Improved Infrastructure Assessments of Water Main Breaks Using Data Mining and Machine Learning Algorithms

点击翻译按钮获取中文摘要

Maintaining the integrity of buried water infrastructure is at the forefront of priority issues for municipalities across Canada. However, substantial portions of our buried water infrastructure are in need of repair/ rehabilitation, and possible replacement, all of which require significant expenditures. Increased vulnerability to imposed stresses as a result of climate change and urban densification, and avoiding compromised water quality, are examples of challenges complicating the prioritization of investments. In response, this research proposal is focused on developing guidance procedures to most effectively direct such expenditures. The merits of dramatically improved data mining and machine learning algorithms, and including the potential to improve the utilization of information now available from evolving smart data acquisition procedures becoming available, will be explored in this research. Research insights will include the merits of prioritization to include factors such as health risks, external factors of adjacency of high-cost impacts should pipe failure occur, climate change influencing water demand patterns, urban intensification and the concomitant increased demands that will arise, all of which contribute to the merits of a performance-based management approach for failure prediction assessments. Data mining models will be used to improve discovery of patterns, and machine language models will be employed to improve prediction efforts, all with the intent to improve pipe break prediction modeling and to provide better prioritization approaches for types of pipe rehabilitation. The utility of smart infrastructure technology will also be explored to determine the potential to decrease water distribution issues of pipe breaks and maintenance issues. This research will train HQP students in municipal engineering water infrastructure (condition evaluation, lifecycle costing, lifecycle prediction, deterioration modeling, machine learning, and data mining approaches). The proposed research will provide direct funding to assist in the training of three PhD and four masters students,one U/G, and a post-doc. The goal is to include the mentoring of grad students in research management and effective communication skills in this critical area of Canadian needs.