Performance Evaluation of Pipe Break Machine Learning Models Using Datasets from Multiple Utilities

Chen, Thomas Ying‐Jeh; Vladeanu, Greta; Yazdekhasti, Sepideh; Daly, Craig

doi:10.1061/(asce)is.1943-555x.0000683

Cited by 8 publications

(6 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Previous research by T. Y. J. Chen and Guikema (2020); T. Y.‐J. Chen et al (2022) suggests that older break records are not as useful as recent ones when predicting where future breaks will occur. By applying a greater weight to recent breaks, it will cause the heatmap to skew toward these locations, and by extension impact the allocation of sensors to better cover these areas.…”

Section: Case Studymentioning

confidence: 99%

Data‐efficient and risk‐based pressure sensor placement optimization for water systems

Chen,

Dryer,

Rathor

et al. 2024

AWWA Water Science

View full text Add to dashboard Cite

Deployment of remote metering infrastructure can help utilities improve pressure management and limit pipe break impacts. Since it is cost‐infeasible to install sensors to fully cover an entire system, a risk‐based approach where devices are deployed in areas of systemic vulnerabilities will maximize the benefit of these limited resources. Previous work on sensor allocations requires well‐calibrated hydraulic models that contain a full asset inventory. This high barrier of data requirement makes it difficult to apply existing methods. In this research, we present a more efficient methodology with lower data requirements. Proxies for system vulnerabilities are generated using historic break records, and GIS inventories are used to derive candidate installation points. The problem is presented as a generalized version of the well‐studied maximal coverage location problem, where locations are selected for optimal coverage of risky areas. The workflow is demonstrated on a small suburban utility in Pennsylvania, USA.

show abstract

Section: Case Studymentioning

confidence: 99%

Data‐efficient and risk‐based pressure sensor placement optimization for water systems

Chen,

Dryer,

Rathor

et al. 2024

AWWA Water Science

View full text Add to dashboard Cite

show abstract

“…In order to solve the problem of limited historical failure data experienced in some water utilities, Chen et al. (2022) combined historical data of six utilities to make failure probability predictions. Three algorithms were used: RF, GBT, and XGBoost.…”

Section: Machine Learning‐based Modelsmentioning

confidence: 99%

“…Hence, low quality and limited data will result in inaccurate predictions. Besides, ML models developed with limited data and an algorithm that optimizes its parameters may lead to overfitting (Chen et al., 2022). Hence, such a model cannot be generally applied to other historical failure data.…”

Section: Machine Learning‐based Modelsmentioning

confidence: 99%

Toward Sustainable Water Infrastructure: The State‐Of‐The‐Art for Modeling the Failure Probability of Water Pipes

Taiwo

Seghier

Zayed

2023

Water Resources Research

View full text Add to dashboard Cite

Water distribution networks (WDNs) are undoubtedly a critical part of water utility assets, as they are responsible for water transmission (Atef et al., 2016;Berardi et al., 2014). It has been reported that, on average, 80% of water utility expenditures are spent on the management of WDNs (Poulakis et al., 2003). However, available evidence shows that the failure of WDNs is increasing at a fast rate, which negatively impacts individuals' social, economic, and health status (Steffelbauer et al., 2022;Yazdani & Jeffrey, 2012).In the USA and Canada, around 700 water pipes fail daily, thereby contributing to the loss of over 2 trillion gallons of clean water annually (Fan et al., 2022). In 2017, more than 2.2 billion m 3 of water was loss in China (Liao et al., 2021). The enormous financial commitment associated with the failure of water distribution networks cannot be overemphasized. According to the American Water Work Association (AWWA), the USA needs to invest around $1 trillion in replacing and repairing the deteriorating components of their WDNs (Fan et al., 2022). In Australia, the estimated cost of repairing and maintaining WDNs is about AUD 1.4 billion (Weeraddana et al., 2020). In South Korea, 52.5% of the water pipes will require rehabilitation by 2024, as indicated in a study by Seo et al. (2015). In the case of Colombia, a developing country, approximately 50% of water is lost due to water pipe failures (Giraldo-González & Rodríguez, 2020). From the aforementioned, the failure of WDNs, which mostly consist of water pipes, is a global issue that requires utmost attention. WDNs are founded underground; therefore, these infrastructures need to be designed to resist traffic, soil, internal, and overburden pressures, and other environment and operation related loads (Berardi et al., 2008). In essence, the reliability of WDNs must not be compromised during their service life; hence, catastrophic failure

show abstract

“…The underlying models rely on historic records of asset failure to train and validate the predictions for which assets will fail in the future. A case study by Chen et al (2022) examines the relationship between data quantity and model performance. On the other hand, see literature reviews by Tscheikner-Gratl et al (2019) and Ana and Bauwens (2007) that summarize the state of the art in sewer main risk modeling.…”

Section: Article Impact Statementmentioning

confidence: 99%

Multi‐objective optimization models for the renewal planning of multiple asset classes

et al. 2023

View full text Add to dashboard Cite

show abstract

Performance Evaluation of Pipe Break Machine Learning Models Using Datasets from Multiple Utilities

Cited by 8 publications

References 38 publications

Data‐efficient and risk‐based pressure sensor placement optimization for water systems

Data‐efficient and risk‐based pressure sensor placement optimization for water systems

Toward Sustainable Water Infrastructure: The State‐Of‐The‐Art for Modeling the Failure Probability of Water Pipes

Multi‐objective optimization models for the renewal planning of multiple asset classes

Contact Info

Product

Resources

About