Article: Optimal residual subspace model for structural damage diagnosis: an approach independent of operational and environmental variations Journal: International Journal of Structural Engineering (IJSTRUCTE) 2022 Vol.12 No.1 pp.44 - 61 Abstract: Several machine learning algorithms have been proposed to detect damage in civil structures that implicitly learn changes in dynamic characteristics of structures due to varying operational and environmental conditions. However, despite the intensive computational load, the methods were not able to completely mitigate the said variations. In contrast to that, here, we introduce a new methodology based on percentage of total variance (PTV) criterion for damage detection to overcome the influence of varying operational and environmental conditions on the vibration-based damage sensitive features. Using PTV criterion, an optimal residual subspace (ORS) is modelled which is applied to Mahalanobis squared distance (MSD) and singular value decomposition (SVD)-based damage detection methods. Interestingly, we find that in comparison with similar machine learning-based damage detection approaches, the proposed approach outperforms in terms of false positive reduction and overall accuracy. Inderscience Publishers - linking academia, business and industry through research

Title: Optimal residual subspace model for structural damage diagnosis: an approach independent of operational and environmental variations

Authors: Kundan Kumar; Sumanshu Agarwal

Addresses: Department of Electronics and Communication Engineering, Institute of Technical Education and Research, Siksha 'O' Anusandhan (Deemed to be University), Bhubaneswar-751030, India ' Department of Electronics and Communication Engineering, Institute of Technical Education and Research, Siksha 'O' Anusandhan (Deemed to be University), Bhubaneswar-751030, India

Abstract: Several machine learning algorithms have been proposed to detect damage in civil structures that implicitly learn changes in dynamic characteristics of structures due to varying operational and environmental conditions. However, despite the intensive computational load, the methods were not able to completely mitigate the said variations. In contrast to that, here, we introduce a new methodology based on percentage of total variance (PTV) criterion for damage detection to overcome the influence of varying operational and environmental conditions on the vibration-based damage sensitive features. Using PTV criterion, an optimal residual subspace (ORS) is modelled which is applied to Mahalanobis squared distance (MSD) and singular value decomposition (SVD)-based damage detection methods. Interestingly, we find that in comparison with similar machine learning-based damage detection approaches, the proposed approach outperforms in terms of false positive reduction and overall accuracy.

Keywords: structural health monitoring; damage detection; outlier detection; optimal residual space; percentage of total variance; PTV; eigenspace; Mahalanobis squared distance; MSD; singular value decomposition; SVD.

DOI: 10.1504/IJSTRUCTE.2022.119285

International Journal of Structural Engineering, 2022 Vol.12 No.1, pp.44 - 61

Received: 19 Feb 2021
Accepted: 25 Feb 2021
Published online: 30 Nov 2021 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article

Title: Optimal residual subspace model for structural damage diagnosis: an approach independent of operational and environmental variations

Keep up-to-date