Work

Physics-informed machine learning for the structural health monitoring and early warning of a long highway viaduct with displacement transducers

Year: 2025

Type: article

Abstract: In Bridge Structural Health Monitoring (SHM), damage detection is often hindered by environmental and operational variability. Confounding influences such as traffic, wind, and especially temperature can significantly affect measured structural responses, making it difficult to distinguish true damage-related anomalies. This is particularly critical in the static monitoring of long steel-made viaducts, where thermal effects dominate the observed displacements, making small damage-induced perturbations almost impossible to notice in the raw data. To address this challenge, the present study introduces a predictive Physics-Informed Machine Learning (PIML) model, aiming to establish a reliable baseline for the ‘normal conditions’ of the monitored infrastructure. This depurated baseline can then be used to properly isolate anomalies attributable to structural damage, while accounting for temperature-induced movements. The proposed grey-box approach combines data-driven modelling with physical knowledge of thermal behaviour, enhancing both accuracy and interpretability. A real-world application is demonstrated on a long-span highway viaduct, where longitudinal displacements are monitored using temperature and displacement sensors. By using only temperature and time as input variables, the model effectively captures the nonlinear effects of daily and seasonal thermal cycles without relying on additional instrumentation. Furthermore, to assess the model’s reliability, an Early Warning System (EWS) is developed, based on displacement anomaly thresholds and using Kernel Density Estimation (KDE). The PIML model is evaluated against both black-box (purely data-driven) and white-box (purely physics-based) alternatives. Damage scenarios are simulated by introducing anomalies into the experimental data to test each model's capability to detect abnormal behaviour while filtering out environmental influences. The results show that the grey-box PIML model consistently outperforms the black- and white-box approaches in terms of accuracy, robustness, and anomaly discrimination. These findings highlight the potential of PIML methods to enhance SHM practices and support the development of automated, reliable EWSs for bridge SHM.

Source: Mechanical Systems and Signal Processing

DOI: 10.1016/j.ymssp.2025.113659

Authors: Enrico Cianci, Marco Civera, Valerio De Biagi, Bernardino Chiaia

Institution: Politecnico di Torino

Language: English

Topic: Structural Health Monitoring Techniques

Subfield: Civil and Structural Engineering

Field: Engineering

Domain: Physical Sciences

SDGs: Industry, innovation and infrastructure, Sustainable cities and communities

Open Access status: hybrid

Estimated APC paid: $4,730