Integrated machine learning-based life-cycle cost analysis models for comparative prediction of sewer rehabilitation methods for resilient underground infrastructure planning

Abstract Aging wastewater networks require sustainable and low-impact rehabilitation strategies. Open-cut pipeline installation (OCPI) is a conventional rehabilitation method associated with significant construction-related disruptions, whereas Cured-in-Place Pipe (CIPP) is a trenchless alternative that minimizes surface disturbance. However, existing studies rarely integrate construction costs (CC), environmental costs (EC), and social costs (SC) within a unified, sustainability-oriented life-cycle cost analysis (LCCA) framework. This study develops supervised machine-learning (ML) models—k-Nearest Neighbors (KNN), Decision Tree Regression (DTR), Gradient Boosting Regression (GBR), and Multiple Linear Regression (MLR)—to predict construction costs for OCPI and CIPP across pipe diameters ranging from 150 to 2,130 mm (6–84 in.). These predictions are integrated into a modular LCCA framework incorporating EC quantified using SimaPro 2024 with the ReCiPe 2016 Midpoint (H) method and SC estimated using established formulations. Under the stated planning-level assumptions and included cost components, integrating CC, EC, and SC demonstrated. A national dataset comprising more than 750 installations was preprocessed using Multiple Imputation by Chained Equations (MICE), five-fold cross-validation, and hyperparameter optimization. Model performance was evaluated using the coefficient of determination (R²), root mean square error (RMSE), mean absolute percentage error (MAPE), and prediction accuracy based on a ± 20% relative error threshold. KNN achieved the highest predictive performance for OCPI, while GBR performed best for CIPP. OCPI generated higher environmental impacts in 16 of the 18 ReCiPe midpoint categories for pipe diameters of 150–525 mm (6–21 in.). Under the assumptions and cost components considered in this planning-level assessment, integrating CC, EC, and SC resulted in CIPP having an average LCC approximately 56% lower than OCPI, with peak savings of 69% for pipe diameters of 150–1,500 mm (6–60 in.). Under these same assumptions, CIPP rehabilitated approximately 2.3 times more pipeline length per dollar than OCPI. The proposed integrated ML–LCCA framework provides a practical decision-support tool for sustainable and resource-efficient sewer infrastructure planning.