A Robust PCA-Based Framework for Long-Term Condition Monitoring of Civil Infrastructures

Mousavi, M; Gandomi, AH

A Robust PCA-Based Framework for Long-Term Condition Monitoring of Civil Infrastructures

Mousavi, M

Gandomi, AH

Permalink

Publisher:: Springer International Publishing
Publication Type:: Chapter
Citation:: Data Science in Engineering, Volume 9, 2022, pp. 79-85
Issue Date:: 2022-01-01

Open Access

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is open access.

The embargo period expires on 1 Jan 2024

Adobe PDF

Download Accepted versionAdobe PDF (348.59 kB)

View on publisher's site

View statistics

Full metadata record

Field	Value	Language
dc.contributor.author	Mousavi, M https://orcid.org/0000-0002-0436-5176
dc.contributor.author	Gandomi, AH
dc.date.accessioned	2023-01-19T08:22:45Z
dc.date.available	2023-01-19T08:22:45Z
dc.date.issued	2022-01-01
dc.identifier.citation	Data Science in Engineering, Volume 9, 2022, pp. 79-85
dc.identifier.isbn	9783031041211
dc.identifier.uri	http://hdl.handle.net/10453/165259
dc.description.abstract	This chapter proposes an output-only method for condition monitoring of civil infrastructures through studying a couple of lowest structural natural frequency signals. The main challenge in this sort of problem is to mitigate the effect of the Environmental and Operational Variations (EOV) on the structural natural frequencies to avoid misinterpretation of these effects as damage. To this end, a robust Principal Component Analysis (PCA)-based approach is proposed that uses a couple of lowest structural natural frequency signals obtained from vibration data over a long period of time. First, the proposed method utilizes a truncated transformation matrix of the robust local PCA of a portion of the dataset corresponding to the healthy state of the structure to remove the EOV effects by mapping the dataset to a new space. The difference between the mapped signals and the original signals is deemed to minimize the effect of the EOV. As such, extracting the mapped data from the original data, termed error signals, will remove the EOV effects and can be further used for damage detection. To this end, the Mahalanobis distances of the errors in the test set from the distribution of the errors in the baseline data are used for condition monitoring through constructing a Hotelling (T2) control chart. The proposed PCA-based method does not apply the covariance matrix and the mean vector of the entire dataset, but instead the Minimum Covariance Determinant (MCD) algorithm, in its fast mode (FastMCD), is employed to obtain a robust covariance matrix and mean vector of the dataset. It is shown through solving the benchmark problem of the Z24 bridge that the proposed method can effectively increase the accuracy of the damage detection compared with the case when the normal PCA is used.
dc.language	en
dc.publisher	Springer International Publishing
dc.relation.ispartof	Data Science in Engineering, Volume 9
dc.relation.isbasedon	10.1007/978-3-031-04122-8_9
dc.rights	info:eu-repo/semantics/embargoedAccess
dc.title	A Robust PCA-Based Framework for Long-Term Condition Monitoring of Civil Infrastructures
dc.type	Chapter
pubs.organisational-group	/University of Technology Sydney
pubs.organisational-group	/University of Technology Sydney/Faculty of Engineering and Information Technology
utslib.copyright.status	open_access	*
utslib.copyright.embargo	2024-01-01T00:00:00+1000Z
dc.date.updated	2023-01-19T08:22:44Z
pubs.publication-status	Published

Abstract:

This chapter proposes an output-only method for condition monitoring of civil infrastructures through studying a couple of lowest structural natural frequency signals. The main challenge in this sort of problem is to mitigate the effect of the Environmental and Operational Variations (EOV) on the structural natural frequencies to avoid misinterpretation of these effects as damage. To this end, a robust Principal Component Analysis (PCA)-based approach is proposed that uses a couple of lowest structural natural frequency signals obtained from vibration data over a long period of time. First, the proposed method utilizes a truncated transformation matrix of the robust local PCA of a portion of the dataset corresponding to the healthy state of the structure to remove the EOV effects by mapping the dataset to a new space. The difference between the mapped signals and the original signals is deemed to minimize the effect of the EOV. As such, extracting the mapped data from the original data, termed error signals, will remove the EOV effects and can be further used for damage detection. To this end, the Mahalanobis distances of the errors in the test set from the distribution of the errors in the baseline data are used for condition monitoring through constructing a Hotelling (T2) control chart. The proposed PCA-based method does not apply the covariance matrix and the mean vector of the entire dataset, but instead the Minimum Covariance Determinant (MCD) algorithm, in its fast mode (FastMCD), is employed to obtain a robust covariance matrix and mean vector of the dataset. It is shown through solving the benchmark problem of the Z24 bridge that the proposed method can effectively increase the accuracy of the damage detection compared with the case when the normal PCA is used.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/165259