PSI - Issue 84

Available online at www.sciencedirect.com

ScienceDirect

Procedia Structural Integrity 84 (2026) 1119–1126

© 2026 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the scientific committee of the Conference Abstract Prestressed reinforced concrete bridge systems are widely used in medium-span applications due to their high structural efficiency and durability. However, early-stage damage detection, such as prestress loss and localized stiffness degradation, remains challenging, particularly under service-level loading where deterioration induces only subtle changes in the essentially linear global structural response. To address this challenge, a comprehensive full-scale experimental program is being conducted on a 20 m long bridge consisting of three prestressed concrete girders and a cast-in-place reinforced concrete deck slab. The structure is adopted as a benchmark case study for developing a reliable numerical model and will be subjected to incremental loading and controlled damage scenarios. A dense structural health monitoring system comprising accelerometers, strain gauges, load cells, inclinometers, and temperature sensors is deployed to capture static and dynamic responses, from which damage-sensitive features such as modal parameters, rotational responses, and strain distributions are systematically extracted. In parallel, a finite element model is developed and carefully calibrated against experimental data. Based on this calibrated model, a supervised multi-class damage classification framework is formulated using ensemble learning techniques. The classifiers are trained on parametrized simulations of healthy and damaged states that explicitly account for prestress loss and localized stiffness degradation. Four complementary base classifiers, namely multilayer perceptron, k-nearest neighbours, support vector machine, and random forest, are combined using soft and hard voting strategies. The results demonstrate that the proposed ensemble framework consistently outperforms individual classifiers, providing improved damage sensitivity and robust performance for structural health monitoring of prestressed concrete bridge systems. Future work will focus on testing the framework using real data from controlled damage scenarios on the bridge mock-up. III Fabre Conference: Existing Bridges, Viaducts, and Tunnels: Research, Innovation, and Applications An Ensemble-Learning Framework for Structural Damage Classification in Prestressed Bridge Deck Prajwal Giri a, *, Laura Ierimonti a , Francesco Mariani a , Enrique Garcìas-Macìas b , Manuel Boccolini c , Leonardo Casali c , Filippo Ubertini a , Ilaria Venanzi a a Department of Civil and Environmental Engineering, University of Perugia, Via Go ff redo Duranti 93, 06125 Perugia, Italy b Department of Structural Mechanics and Hydraulics, University of Granada, Campus de Fuentenueva, 18071 Granada, Spain c Manini Prefabbricati S.p.A., Via San Bernardino da Siena, 33, 06088 Santa Maria degli Angeli (PG), Italy

* Corresponding author. E-mail address: prajwal.giri@dottorandi.unipg.it

2452-3216 © 2026 The Authors. Published by ELSEVIER B.V. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-review under responsibility of the scientific committee of the Conference 10.1016/j.prostr.2026.06.143