| Author: | Gong, Yi |
| Title: | Model-free damage identification of high-speed railway enabled by on-line monitoring and Bayesian framework |
| Advisors: | Ni, Yi-qing (CEE) |
| Degree: | M.Sc. |
| Year: | 2019 |
| Subject: | High speed trains Railroads -- Maintenance and repair Railroad engineering Hong Kong Polytechnic University -- Dissertations |
| Department: | Department of Civil and Environmental Engineering |
| Pages: | xiii, 148 pages : color illustrations |
| Language: | English |
| Abstract: | Claimed as the "4 new great inventions", high-speed railway (HSR) has not only significantly shortened the travel times between cities but also pushed the development of social and various economical industry chains. Whereas, due to the harsh environment and frequently operation schedule, the components of railway engineering would encounter high probability of failure and further affect the safety of the train's operation, passenger comfort. Thus, during the operation and maintenance stage, the Structural health monitoring (SHM) of the HSR is necessary and vital. Generative adversarial network (GAN) and variational auto-encoder (VAE) are two kinds of generative models in terms of latent variables, which are amenable to extracting intrinsic feature and handling uncertainty inherent in data. Recent years have witnessed the huge successful progress of generative models in many domains, including image generation, text generation and other fascinating applications. However, the application of them in the structural health monitoring field is so rare. In this research work, the emphasis is put on the built of framework for wheel defect detection based on VAE. The proposed framework contains following major parts: data normalization, sliding window, model training and damage detection. The VAE for structural health monitoring using only dynamic response data that contains uncertainty of different sources is explored based on a reconstruction-based theory. The monitoring data collected under healthy condition is used to train a model in the auto-encoding framework, and the structural health index based on reconstruction probability is operated as a core indicator for identifying possible structural damage when a new set of data from an unknown status of the structure is available. Lower reconstruction probability refers to the existence of flat spot. In order to illustrate the applicability of the proposed method, a thorough case study on the detection of high-speed train wheel defects using the online monitoring data of dynamic strain on the rail track collected during the passage of the train is provided. Through comparison between the predicted location of wheel flat spot and the off-line inspection results, the proposed framework is proved to be able to estimate the relative location of the wheel flats around the wheel's circumference. This dissertation proves that the generative model could be exploited into the damage identification for Structural health monitoring and remained a lot of potential to be discovered in the future. |
| Rights: | All rights reserved |
| Access: | restricted access |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| 5274.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 3.72 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/200/10837