Anomaly discrimination via spatiotemporal correlation analysis of time-varying sensor systems under sparse traffic loads

Structural Health Monitoring (SHM) systems are crucial for safeguarding large-span bridges, yet accurately distinguishing structural damage from sensor faults remains a significant challenge, especially under sparse traffic loads where environmental effects dominate. To address this persistent problem in engineering practice, this study proposes a novel dual-discrimination framework combining spatial pattern recognition with correlation analysis for real-time SHM anomaly source identification. A core innovation involves incorporating a proximity constraint condition within the spatial anomaly assessment to prevent misclassification of widespread structural damage as sensor faults. Crucially, the framework leverages the spatial disruption patterns of correlated measurements and introduces specific constraints to differentiate between sensor malfunction signatures and actual structural damage indicators. The proposed method is rigorously evaluated using extensive field monitoring data acquired from the Heilongjiang River Bridge Health Monitoring System (HLJ-SHM). Experimental results demonstrate its superior effectiveness in accurately distinguishing sensor anomalies from structural damage compared to established baselines, achieving Accuracy of 92.3 %, Precision of 93.5 %, Recall of 89.7 %, and F1-score of 91.5 % in sparse traffic conditions. This research provides a robust and practical solution for enhancing the reliability of SHM-based early warnings, offering significant value for bridge maintenance decision-making.