A Two-Stage Data-Driven Spatiotemporal Analysis to Predict Failure Risk of Urban Sewer Systems Leveraging Machine...

Area of Science:

• Environmental Engineering
• Urban Infrastructure Management
• Data Science

Background:

• Urban sewer systems require effective asset management for optimal performance.
• Previous failure risk analyses were limited by unidimensional (spatial or temporal) approaches, neglecting complex spatiotemporal interactions.
• Data imperfections like missing values and outliers hinder accurate failure risk assessment.

Purpose of the Study:

• To develop a generalized framework for robust spatiotemporal analysis of urban sewer system failure risk.
• To address challenges posed by imperfect data, including missing data, outliers, and imbalanced information.
• To integrate machine learning and optimization techniques for improved sewer maintenance planning.

Main Methods:

• A two-stage, data-driven modeling technique utilizing a bidimensional space-time approach.
• Implementation and validation of various machine learning algorithms (logistic regression, decision trees, random forests, XGBoost).
• Application of optimization techniques for scheduling sewer system maintenance operations.

Main Results:

• A robust framework capable of handling data imperfections for accurate spatiotemporal failure risk prediction.
• Identification of the best-performing machine learning model based on goodness-of-fit and predictive accuracy.
• Demonstration of the framework's utility in planning and scheduling maintenance operations.

Conclusions:

• The proposed framework enables comprehensive spatiotemporal analysis of sewer system failure risk.
• Stakeholders can gain managerial insights into model performance, spatial resolution, and decentralized management strategies.
• This approach enhances the efficiency and effectiveness of urban sewer asset management.