Application of Regression-Based Machine Learning Algorithms in Sewer Condition Assessment for Ålesund City, Norway

Abstract

Predicting the condition of sewer pipes plays a vital role in the formulation of predictive maintenance strategies to ensure the efficient renewal of sewer pipes. This study explores the potential application of ten machine learning (ML) algorithms to predict sewer pipe conditions in Ålesund, Norway. Ten physical factors (age, diameter, depth, slope, length, pipe type, material, network type, pipe form, and connection type) and ten environmental factors (rainfall, geology, landslide area, population, land use, building area, groundwater, traffic volume, distance to road, and soil type) were used to develop the ML models. The filter, wrapper, and embedded methods were used to assess the significance of the input factors. A dataset consisting of 1159 inspected sewer pipes was used to construct the sewer condition models, and 290 remaining inspections were used to verify the models. The results showed that sewer material and age are the most significant factors, otherwise the network type is the least contributor affecting the sewer conditions in the study area. Among the considered ML models, the Extra Trees Regression (R2 = 0.90, MAE = 11.37, and RMSE = 40.75) outperformed the other ML models and it is recommended for predicting sewer conditions for the study area. The results of this study can support utilities and relevant agencies in planning predictive maintenance strategies for their sewer networks.