Outlier detection of monitored data and unsupervised recognition of construction activities during seismic performance enhancement of historic stone monuments

Abstract To enhance the seismic resilience of historical and cultural heritage sites, protective measures were implemented through the installation of advanced heritage protection platform facilities. A structural health monitoring system was developed to safeguard historical relics during construction activities by continuously monitoring the overall condition of the relics and the integrity of critical components. Key parameters, such as settlement differences, tilt, crack width, and acceleration, were meticulously tracked, with predefined warning and alarm thresholds established. Alerts were triggered whenever these parameters exceeded their respective thresholds, ensuring timely interventions. To ensure the reliability and consistency of the collected data, this study proposes an evaluation method that integrates multi-source data fusion with statistical analysis techniques. Building on this foundation, an unsupervised algorithm was employed to identify construction activities impacting the structural health of the relics. The results demonstrate the effectiveness of combining multi-source data and intelligent algorithms for reliable monitoring and early detection of risks during conservation. The developed system offers automated, real-time assessments and can serve as a model for future heritage protection projects. Looking forward, integrating wireless sensors and diverse data sources could improve system accuracy, efficiency, and cost-effectiveness, further enhancing the protection of cultural heritage.