A Semi‐Supervised Graph‐Based Method for Structural Damage Detection.
Saved in:
| Title: | A Semi‐Supervised Graph‐Based Method for Structural Damage Detection. |
|---|---|
| Authors: | Mansouri Hanis, Fateme1 (AUTHOR), Ghodrati Amiri, Gholamreza1 (AUTHOR) ghodrati@iust.ac.ir, Darvishan, Ehsan2 (AUTHOR), De Michele, Carlo3 (AUTHOR), Khandelwal, Manoj (AUTHOR) m.khandelwal@federation.edu.au |
| Source: | Advances in Civil Engineering. 7/26/2026, Vol. 2026, p1-17. 17p. |
| Subjects: | Structural health monitoring, Graph algorithms, Model validation, Machine learning, Fault diagnosis, Feature extraction, Supervised learning, Signal processing |
| Abstract: | Data‐driven and machine learning‐based methods have proven effective in structural health monitoring (SHM). However, most rely on supervised learning, which demands sufficient labeled data from healthy and damaged structures, often impractical in real‐world scenarios. Semi‐supervised learning provides a potential solution to this challenge by requiring less labeled data to generate accurate predictions. This study presents a novel SHM framework using semi‐supervised local and global consistency (LGC) algorithm. The acceleration signals are further processed to extract features in the time, frequency, and time–frequency domains using fast Fourier transform (FFT) and short‐time Fourier transform (STFT). The features are used as inputs to the LGC algorithm. The algorithm is first trained on a small labeled dataset and then the labels are propagated from this dataset to the remaining unlabeled dataset using a graph‐based approach. LGC is compared to three supervised algorithms including support vector machine (SVM), multilayer perceptron (MLP), and k‐nearest neighbors (K‐NNs), as well as two semi‐supervised algorithms including label propagation and self‐training. Experimental validation is conducted on three case studies: a numerical model bridge, a laboratory‐scale grandstand model, and a real cable‐stayed bridge. The experimental results show that the proposed method enhances the accuracy of the damage detection, and significantly decreases the labeling cost. The contributions of this study are three‐fold: (i) the novel adaptation of LGC to SHM, (ii) the analytical comparison of the feature domains, and (iii) a proven semi‐supervised SHM framework over several case studies, thus expanding its applicability in real‐world scenarios. [ABSTRACT FROM AUTHOR] |
| Copyright of Advances in Civil Engineering is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | Data‐driven and machine learning‐based methods have proven effective in structural health monitoring (SHM). However, most rely on supervised learning, which demands sufficient labeled data from healthy and damaged structures, often impractical in real‐world scenarios. Semi‐supervised learning provides a potential solution to this challenge by requiring less labeled data to generate accurate predictions. This study presents a novel SHM framework using semi‐supervised local and global consistency (LGC) algorithm. The acceleration signals are further processed to extract features in the time, frequency, and time–frequency domains using fast Fourier transform (FFT) and short‐time Fourier transform (STFT). The features are used as inputs to the LGC algorithm. The algorithm is first trained on a small labeled dataset and then the labels are propagated from this dataset to the remaining unlabeled dataset using a graph‐based approach. LGC is compared to three supervised algorithms including support vector machine (SVM), multilayer perceptron (MLP), and k‐nearest neighbors (K‐NNs), as well as two semi‐supervised algorithms including label propagation and self‐training. Experimental validation is conducted on three case studies: a numerical model bridge, a laboratory‐scale grandstand model, and a real cable‐stayed bridge. The experimental results show that the proposed method enhances the accuracy of the damage detection, and significantly decreases the labeling cost. The contributions of this study are three‐fold: (i) the novel adaptation of LGC to SHM, (ii) the analytical comparison of the feature domains, and (iii) a proven semi‐supervised SHM framework over several case studies, thus expanding its applicability in real‐world scenarios. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 16878086 |
| DOI: | 10.1155/adce/6935378 |
