SHM is a cross-disciplinary domain that draws upon expertise from civil engineering, signal processing, sensor technology, ML, and data structures.
SHM's potential lies in its ability to conduct structural condition assessments, which minimizes uncertainties in design, maintenance, and management planning.
The evaluation of structural condition can serve both for damage detection and the assessment of remaining useful life.
Identifying and quantifying damage in structures using a reduced number of variables is a big deal.
The combination of influence lines-based methods to extract damage location with neural networks and genetic algorithms is a valuable solution.
Accurate damage localization and severity estimation.
Few measurement points-based damage identification method.
Robust damage identification method (Validation of structures with varying geometric properties ).
Development of advanced structural models.
Utilization of influence surfaces.
Exploration of innovative damage indicators.
Consideration of complex damage scenarios.
Marasco, G., Piana, G., Chiaia, B., and Ventura, G. “Genetic Algorithm Supported by Influence Lines and a Neural Network for Bridge Health Monitoring”, Journal of Structural Engineering, 9, 04022123 (2022).
Early detection of structural anomalies enables proactive maintenance, preventing both economic loss and potential harm to human life—especially for critical infrastructure like railway bridges. In long-term monitoring, static measurements (e.g., displacements, strains, rotations) and environmental data (e.g., temperature) can train machine learning models (e.g., Extreme Gradient Boosting, Multi-Layer Perceptron) to build regression models that represent the undamaged state of the structure.
Alarm thresholds are defined based on the expected residuals between predicted and measured structural quantities.
These thresholds effectively detect early-stage anomalies that traditional damage thresholds, based on design values, may overlook.
The proposed damage index, calculated as the moving median of the residuals, significantly reduces false alarms.
The correlation models serve as a foundation for extending this approach to other bridges of the same static typology.
Marasco, G., Oldani, F., Chiaia, B., Ventura, G., Dominici, F., Rossi, C., Iacobini, F., and Vecchi, A. “Machine learning approach to the safety assessment of a prestressed concrete railway bridge”, Structure and Infrastructure Engineering, 1-15 (2022).
Visual inspections do not reliably and objectively assess the conservation state of structures. Indirect techniques, such as ground-penetrating radar (GPR), have become fundamental to supporting limited direct tests. The analysis of the GPR tunnel lining profiles is mainly hand-operated. Artificial Intelligence (AI) can offer a classification framework to improve the efficiency of this powerful indirect surveying method.
Accurate CNN-based road tunnel assessment.
Accurate ViT-based road tunnel assessment.
Damage Localization
Rosso, M.M., Marasco, G., Aiello, S., Aloisio, A., Chiaia, B., and Marano, G.C. “Convolutional networks and transformers for intelligent road tunnel investigations”, Computers & Structures, 275, 106918 (2023).
Chiaia, B., Marasco, G., and Aiello, S. “Deep convolutional neural network for multi-level non-invasive tunnel lining assessment”, Frontiers of Structural and Civil Engineering, 16(2), 214-223 (2022).
Continuous monitoring by crowdsourcing allows unprecedented data volumes, increased time and space resolutions, repeatable data collection, and very low cost.
System identification using data streams from mobile sensors.
Study of vehicle bridge interaction, impact of vehicle dynamics, and impact of road profile.
Cronin, L., Sen, D., Marasco, G., Matarazzo, T., and Pakzad, S. “Bridge monitoring using mobile sensing data with traditional system identification techniques”, Computer-Aided Civil and Infrastructure Engineering, (2024).
Cronin, L., Sen, D., Marasco, G., Dabbaghchian, I., Benedetti, L., Matarazzo, T., and Pakzad, S. “A roadmap for ubiquitous crowdsourced mobile sensing-based bridge modal identification”, Sensors (2025).
Bridge authorities normally own a huge amount of similar bridges, like the pedestrian ones. This creates the opportunity to transfer knowledge from one or a small group of well-known bridges to help make more effective decisions in new bridges. Transfer learning offers a novel solution to periodically evaluate the structural condition of all pedestrian bridges using long-term monitoring data from one or a small group of pedestrian bridges.
Analysis of the impact of geometric and material parameters on Transfer Learning performance.
Successful application of unsupervised transfer learning on data from two similar pedestrian prestressed concrete bridges, using two domain adaptation techniques.
The domain adaptation approach improved classification accuracy for the target domain, compared to methods without domain adaptation, by reducing the distribution mismatch between the source and target domains.
Definition of the similarity concept for several bridge typologies.
Marasco, G., Moldovan, I., Figueiredo, E., and Chiaia, B. “Unsupervised transfer learning for structural health monitoring of urban pedestrian bridges”, Journal of Civil Structural Health Monitoring, 1-17 (2024).
Life-Cycle Assessment
Traditionally, strain or stress time histories are used to evaluate the remaining fatigue life of the structure. Obtaining these required measurements often poses challenges, including difficulties in accessing critical locations, measurement errors, significant labor costs, and other related issues. To overcome such limitations, the use of virtual sensing has proven to be highly effective.
Prediction of strain signals with artificial neural networks.
Toward physics-informed artificial neural networks.
Marasco, G., Sen, D., Sadeghi Eshkevari, S., Dabbaghchian, I., and Pakzad, S. “A framework for bridge condition assessment using Graph Neural Network”, Bridge Maintenance, Safety, Management, Digitalization and Sustainability - Proceedings of the 12th International Conference on Bridge Maintenance, Safety and Management, IABMAS 2024, 2528–2535 (2024).