Classification of Defects in Sewer Pipes Using Neural Networks

Abstract

Deterioration of underground infrastructure facilities such as sewer pipes poses a serious problem to most developed urban centers today. As distribution piping networks age, they deteriorate and may ultimately fail to fulfill their intended functions. To ensure continuity of services and protect the investment made in these networks, municipalities check their conditions regularly. The current practice that is being followed in those checkup programs is usually time consuming, tedious, and expensive. This paper presents an automated system designed for detecting defects in underground sewer pipes and focuses primarily on the application of neural networks in the classification of those defects. A three-layer (i.e., one hidden layer) neural network has been developed and trained using a back-propagation algorithm to classify four categories of defects, namely cracks, joint displacements, reduction of cross-sectional area, and spalling. A total of 1,096 patterns were used in developing the neural network. An example application is described to demonstrate the use and capabilities of the developed system.