A SYSTEMATIC REVIEW OF ARTIFICIAL INTELLIGENCE APPLICATIONS IN STRUCTURAL HEALTH MONITORING OF CIVIL INFRASTRUCTURE

Abstract

Structural Health Monitoring (SHM) has evolved from traditional periodic inspection methods to intelligent, data-driven systems powered by artificial intelligence (AI). This systematic review synthesizes recent advances in AI applications for SHM across civil infrastructure including bridges, buildings, tunnels, and dams. A comprehensive literature search covering 2020-2025 across Web of Science, Scopus, and IEEE Xplore identified 31 high-quality studies for qualitative analysis. The review examines AI methodologies including convolutional neural networks (CNNs), recurrent neural networks (RNNs), long short-term memory (LSTM) networks, and physics-informed neural networks (PINNs) applied to damage detection, localization, classification, and prognostic tasks. Sensor technologies—including accelerometers, fiber optics, piezoelectric transducers, and vision-based systems—are integrated with IoT platforms and digital twins for real-time monitoring. Key findings indicate that AI-driven SHM achieves damage detection accuracies exceeding 95%, with significant reductions in inspection time and maintenance costs. However, critical challenges persist: data scarcity in operational environments, model generalization across diverse infrastructure types, uncertainty quantification, cybersecurity vulnerabilities, and limited model interpretability for safety-critical decisions. Physics-informed and hybrid AI-physics models emerge as promising solutions to enhance robustness and transferability. This review identifies interdisciplinary opportunities including federated learning for decentralized monitoring, explainable AI for stakeholder trust, and autonomous inspection systems. Future research must prioritize standardized benchmark datasets, field-validated deployment studies, and sustainable practices for infrastructure resilience in the context of aging, climate-stressed, and increasingly complex built environments.