PSI - Issue 78

Available online at www.sciencedirect.com

ScienceDirect

Procedia Structural Integrity 78 (2026) 387–394

© 2025 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 XX ANIDIS Conference organizers Keywords: Structural health monitoring; masonry structures; strain measurements; numerical modelling; deep learning; neural network. Abstract The vulnerability of historic masonry structures to seismic events represents a significant challenge for the preservation of built her itage. SHM techniques o ff er a promising solution for detecting degradation and damage progression, enabling timely and targeted retrofit interventions. However, the limited availability of real-world labeled data related to damage conditions hinders the develop ment of reliable predictive models for structural anomaly detection. To address this limitation, this work proposes a methodology based on finite element micromechanical modelling of an archetypal masonry panel and a realistic masonry frame, consisting of a typical facade, both subjected to in-plane loading to simulate seismic-induced damage. Specifically, damage is simulated at various levels of severity, representing di ff erent stages of structural degradation. Monitoring data are generated numerically by simulating the use of smart bricks, a new type of strain sensor for SHM of masonry structures. The information obtained from the archetypal panel is used to train a Domain-Adversarial Neural Network, enabling knowledge transfer from the archetypal panel to the fa cade model. This domain adaptation strategy e ff ectively overcomes data scarcity and enhances the generalization capability of the damage classifier. The obtained results demonstrate the e ff ectiveness of the method in detecting structural anomalies. Overall, the proposed approach shows strong potential for real-world SHM applications in historic masonry structures, supporting continuous monitoring. XX ANIDIS Conference Domain adversarial neural network for strain-based seismic damage detection in masonry structures: first proposal and preliminary results Alina Elena Eva a, ∗ , Andrea Meoni a , Valentina Giglioni a , Ilaria Venanzi a , Filippo Ubertini a a Department of Civil and Environmental Engineering, University of Perugia, Italy

1. Introduction

The built heritage of Italy and Europe is largely made up of masonry structures, many of which are in poor condition due to aging materials and lack of maintenance. Their location in seismic zones further increases their vulnerability to earthquake-induced damage. For this reason, e ff ective monitoring strategies are essential to protect these structures. Structural Health Monitoring (SHM) techniques make it possible to assess the behavior of masonry buildings under

∗ Corresponding author. Tel.: + 39-328-329-9142. E-mail address: alinaelena.eva@dottorandi.unipg.it

2452-3216 © 2025 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 XX ANIDIS Conference organizers 10.1016/j.prostr.2025.12.050