TY - JOUR
T1 - Prediction of Breaks in Municipal Drinking Water Linear Assets
AU - Karimian, Farzad
AU - Kaddoura, Khalid
AU - Zayed, Tarek
AU - Hawari, Alaa
AU - Moselhi, Osama
N1 - Publisher Copyright:
© 2020 American Society of Civil Engineers.
Copyright:
Copyright 2020 Elsevier B.V., All rights reserved.
PY - 2021/2/1
Y1 - 2021/2/1
N2 - Improper asset management practices increase the probability of water main failures due to inactive intervention actions. The annual number of breaks of each pipe segment is known as one of the most important criteria for the condition assessment of water pipelines. This metric is also considered one of the major performance measures in levels of service (LoS) studies. In an effort to maximize the benefits of historical data, this research utilized the evolutionary polynomial regression (EPR) method in determining the best mathematical expression for predicting water pipeline failures. The prediction model was trained and tested on the city of Montreal water network. After determining the best independent variables through the best subset regression, pipelines were clustered based on their attributes (length, diameter, age, and material). The majority of the models provided high R2 values, but the highest performing model's R2 was 89.35%. Further, a sensitivity analysis was also performed and showed that the most sensitive parameter was the diameter, and the most sensitive material type to age was ferrous material. The tools and stages performed in this research showed promising results in predicting the expected water main failures using four different asset attributes. Therefore, this research can be implemented in asset management best practices and in LoS performance measures to predict the number of water pipeline failures. To further improve the prediction model, additional explanatory variables could be considered along with leveraging multiple artificial intelligence tools.
AB - Improper asset management practices increase the probability of water main failures due to inactive intervention actions. The annual number of breaks of each pipe segment is known as one of the most important criteria for the condition assessment of water pipelines. This metric is also considered one of the major performance measures in levels of service (LoS) studies. In an effort to maximize the benefits of historical data, this research utilized the evolutionary polynomial regression (EPR) method in determining the best mathematical expression for predicting water pipeline failures. The prediction model was trained and tested on the city of Montreal water network. After determining the best independent variables through the best subset regression, pipelines were clustered based on their attributes (length, diameter, age, and material). The majority of the models provided high R2 values, but the highest performing model's R2 was 89.35%. Further, a sensitivity analysis was also performed and showed that the most sensitive parameter was the diameter, and the most sensitive material type to age was ferrous material. The tools and stages performed in this research showed promising results in predicting the expected water main failures using four different asset attributes. Therefore, this research can be implemented in asset management best practices and in LoS performance measures to predict the number of water pipeline failures. To further improve the prediction model, additional explanatory variables could be considered along with leveraging multiple artificial intelligence tools.
KW - Asset management
KW - Evolutionary polynomial regression
KW - Levels of service
KW - Prediction
KW - Water pipelines
UR - http://www.scopus.com/inward/record.url?scp=85092314201&partnerID=8YFLogxK
U2 - 10.1061/(ASCE)PS.1949-1204.0000511
DO - 10.1061/(ASCE)PS.1949-1204.0000511
M3 - Journal article
AN - SCOPUS:85092314201
SN - 1949-1190
VL - 12
JO - Journal of Pipeline Systems Engineering and Practice
JF - Journal of Pipeline Systems Engineering and Practice
IS - 1
M1 - 04020060
ER -