a numerical simulation, the Tennessee Eastman (TE) process, and a real-world industrial wastewater treatment plant (WWTP) demonstrate that the proposed framework outperforms advanced benchmarks in both monitoring performance and diagnostic interpretability.