A Machine Learning-based Framework for Preventive Maintenance of Sewer Pipe Systems

• Author / Creator

  Yin, Xianfei

• The municipal drainage system is a key component of every modern city’s infrastructure. However, as the drainage system ages, its pipes gradually deteriorate at rates that vary based on the conditions of utilization (i.e., intrinsic conditions) and other extrinsic factors such as the presence of trees with deep roots or the traffic load above the sewer lines, which collectively can impact the structural integrity of the pipes. As a result, regular preventive maintenance of the drainage system is extremely important since replacement is not only costly, but, more importantly, can disturb the daily routines of citizens. As for preventive maintenance, closed-circuit television (CCTV) inspection has been widely accepted as an effective inspection technology for buried infrastructure. In practice, the CCTV inspection of sewer pipes is scheduled by the municipal department according to a set of criteria that prioritize the pipe sections and the frequency of the visits. After the onsite CCTV inspection process, the CCTV video footage is sent to the offsite office, where technologists (trained professionals) must watch through the entirety of the video footage in order to assess the condition of the corresponding pipes. When the technologists watch the video, pipe defects (such as cracks, fractures, roots, deposits, broken, and holes) that appear in the video need to be classified according to certain standardized nomenclature (such as PACP, WRc). After recording the existing defects, a pipe score is calculated, which will serve as the foundation for scheduling the pipe inspection and repair plan in the future. The above-mentioned preventive maintenance process is a time-consuming and costly operation. In the meantime, machine-learning technologies and computing power have increased rapidly in recent years and both are used in various engineering areas to improve productivity and the level of automation. In this context, this research proposes a machine learning-based framework to facilitate the preventive maintenance of sewer pipe systems. The ultimate goal of the research is to improve the productivity, consistency, and automation of the CCTV inspection-based sewer pipe preventive maintenance. To accomplish this aim, the following five objectives targeting on the optimization of each step of the maintenance process guide the activities of the research: 1) Develop a data-driven framework for modeling the productivity of the CCTV recording process for sewer pipes. This involves modelling the video recording process to predict the video duration. 2) Develop a deep learning-based framework for an automated defect detection system for sewer pipes. The targeted defects and construction features will be detected and labeled in the CCTV video. 3) Develop a video interpretation algorithm and corresponding software program. The text information, defect information, and other information included in the CCTV video will be exported to a tabulated format (e.g., EXCEL, database), which will serve as the source of information for pipe rating purposes. 4) Develop and analyze a data-driven bi-level sewer pipe deterioration model, which will provide city managers with a basis for scheduling preventive maintenance at the neighborhood- and individual-level. 5) Develop an input model of the CCTV inspection data of sewer pipes by examining the inherent characteristics of the historical dataset employing Markov chain-based random generation and validation. The generated dataset will be the input for scheduling preventive maintenance in a situation where there is insufficient data. The outcome of this research is expected to make significant contributions by proposing a machine learning-based framework for sewer pipe preventive maintenance by reducing human work, increasing productivity, and increasing assessment consistency.

• Subjects / Keywords

  • Sewer Pipe
  • CCTV
  • Preventive Maintenance
  • Machine Learning
  • Deterioration

• Graduation date

  Fall 2020

• Type of Item

  Thesis

• Degree

  Doctor of Philosophy

• DOI

  https://doi.org/10.7939/r3-vm4h-2w93

• License

  Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.