Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.629310
Title: Damage and repair identification in reinforced concrete beams modelled with various damage scenarios using vibration data
Author: Al-Ghalib, A. A.
Awarding Body: Nottingham Trent University
Current Institution: Nottingham Trent University
Date of Award: 2013
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
This research aims at developing a novel vibration-based damage identification technique that can efficiently be applied to real-time large data for detection, classification, localisation and quantification of the potential structural damage. A complete testing procedure of the Experimental Modal Analysis (EMA) in freely supported beam based on impact hammer, as a relevant excitation source for field measurements, was established and the quality of its measurements was ensured. The experimental data in this research was collected from five laboratory-scale reinforced concrete beams modelled with various ranges of common defects. Reliable finite element beam models for the five beams in their normal conditions were developed correlated and updated using the results of the experimental tests. As a first round of investigation of the damage identification methods, the results of the modal parameters along with a number of their formulations and combinations were evaluated as model-based damage characterisation systems. Different ways for the representation and visualisation of the measurements in the time- or frequency-domain in a format pertinent for pattern identification were assessed. A two-stage combination between principal component analysis and Karhunen-Loéve transformation (also known as canonical correlation analysis) was proposed as a statistical-based damage identification technique. The suggested technique attempts to detect features regarding outliers or variation in the structural dynamic behaviour. In addition, it is used to serve as an unsupervised classification tool for data representing different structural conditions. Vibration measurements from time- and frequency-domain were tested as possible damage-sensitive features in an effort to avoid the expensive prolonged calculations of the modal parameters. In the first stage of the algorithm, principal component analysis is conducted on data from frequency response functions or response power spectral density functions in order to reduce the size of the data. The first prominent principal components that account for a reasonable percentage of the variance in the original data are preserved. In the second stage, the important principal components are provided as inputs to Karhunen-Loéve transformation to constitute the new transformed space. Within-class and between-class covariance matrices are exploited for maximising the discriminant capacity between subgroups. The new generated sets of data are analysed as a typical mathematical eigenproblem to account for the first two or three principal components that retain the major part of the variance. These components are next being employed for significant visualisation of the original data. The proposed system would provide unsupervised means that is capable to process, compare and discriminate between different periodically-collected immense data without considerable unnecessary effort for computations and modelling. The results of this statistical system help in distinguishing between normal and damaged patterns in structural vibration data. Most importantly, the system further dissects reasonably each main group into subgroups according to the levels of damage. The performance of this technique was credibly tested and verified on real measurements collected from the five beams with various detailed damage states. Its efficiency was conclusively proved on data from both frequency response functions and response-only functions. The outcomes of this two-stage system show realistic detection and classification and outperform results from the rival principal component analysis-only.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.629310  DOI: Not available
Share: