ORCID Identifier(s)

0000-0002-0188-0842

Graduation Semester and Year

Spring 2025

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Civil Engineering

Department

Civil Engineering

First Advisor

Mohammad Najafi

Second Advisor

Jessica Eisma

Third Advisor

Michelle Hummel

Fourth Advisor

Yonghe Liu

Abstract

Based on optimistic scenarios by national agencies, by the mid-21 century, water and wastewater utilities will be expected to achieve a pipeline replacement rate that exceeds 2\% to counter the effect of deterioration of our water infrastructure. However, even in the best-case scenarios, determining which assets require replacement monitoring remains a complex task since utilities strive to maximize return on investment from installed asset value and may prefer to replace near the end of asset's remaining useful life. Therefore, determining the optimal time to replace a watermain ultimately relies on effective Condition Assessment (CA) and Condition Monitoring (CM) technologies capable of providing insights into asset distress characteristics.

For decades, utilities have used CM technologies for their most strategic assets, such as raw-water mains or transmission mains. Because of the severe consequences of pipe failures in transmission systems, most systems require some level of monitoring; however, these systems are often designed to higher standards in light of achieving longer lifespans and withstand harsh environments. In contrast, monitoring the condition of water distribution systems is considerably more challenging due to their vast, dispersed, and intricate design. Although failures in distribution systems are usually less catastrophic, their sheer volume often results in daily losses in the millions of dollars, just due to drinking water losses. Since monitoring distribution systems require a holistic approach to oversee a large number of assets, conventional approaches for selecting and placing CM technology, such as the ones used in Structural Health Monitoring (SMH), are generally not directly applicable. To address this challenge, utilities have developed various strategies, including digital twins favored by thousands of sensors. However, aggregated big-data approaches for analyzing the multitude of factors and subtle system changes affecting deterioration, is a constantly investigated field, since it holds promise at optimizing resources.

In an attempt to leverage water system data and the value of mining historical maintenance records, this study introduces a novel methodology for the optimal selection and deployment of CM technology, particularly for distribution systems. This paper-based dissertation comprehensively addresses the research problem by presenting a state-of-the-art review on CA and CM, examining current techniques for strategic sensor placement in water networks, and developing a framework that integrates technology selection with placement based on the R-E-R-A-V approach (Redundant, Established, Reliable, Accurate, and Viable). The approach is demonstrated using both a synthetic and a real-world water utility system. Furthermore, this dissertation explores more than 50 water pipeline CA and CM technologies and proposes a set of methods to assess their level of technical maturity while aligning with utility priorities with respect to overall functionality, cost efficiency, and operating conditions.

To achieve these objectives, the study introduces a Technology Readiness Level assessment tailored for CM technology in water systems and employs a Spherical Fuzzy Analytical Hierarchy Process (SFAHP) based on six criteria: installation complexity, power requirements, reliability and responsiveness, system and data integration, capital expenses (CAPEX), and operational expenses (OPEX). The optimal placement of CM technology is further supported by a predictive Machine Learning (ML) model that integrates an k-Nearest Neighbors (kNN) algorithm with the OPTICS (Ordering Points To Identify the Clustering Structure) method. The unsupervised OPTICS algorithm is employed as a feature engineering tool that, while evaluating the vulnerability of the pipe segment based on proximity to failure-prone areas, generates key independent variables to the kNN model, such as a categorical failure cluster membership and a numerical Density-Weighted Cluster Proximity Index (FDW-CIP). To effectively combine data from CM technology selection and placement, the dissertation introduces an integrative algorithm based on asset management principles that assigns technologies to the most vulnerable pipe segments with the aim of proactively predicting failures. When tested in 7,558 pipe segments of a real water system, the methodology demonstrated robust performance metrics and achieved statistical significance, indicating that its predictive power was not attributable to chance.

Keywords

Water, Pipelines, Condition Assessment, Condition Monitoring, Sensors, Utility, TRL, SFAHP, OPTICS, Machine Learning

Disciplines

Artificial Intelligence and Robotics | Civil Engineering | Data Science | Technology and Innovation

Download

Available for download on Saturday, May 29, 2027

Share

COinS