Evaluation of Decision-Making Prediction Models for Sewer Pipes Asset Management

چکیده مقاله

Wastewater collection systems deteriorate over time, requiring continuous adjustments and the development of asset management frameworks by utility owners to maintain the performance of their assets. While closed-circuit television (CCTV) is the primary method for inspecting sewer pipes in the U.S., it is both costly and time-consuming. Therefore, the primary objective of this research is to develop predictive models based on various machine learning algorithms that can forecast the future conditions of sewer pipes. The results of the models can be used to prioritize the need for sanitary sewer pipe inspections, rehabilitation, and replacements. Data collected from the cities of Dallas and Tampa were combined in this research. This dataset included nine independent variables: pipe age, size, length, material, surrounding soil type, soil pH, depth, slope, and surface conditions. The dependent variable was the condition rating of the sewer pipe based on PACP scores, which ranged from 1 to 5. Among models, tree-based models performed better than other models, and the bagging approach was more efficient than boosting techniques. Additionally, it was found that the age and length of the pipes had the greatest effect on the condition rating of sewer pipes, while the pipe location had the least impact.