Prioritizing inspection of sewer pipes based on self-cleansing criteria.

This paper aims to develop a sediment deposits hydraulic deterioration model based on self-cleansing criteria to prioritize the inspection of sewer systems. The model was trained with benchmarking literature values from earlier experiments and validated with household connections complaints data from Bogotá, Colombia. Recursive Feature Elimination with Cross-Validation (RFECV) and Bayesian Optimization (BO) were used to construct a Random Forest (RF) model to predict, at pipe level, the likelihood for a pipe to present sediment deposits. To evaluate the model's prediction accuracy, two different performance indicators were used: (i) the Percentage of Effective Inspections, and (ii) Pipes per Inspection with sediments. The sediment deposits hydraulic deterioration model shows good overall performance with buffer zones radiuses of 250 m predicting which pipes tend to present sediment deposits over time. This model improves the understanding of sediment deposits in hydraulic deterioration models and can be used to prioritize inspection of sewer systems. [ABSTRACT FROM AUTHOR]