Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor | Faculty of Engineering | en_US |
| dc.contributor | Department of Electronic and Information Engineering | en_US |
| dc.contributor.advisor | Lun, Daniel (EIE) | en_US |
| dc.creator | Choy, Wai Hing Alex | - |
| dc.identifier.uri | https://theses.lib.polyu.edu.hk/handle/200/10983 | - |
| dc.language | English | en_US |
| dc.publisher | Hong Kong Polytechnic University | en_US |
| dc.rights | All rights reserved | en_US |
| dc.title | Deep learning techniques for structural health monitoring based on electromechanical impedance method | en_US |
| dcterms.abstract | In this research, new approaches that make use of the machine learning techniques for structural damage detection are developed. Structural damage detection is vital to the safety and integrity of civil and mechanical structures, such as the transportation system. Among different structural health monitoring (SHM) methods for detecting the structural damages, the electromechanical impedance (EMI) method is considered in this research as it is applicable to complex structures and is not based on any other models. It has high sensitivity to incipient damages in a structure. In particular, the EMI method is very suitable for applying to the bolt loosening monitoring problem, which is an essential SHM application in transportation carriers and civil structures. However, traditional statistical metric-based EMI methods often have problem in practical applications due to their sensitivity to the variation of the working environment, such as the change in temperature. Recently, machine learning-based EMI approaches are proposed to allow the damage detection process to be guided by robust statistics derived from large amount of training data. However, existing machine learning-based EMI approaches all adopt shadow-learning models which exhibit much difficulty when extracting the features of complex input signals. In this research, we propose to use deep learning-based methods, more specifically convolutional neural networks (CNN), to improve the robustness of the EMI applications. In this thesis, we first demonstrate the difficulty of the existing machine learning-based EMI method in a simple binary level Bolt Loosening Detection problem. For creating the testing environment, physical hardware setup is developed, and real EMI signals are acquired for the experiment. Then we propose the CNN-based EMI methods and show the significant improvement compared with the existing approach. Furthermore, a novel 1D-to-2D image encoding method is proposed for applying the 2D-CNN technique to the 1D spectral data. Our experimental results show that a prediction accuracy of over 95% can be achieved using the proposed CNN-based EMI methods in that Bolt Loosening Detection problem. Although the problem is relatively simple, the encouraging results have demonstrated the feasibility of the deep learning techniques in EMI applications. To further investigate the practicality of the proposed CNN-based EMI methods, we extend the simple binary level classification problem described above to a more practical multi-level Bolt Loosening Detection problem. Besides, we also suggest using a low-cost field-deployable device for EMI measurement. It simulates the practical situation since many EMI applications are outdoor and will not allow to use expensive and heavy laboratory-grade measurement equipment. For such a problem, it is found that the prediction accuracy decreases for the proposed CNN-based EMI methods. A multi-sensor fusion scheme is then proposed to improve the prediction accuracy. The scheme encodes the EMI signals contributed by different bolted nuts to the input images such that the CNN model can consider the signal patterns of all sensors at the same time. As a result, a prediction accuracy of about 92% is achieved. The promising result demonstrates the flexibility and extendibility of the proposed CNN-based EMI methods for tackling the SHM problem. | en_US |
| dcterms.extent | xvii, 126 pages : illustrations | en_US |
| dcterms.isPartOf | PolyU Electronic Theses | en_US |
| dcterms.issued | 2021 | en_US |
| dcterms.educationalLevel | Eng.D. | en_US |
| dcterms.educationalLevel | All Doctorate | en_US |
| dcterms.LCSH | Structural analysis (Engineering) | en_US |
| dcterms.LCSH | Structural health monitoring | en_US |
| dcterms.LCSH | Hong Kong Polytechnic University -- Dissertations | en_US |
| dcterms.accessRights | restricted access | en_US |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| 5447.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 7.14 MB | Adobe PDF | View/Open |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/10983