Pipeline condition monitoring using depth camera technology
File(s)
Author(s)
Wong, Boon
Type
Thesis
Abstract
In the Housing and Development Board (HDB) housing estates in Singapore, there has been an increasing number of pneumatic waste conveyance systems (PWCS) being commissioned to manage waste collection more efficiently. In a survey conducted at a PWCS network that had only been in service for 4 years, moderate to severe degradation, caused by abrasive corrosion induced by the fast movement of refuse up to 80 km/h, was observed at certain areas of the pipelines. The requirement to evaluate the rate of degradation of these PWCS pipelines motivated the research in this thesis, specifically the need to develop a computer vision technique that is not only capable of accurately reconstruct high-fidelity digital models of actual pipeline networks, but also providing data to HDB to develop corrosion growth model for predictive maintenance using data analytics. In this thesis, the suitability of the depth camera technology in the in-pipe environment is investigated using a robotic prototype, deployed in a decommissioned PWCS pipeline segment. The baseline accuracy of measurement of the depth camera is benchmarked using artificially made defects with known dimensions made using a computerised numerical control (CNC) cutting machine. A data processing scheme is designed to convert the raw depth point cloud from a depth camera to a continuous mesh to reconstruct a high-fidelity 3-dimensional model of an actual pipe with the accuracy of defect modelling up to 85\% of the actual depths and surface areas. To the best of our knowledge, this thesis presents the first 3-dimensional model of the inside of an actual pipe using a depth camera that could be deployed via a robot through a single access point. Using the methods and the conditions proposed in this thesis, the best accuracy to date of the quantitative characterisation of sub-centimetre defects has been achieved.
Version
Open Access
Date Issued
2023-05-16
Date Awarded
01/03/2024
Copyright Statement
Creative Commons Attribution NonCommercial Licence
License URL
Advisor
A. McCann, Julie
Sponsor
Ministry of National Development Singapore
National Research Foundation Singapore (Trust)
Grant Number
L2NICTDF1-2017-3
Publisher Department
Department of Computing
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)