RMIT University
Browse
- No file added yet -

Artificial Intelligence-based Prediction of Bridge Rating

Download (2.1 MB)
thesis
posted on 2024-05-14, 04:18 authored by Irvin Ming Yii Wong
With the worsening situation of bridge conditions in Victoria, Australia, and the increasing potential of utilizing artificial intelligence with viable results due to the advancement of knowledge and technology, the construction sector stands to benefit greatly. Not only was it reported that bridge conditions have been deteriorating, there has been an increasing backlog of bridges which has not been inspected in over 5 years; the longest allowable timeframe set by VicRoads. The current standard practice for bridge data recording in Victoria includes obtaining data using visual inspections and manually inputting in the VicRoads Bridge Inspection System (BIS), which are then uploaded into the VicRoads Road Asset System (RAS) database alongside other structural road fixtures such as culverts and signages. Solely relying on visual inspections to deduce the bridges’ condition does not allow governments or asset owners to prepare budgets and forecast labour requirements ahead of time for bridge maintenance. Bridge management systems such as Pontis and BRIDGIT does include a bridge rating prediction modelling which utilizes Markov Chain. However, this is probabilistic-based in its approach and does not include physical parameters. This study aims to investigate the feasibility of implementing artificial neural network to improve future predictions of bridge conditions in Victoria. With bridge data collected via VicRoads online databases and physical site inspections, an artificial neural network analysis was run using the Multi-Layer Perceptron method. Results shows that the model is able to determine a pattern and predict future bridge conditions and subsequently, prioritization of bridges is performed. The results of prioritization depends on the accuracy of available data. Therefore, comprehensive record keeping of past bridge conditions and affecting factors are essential in ensuring proper estimations. The methodology of utilizing artificial neural network for bridge rating prediction is a novel approach in the field of civil engineering and infrastructure asset management.

History

Degree Type

Masters by Research

Copyright

© Irvin Ming Yii Wong 2023

School name

Engineering

Usage metrics

    Theses

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC