PSI - Issue 78

Alina Elena Eva et al. / Procedia Structural Integrity 78 (2026) 387–394

394

4. Conclusion

This work presents a methodology for SHM of masonry structures that combines smart brick sensor technology with artificial intelligence, specifically domain-adversarial neural networks (DANN). A damage-sensitive classifier was developed using data from a finite element model of a masonry panel, where several damage scenarios—including bending, shear, and compression—were simulated. The classifier demonstrated strong performance in detecting and distinguishing between these scenarios, validating its integration into a DANN framework. The classifier was subsequently integrated into the DANN network to enable knowledge transfer from the source domain—fully labeled—to the target domain, which lacks labels. This transfer was explored using a full-scale ma sonry facade model. Although a low level of damage was introduced in the target domain, the DANN network suc cessfully detected the presence of damage, demonstrating good sensitivity to structural anomalies. However, due to the limited extent of the damage, the network was unable to accurately identify its specific type. This behavior highlights both the robustness of the method in detecting anomalies and its limitations in low-damage scenarios, particularly in the absence of calibrated material properties. These preliminary results confirm the feasibility of using domain adaptation techniques for the SHM of masonry structures, especially in the context of historical heritage, where real labeled data are often scarce or entirely unavail able. The proposed approach enables the transfer of damage-related information learned in a simulated domain to a real structure, thus reducing the need for extensive experimental campaigns. However, the limited accuracy in identi fying specific damage types under low-severity conditions highlights the need to enrich the classifier with a broader variety of damage scenarios and severity levels. Incorporating this diversity in the training dataset will enhance the model’s ability to discriminate between subtle and more significant structural anomalies. Future developments will focus on improving material calibration, extending simulations to more realistic damage scenarios in the fac¸ade, and further training the DANN network to improve classification accuracy across a wider range of structural states.

Acknowledgements

The authors would like to gratefully acknowledge the support of the Italian Ministry of University and Research (MUR) via the FIS2021 Advanced Grant “SMS-SAFEST - Smart Masonry enabling SAFEty-assessing STructures after earthquakes” (FIS00001797). Dr. Meoni also acknowledges the European Union - NextGenerationEU and the University of Perugia for supporting his research activity through the project Vitality framed within the Italian Ministry of University and Research (MUR) National Innovation Ecosystem grant ECS00000041 - VITALITY.

References

Di Trapani, F. et al, 2024, Structural health monitoring of masonry structures using stress sensors: Experimental induced damage tests and proposed approach for real-time monitoring. Construction and Buildings Materials. 449. Meoni, A. et al, 2021, Strain Field reconstruction and damage identification in masonry walls under in pane loading using dense sensor network of smart bricks: Experiments and simulations. Engineering Structures . 239. Meoni, A. et al, 2020, Characterization of the strain-sensing behavior of smart bricks: A new theoretical model and its application for monitoring of masonry structural elements. Construction and Buildings Materials. 250. Ozdagli, I.A and Koutsoukos, X, 2020 Characterization of the strain-sensing behavior of smart bricks: A new theoretical model and its application for monitoring of masonry structural elements. Construction and Buildings Materials. 12. Giglioni, V. et al, 2025, An Adversarial-Based Method for Knowledge Transfer Across Bridges. International journal of bridge engineering man agement and research. 2. Garden, P. et al, 2020, On the application of domain adaptation in structural health monitoring. Mechanical Systems and Signal Processing. 138. D’altri, A.M. et al, 2018, A 3D detailed micro-model for the in-plane and out-of-plane numerical analysis of masonry panels. Computers and Structures. 206. Drougkas, A., Roca, P., Molins, C., 2015. Numerical prediction of the behavior, strength and elasticity of masonry in compression. Engineering Structures. 90, 15–28. Meoni, A. et al, 2025. Automated damage detection in masonry structures using cointegrated strain measurements from smart bricks: Application to a full-scale building model subjected to foundation settlements under changing environmental conditions. Journal of Building Engineering. 100. ABAQUS, ® Theory Manual(1999),Version 6.6.