Tram track degradation prediction

Hitihamillage, L 2018, Tram track degradation prediction, Masters by Research, Engineering, RMIT University.

Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Hitihamillage.pdf Thesis application/pdf 2.34MB
Title Tram track degradation prediction
Author(s) Hitihamillage, L
Year 2018
Abstract Transport organisations have historically focused on major construction and expansion of infrastructure. After completing the expansion of transport networks, the emphasis has increasingly shifted from developing new infrastructure to intelligently maintaining the existing ones. In recent years, economic constraints have influenced budget allocation to transport sectors. This has resulted in emphasising the development of maintenance management systems in transport sectors, particularly in transport infrastructure. Maintenance management systems assist organisations in deciding when and how to maintain transport infrastructure facilities to enhance cost efficiency.

The Melbourne tram network is the largest urban tram network in the world, consisting of around 250 kilometres of double track and running 31,400 scheduled tram services every week (Yarra Trams 2018). Yarra Trams operates and maintains Melbourne's iconic tram network on behalf of Public Transport Victoria. Yarra Trams organises the timetables, service deliveries and changes, tram arrival information via tram TRACKER, tram maintenance, as well as the construction and maintenance of the tram infrastructure. Many parameters are involved in ensuring that the Melbourne tram system operates to its safe and best practice standards. Track infrastructure is one of the fundamental elements of the tram system. The condition of the tram infrastructure can influence the system¿s operation, either directly or indirectly. To keep the track infrastructure in a reasonable condition over years and to obtain the most benefit out of its life cycle, a maintenance and renewal regime is required. The provision of a maintenance plan to recover the serviceability of tram tracks from defects and damages and preventing further wear is essential for such a large network. Currently, tram track maintenance activities are achieved by manual inspections across the network. Yarra Trams usually has a fixed number of maintenance teams who are responsible for visual inspection of the status of the tram tracks and identification of whether tracks need maintenance. Furthermore, they estimate an approximate time period during which maintenance should be carried out. Since the inspections are done visually, human error is inevitable. Mistakes in the inspection and detection of track faults and inaccurate prediction of maintenance time frames are challenges of the current maintenance system. In addition, prioritising maintenance projects is often a significant challenge. Poorly planned maintenance schedules may result in high maintenance and operational costs due to very early or late maintenance of tram tracks. Occasional unnecessary maintenance and replacement of tram tracks or maintenance at very late stages of damage is very costly.

The case study for this research is the Melbourne tram network and the necessary data for the research were collected from the entire network. The aim of this research is to develop an artificial intelligence model (Adaptive Network-based Inference System (ANFIS)) to predict conditions of track in the future years based on the most influential parameters identified via statistical analysis and a literature review. According to the literature and the statistical analysis, gauge and total annual loading have been utilized as the two key parameters in the model development. The total data set was randomly divided into two separate sets which were used as the training and testing sets. Two ANFIS models are developed for straight and curve sections. The model developed through this research is capable of predicting the future gauge values with an r-square value for the curve model of 0.60 while that of the straight model is around 0.78. A simple Artificial Neural Network (ANN) model is then proposed which managed to produce an r-square value of 0.4587 for curves and 0.5813 straight sections.
Degree Masters by Research
Institution RMIT University
School, Department or Centre Engineering
Subjects Transport Engineering
Keyword(s) artificial neural network
adaptive network-based fuzzy inference system
fuzzy inference system
million gross tons
neural network
Version Filter Type
Access Statistics: 76 Abstract Views, 93 File Downloads  -  Detailed Statistics
Created: Wed, 06 Feb 2019, 07:59:03 EST by Keely Chapman
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us