From traditional damage detection methods to Physics-Informed Machine Learning in bridges: A review

Abstract

Structural Health Monitoring (SHM) of bridges plays a crucial role in infrastructure management, ensuring the safety and durability of bridges under diverse operational and environmental conditions. A vital aspect of SHM involves Structural Damage Detection (SDD), which focuses on identifying, localizing, and quantifying structural damage such as cracks, corrosion, and other forms of deterioration. While traditional SDD methods, including physics-based and Machine Learning (ML) methods, are effective, they often tend to be challenging in addressing the complex and dynamic nature of bridge systems, particularly when dealing with limited or noisy data. Physics-Informed Machine Learning (PIML) has emerged as a promising approach that integrates the strengths of ML with the reliability of physical constraints and principles, offering more accurate, robust interpretability and generalization capabilities, thereby strengthening the SHM framework. This paper provides a comprehensive overview of the evolution of SHM, from traditional SDD methods to the application of PIML. By analyzing key case studies and examining the strengths and limitations of each method, this review highlights the potential of PIML to address the challenges of real-world bridge monitoring and improve the early detection of structural damage.