PSI - Issue 84

Prajwal Giri et al. / Procedia Structural Integrity 84 (2026) 1119–1126

1126

The ensemble achieves the highest AUC ( > 0 . 996), followed by MLP and RF (0.995), SVM (0.976), and KNN (0.952). Its curve lies closest to the ideal upper-left corner, indicating superior global discriminative capability and an e ff ective balance between detection accuracy and misclassification rate. Overall, the soft voting ensemble provides the most stable and balanced classification across structural conditions, highlighting the benefit of combining complementary learners to reduce model bias and variance, and delivering improved robustness, generalization, and reliability over individual classifiers for bridge damage classification. Conclusions This study presents a supervised ensemble-based damage classification framework for a full-scale prestressed bridge deck under multiple damage scenarios. By integrating numerical simulation with ML, five structural states with increasing damage severity are accurately classified. The ensemble approach enhances predictive reliability by combining complementary base learners and demonstrates strong generalization across varying damage conditions. In particular, the soft voting ensemble consistently outperforms individual classifiers, confirming its e ff ectiveness for robust structural health monitoring. The main findings are summarized as follows: • The proposed framework, particularly the soft voting ensemble, achieves superior performance with an overall accuracy of 96.0% and an AUC of 0.998. • The framework maintains stable and balanced classification across all damage states (DS0–DS4), even when individual models exhibit diverse predictive behaviour. As the current approach is trained in a supervised manner, future work will focus on evaluating its performance on real structures and addressing domain mismatch between numerical models and field measurements. Acknowledgment The authors gratefully acknowledge the support of Manini Prefabbricati S.p.A. and the Italian Ministry of U v y U j “ –Time Evolution Laws for Structural Reliability of Post- C ” . P20223Y947), as well as FABRE within the FABRE–ANAS 2021–2026 research program. The views expressed in this paper do not necessarily reflect those of the funding bodies. References García-Macías, E., & Ubertini, F. (2020). MOVA/MOSS: Two integrated software solutions for comprehensive structural health monitoring of structures. Mechanical Systems and Signal Processing, 143 , 106830. https://doi.org/10.1016/j.ymssp.2020.106830 . Gautam, D., Bhattarai, A., & Rupakhety, R. (2024). Machine learning and soft voting ensemble classification for earthquake induced damage to bridges. Engineering Structures, 303 , 117534. https://doi.org/10.1016/j.engstruct.2024.117534 . Giri, P., Ierimonti, L., Ubertini, F., Venanzi, I., & García Macías, E. (2025). Multi-class damage classification in prestressed reinforced concrete bridges using multilayer perceptron artificial neural networks: A numerical case study. In Proceedings of the relevant CRC Press volume (1st ed.). CRC Press. https://doi.org/10.1201/9781003595120-155 . Li, Q., & Song, Z. (2022). Ensemble-learning-based prediction of steel bridge deck defect condition. Applied Sciences, 12 . https://doi.org/10.3390/app12115442 . Lu, Q., Zhu, J., & Zhang, W. (2020). Quantification of fatigue damage for structural details in slender coastal bridges using machine learning-based methods. Journal of Bridge Engineering, 25 . https://doi.org/10.1061/(ASCE)BE.1943-5592.0001571 . Mangalathu, S., Hwang, S. H., Choi, E., & Jeon, J. S. (2019). Rapid seismic damage evaluation of bridge portfolios using machine learning techniques. Engineering Structures, 201 . https://doi.org/10.1016/j.engstruct.2019.109785 . Sconocchia, G. G., Mariani, F., Ierimonti, L., Meoni, A., Venanzi, I., & Ubertini, F. (2024). Insights into structural assessment and long-term effects of a post-tensioned multispan concrete box girder bridge with vertically prestressed internal joints. Structures, 63 , 106396. https://doi.org/10.1016/j.istruc.2024.106396 . Yaghoubzadehfard, A., Lumantarna, E., Herath, N., Sofi, M., & Rad, M. (2024). Ensemble learning-based structural health monitoring of a bridge using an interferometric radar system. Journal of Civil Structural Health Monitoring, 14 , 1629–1650. https://doi.org/10.1007/s13349-024 00789-7 . Zhou, Q., Ning, Y., Zhou, Q., Luo, L., & Lei, J. (2013). Structural damage detection method based on random forests and data fusion. Structural Health Monitoring, 12 , 48–58. https://doi.org/10.1177/1475921712464572 . Zhou, Q., Zhou, H., Zhou, Q., Yang, F., & Luo, L. (2014). Structure damage detection based on random forest recursive feature elimination. Mechanical Systems and Signal Processing, 46 , 82–90. https://doi.org/10.1016/j.ymssp.2013.12.013 . Zhou, Z.-H. (2012). Ensemble methods: Foundations and algorithms (1st ed.). Chapman and Hall/CRC. https://doi.org/10.1201/b12207 .