Machine Learning-Driven Structural Health Monitoring in Civil Engineering: A Review of Algorithms, Sensor Networks, and Predictive Maintenance Frameworks

Abstract

Structural health monitoring (SHM) plays an essential role in the safety, reliability and performance of civil engineering structures. The traditional method of monitoring structures has been revolutionized in recent years by new developments in machine learning, intelligent sensor networks, and predictive maintenance. This is a narrative review on the use, potential and drawbacks of machine learning algorithms, sensor technologies and predictive maintenance frameworks in civil engineering. Structural damage detection, anomaly identification, condition assessment and performance prediction are considered in the case of supervised, unsupervised and deep learning approaches. Wireless sensor networks, Internet of Things based monitoring systems and multi sensor data fusion for continuous data acquisition of structures are also discussed in the review. Beyond this, the concept of predictive maintenance using digital twins and intelligent decision support systems is also discussed to improve the planning of maintenance and infrastructure resilience. The use of bridges, buildings, dams, tunnels and civil structures are deemed to be areas that illustrate opportunities for implementation. Challenges include few labelled monitoring data, unreliable sensors, lack of understanding of the models, computational needs, and lack of real-world validation. These limitations can be addressed by new sensing paradigms such as explainable AI, physics-informed learning, and autonomous sensing and digital twin integration. Finally, the review highlights the promising prospects of SHM for more reliable, efficient, and sustainable infrastructure and structural systems using ML, intelligent sensing, and predictive maintenance systems.