Knowledge graph-based software agent for monitoring data real-time diagnosis and reconstruction: A cable net monitoring case

Real-time diagnosis and reconstruction are essential for enhancing the quality of monitoring data in structural health monitoring (SHM) processes. This study aims to employ artificial intelligence techniques to automatically and continuously diagnose and reconstruct SHM data in real time. To this end, a novel framework is proposed that integrates the surrogate model, knowledge graph (KG), shapes graph (SG), and software agent. In this framework, the surrogate model trained by data-driven methods enables real-time prediction of reconstructed targets, KG acts as a hub for dynamically recording monitoring data and sensor observation statuses, SG serves as a rule repository for anomaly diagnosis, and software agent functions as a controller that continuously invokes system functions and makes decisions. Furthermore, a reconstruction mode recognition mechanism is constructed based on these components for the framework, thus tackling the issue of surrogate models becoming ineffective due to random anomalies within partial input features. The proposed methodology is applied to a monitoring case of cable net structure to detail the implementation process, and its efficacy is demonstrated through the developed service evaluation. The testing results confirm that the service can accurately identify predefined abnormal data and their reconstruction scenarios. The software agent then correctly selects the appropriate surrogate model to recover abnormal data, with a maximum reconstruction error of only 0.64%. Meanwhile, the service completes data diagnosis and reconstruction within 1.15 s per timestamp, validating its compliance with real-time performance demands.