Sewer pipeline condition prediction using neural network models

With an aging underground infrastructure, ever-encroaching population areas and increasing economic pressures, the burden on the municipal agencies to efficiently prioritize and maintain the rapidly deteriorating underground utilities is increasing. Accurate forecasting of pipeline performance is essential for prioritizing and risk management of the underground infrastructure. The essential function of a pipeline asset management system is to consider the pipeline maintenance and improvement needs and to arrive at the program of optimal rehabilitation, replacement, and maintenance. Hence, the development of a pipeline condition prediction model will be indispensable to the concerned authorities in prioritizing the care and rehabilitation of pipelines, and in pipeline asset planning and management. This research developed an Artificial Neural Network (ANN) model for predicting the condition of sewer pipes based on the historic condition assessment data. The neural network model was trained and tested with acquired field data. The developed model is intended to aid in identifying the distressed segments of the overall sewer pipeline network using a set of known input values. These can then be directed toward assessing and prioritizing the maintenance measures needed to prevent accelerated future distress and eventual failure of sewer pipes.