Data-driven bi-level sewer pipe deterioration model: Design and analysis

Abstract

To inspect every pipe in a city annually would be a costly and labor-intensive endeavor due to the significant number of sewer pipes underground. In practice, sewer pipes in worse (better) conditions should be assigned a higher (lower) priority for maintenance. A deterioration model is an effective tool for prioritizing sewer pipe maintenance since it is able to predict the current and future conditions of these pipes. As such, this paper proposes a bi-level deterioration model to predict the condition of sewer pipes at a neighborhood level and an individual level. The neighborhood-level prediction is used to facilitate the maintenance schedule at the neighborhood scale, while the individual-level prediction is used to identify the sewer pipes with the highest risk of failure so that maintenance operations can be scheduled to keep them operating at an acceptable level of service. A linear regression model is employed in the development of the neighborhood-level prediction model, and the results are visualized by a geographic information system (GIS). The model for predicting the neighborhood level deterioration is first proposed in this research. A neural network (NN) model with a backward variable elimination process is proposed to predict the individual sewer pipe condition. Deterioration curves are derived from the individual prediction model, which facilitates better decision making with respect to sewer pipe maintenance. The contributions of the research not only lie in developed a bi-level deterioration model for the targeted city but also proposed a framework that can be generalized for municipal departments located in other cities that aim to develop their own deterioration model for sewer pipe systems. Optimized procedures, including results visualization geographically, input variable selection, etc., are adopted in the model development process to get a more accurate and efficient deterioration model for the sewer pipe system.