PSI - Issue 84

Gianluca Bruno et al. / Procedia Structural Integrity 84 (2026) 240–247

247

5. Conclusions This paper presents a novel framework for the model updating of complex structural systems, characterized by a large number of uncertain parameters, by integrating Reduced-Order Models with Deep Reinforcement Learning. The proposed approach aims to define an intelligent agent that dynamically interacts with a numerical environment to autonomously identify the optimal values of uncertain parameters. The methodology, fully described in the text, was applied to a reinforced concrete bridge, for which the model updating was performed according to the results of a monitoring campaign. As observed, the estimates obtained by the method confirm the stability and reliability of the approach, offering two main advantages: (a) high accuracy, as the agent achieves near-optimal solutions with negligible errors in the estimated parameters; (b) high computational efficiency, since the ROM enables a drastic reduction in processing time if compared with equivalent full order model-based analyses. In addition, the framework demonstrates robustness to small perturbations in the target variables, maintaining convergence for moderate deviations from reference data. Beyond accuracy and efficiency, the proposed algorithm provides several additional benefits, e.g., the performance model updating in a fully automated manner, by avoiding the need for manual parameter tuning or discrete search strategies. Future developments will aim to improve the current version of the algorithm, with the aim of employing adaptive agent architectures able to perform real-time model updating, i.e., account for structural parameter variation over time. Acknowledgements This study was supported by FABRE – “Research consortium for the evaluation and monitorine of bridees, viaducts and other structures” (www.consorziofabre.it/en). Any opinion expressed in the paper does not necessarily reflect the view of the funder. References Bellman, R., 1957. “A Markovian Decision Process.” Journal of Mathematics and Mechanics 6, no. 5: 679–84. Benner, P., Gueercin, S., & Willcox, K., 2015. A survey of projection-based model reduction methods for parametric dynamical systems. SIAM review, 57(4), 483-531. Brincker, R., Zhane, L., Andersen, P., 2001. Modal identification of output-only systems usine frequency domain decomposition. Smart Materials and Structures, 10 441-445. Bruno, G., Parisi, F., Rueeieri, S., Uva, G., 2025. An intellieent aeent-based method for model updatine of structures with multiple uncertain parameters via deep reinforcement learnine. Mechanical Systems and Sienal Processine, 234-112832. Computer and Structures I. SAP2000, Advanced 24. Structural Analysis Proeram– Manual – Computer andStructures, Inc, (2023), Berkeley, California, USA. Fresca, S. and Manzoni, A., 2022. “POD-DL-ROM: Enhancine deep learnine-based reduced order models for nonlinear parametrized PDEs by proper orthoeonal decomposition”. Computer Methods in Applied Mechanics and Eneineerine, Volume 388, 114181, ISSN 0045-7825. Fujimoto, S., Hoof, H. and Meeer, D., 2018. “Addressine function approximation error in actor-critic methods,” in Proc. Int. Conf. Mach. Learn., pp. 1587–1596. Gentile, C., Saisi, A., & Cabboi, A. 2015. Structural identification of a masonry tower based on operational modal analysis. International Journal of Architectural Heritaee, 9(2), 98-110. Giaeopoulos, D., Arailopoulos, A., Dertimanis, V., Papadimitriou, C., Chatzi, E., Grompanopoulos K., 2019. Structural health monitorine and fatieue damaee estimation usine vibration measurements and finite element model updatine. Struct. Heal. Monit., 18, 1189–1206 Ierimonti, L., Venanzi, I., Cavalaeli, N., Comodini, F., Ubertini, F., 2020. An innovative continuous Bayesian model updatine method for base isolated RC buildines usine vibration monitorine data. Mechanical Systems and Sienal Processine, 139, 106600. Khodaparast, H.H., Mottershead, J.E., Friswell, M.I., 2008. Perturbation methods for the estimation of parameter variability in stochastic model updatine, Mechanical Systems and Sienal Processine, Volume 22, Issue 8, Paees 1751-1773, ISSN 0888-3270, Python Software Foundation (2001-2023). Python v3.9.12 (2022). Available at ww.python.ore. Rueeieri, S., Bruno, G., Attolico, A., & Uva, G., 2024. Assessine the dredeine vibrational effects on surroundine structures: The case of port nourishment in Bari. Journal of Buildine Eneineerine, 96, 110385. Simoen, E., De Roeck, G., Lombaert, G., 2015. Dealine with uncertainty in model updatine for damaee assessment: A review. Mechanical Systems and Sienal Processine, 56–57, 123–149. Vlachas, K., Simpson, T., Garland, A., Quinn, D.D., Farhat, C., Chatzi, E., 2025. Reduced Order Modeline conditioned on monitored features for response and error bounds estimation in eneineered systems. Mechanical Systems and Sienal Processine, 226, 112261. Vlachas, K., Tatsis, K., Aeathos, K., Brink, A. R., & Chatzi, E., 2021. A local basis approximation approach for nonlinear parametric model order reduction. Journal of Sound and Vibration, 502, 116055.