Online anomaly detection for long-term structural health monitoring of caisson quay walls

To assess the current state and develop maintenance strategies for proactive management of infrastructure, research on Structural Health Monitoring (SHM) has been actively performed. For port facilities, the need for sensor-based monitoring is increasing to analyze the effects of various factors such as aging, ship activities, backfill earth pressure, and waves. However, few researchers have conducted long-term monitoring of caisson quay walls. In this study, an SHM system was developed with different types of sensors installed on two caisson quay walls and monitored over one year. A new online adaptive anomaly detection approach was proposed to identify the anomalous status of each caisson in real-time by analyzing multiple variables. The method was validated with seven simulated anomaly scenarios, demonstrating high accuracy in anomaly detection despite significant environmental variations, outperforming other approaches. These results highlight the potential to provide timely and accurate alerts when anomalous states occur in port structures. • A first step toward the development of long-term SHM on caisson-type quay walls. • Development of anomaly detection methodology for autonomous long-term SHM of port facility. • Adaptive online learning algorithm incorporating multivariate sensors on each caisson. • Robust performance and high accuracy under large environmental variations with timely alerts when an anomalous state occurs.