Railway bridge condition monitoring and fault diagnostics

Vagnoli, Matteo (2019) Railway bridge condition monitoring and fault diagnostics. PhD thesis, University of Nottingham.

[img]
Preview
PDF (Thesis - as examined) - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (14MB) | Preview

Abstract

The European transportation network is ageing continuously due to environmental threats, such as traffic, wind and temperature changes. Bridges are vital assets of the transportation network, and consequently their safety and availability need to be guaranteed to provide a safe transportation network to passenger and freights traffic. The main objective of this thesis is to develop a bridge condition monitoring and damage diagnostics method. The main element of the proposed Structural Health Monitoring (SHM) method is to monitor and assess the health state of a bridge continuously, by taking account of the health state of each element of the bridge. In this way, an early detection of the ongoing degradation of the bridge can be achieved, and a fast and cost-effective recovery of the optimal health state of the infrastructure can be achieved.

A BBN-based approach for bridge condition monitoring and damage diagnostics is proposed and developed to assess and update the health state of the bridge continuously, by taking account of the health state of each element of the bridge. At the same time, the proposed BBN approach allows to detect and diagnose damage of the bridge infrastructure.

Firstly, the BBN method is developed for monitoring the condition of two bridges, which are modelled via two Finite Element Models (FEMs). The Conditional Probability Tables (CPTs) of the BBN are defined by using an expert knowledge elicitation process. Results shows that the BBN allows to detect and diagnose damage of the bridges, however the performance of the BBN can be improved by pre-processing the data of the bridge behaviour and improving the definition of the CPTs.

A data analysis methodology is then proposed to pre-process the data of the bridge behaviour, and to use the results of the analysis as an input to the BBN. The proposed data analysis methodology relies on a five-step process: i) remove of the outlier of the bridge data; ii) identify of the free-vibration of the bridge; iii) extract statistical, frequency-based and vibration -based features from the free-vibration behaviour of the bridge; iv) assess the features trend over time, by using the extracted features as an input to an Empirical Mode Decomposition (EMD) algorithm; v) evaluate of the Health Indicator (HI) of the bridge element. The proposed data analysis methodology is tested on two in-field bridges, a steel truss bridge and a post-tensioned concrete bridge, which are subject to a progressive damage test.

A machine learning method is also developed in order to assess the health state of the bridge automatically. A Neuro Fuzzy Classifier (NFC) is adopted for this purpose. The results of the NFC can potentially be used as an input to the BBN nodes, to select the states of the BBN nodes, and improve the BBN performance. In fact, the NFC shows high accuracy in assessing the health state of bridge elements. An optimal set of HIs, which allows to maximize the accuracy of the NFC, is identified by adopting an iterative Modified Binary Differential Evolution (MBDE) method. The NFC is applied to the post-tensioned concrete in-field bridge that is subject to a progressive damage test.

Hence, the performance of the BBN is improved significantly by pre-processing the bridge data, but also by developing a novel method to continuously update the CPTs of the BBN. The CPTs update process relies on the actual health state of the bridge element, and the knowledge of bridge engineers. Indeed, the CPT updating method aims to merge the expert knowledge with the analysis of the bridge behaviour. In this way, the diagnostic ability of the BBN is improved by merging the expertise of bridge engineer, who can analyse hypothetical damage scenarios of the bridge, and the analysis of a database of known bridge behaviour in different health states. The method is verified on the post-tensioned concrete in-field bridge, by developing a BBN to monitor the health state of the bridge continuously. The damages of the bridge are diagnosed by the proposed BBN.

Finally, a method to analyse database of unknown infrastructure behaviour is finally proposed. An ensemble-based change-point detection method is presented to analyse a database of past unknown infrastructure behaviour. The method aims to identify the most critical change of the health state of the infrastructure, by providing the characteristics of such a change, in terms of time duration and possible causes. The method is applied to a database of tunnel behaviour, which is subject to renewal activities that influence the health state of the infrastructure.

Item Type: Thesis (University of Nottingham only) (PhD)
Supervisors: Remenyte-Prescott, Rasa
Andrews, John
Keywords: Structural Health Monitoring, Bayesian Belief Network, Machine Learning methods, damage detection and diagnostics, data drive approach
Subjects: T Technology > TA Engineering (General). Civil engineering (General) > TA 630 Structural engineering (General)
Faculties/Schools: UK Campuses > Faculty of Engineering
Item ID: 56175
Depositing User: Vagnoli, Matteo
Date Deposited: 18 Jul 2019 04:40
Last Modified: 07 May 2020 12:01
URI: https://eprints.nottingham.ac.uk/id/eprint/56175

Actions (Archive Staff Only)

Edit View Edit View