Answers are grounded in this paper and cite their sources.
Industrial cyber-physical systems (ICPSs) generate interconnected process, control, and communication data in which distributed failures can develop through weak abnormalities across multiple components. This study proposes “Explainable Temporal Graph Learning for Early Detection and Propagation Tracing of Distributed Failures in Industrial Cyber-Physical Systems” to address the limitations of isolated anomaly detection, particularly in identifying failure origins and propagation sequences. The proposed framework represents industrial components as dynamic graph nodes and their physical, functional, process, and communication dependencies as time-varying edges. Multivariate observations are synchronized and transformed into temporal graph sequences, followed by temporal graph learning, anomaly scoring, failure-origin identification, propagation tracing, and attention-based explanation. A simulation-based evaluation is designed using representative public industrial-control datasets and controlled synthetic perturbations to model known propagation scenarios. The simulated results indicate an average F1-score of 0.94 with a mean detection delay of approximately 21 s. Failure-origin Top-1 accuracy reaches 91.7%, while propagation-path accuracy reaches 89.4%. Attention and perturbation analyses further indicate that influential nodes and dependencies can be associated with the detected propagation sequence. Compared with isolated anomaly analysis, the proposed approach provides system-level temporal information about where a failure begins, how it propagates, and which dependencies influence the prediction. The framework can support early-warning systems, root-cause analysis, reliability assessment, and incident investigation in interconnected industrial environments.
No takes yet. Share an insight, caveat, or question.
Abdur Rouf (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: