A neural network-based approach to predict the condition for sewer pipes

Sewer pipe systems are of vital importance in the development of cities, especially in the context of highly populated areas where hygiene is a key factor in avoiding large-scale epidemics. However, the deterioration problem led by age is one of the major issues that need regular monitoring and maintenance to keep the drainage system operational at all time. In order to schedule maintenance effectively, an accurate prediction model of the pipes' condition is required as a reference for prioritization. Given the importance of this assessment, this research proposed a neural network-based approach to automatically predict the condition of sewer pipes. Taking advantage of the availability of both the large volume of data and increasing computational power, a data-driven method is employed in this research. Historical data containing the recorded information of sewer pipes serves as the input of the model, and the output is the condition of the sewer pipes. The final neural network model could serve as a reference for the purpose of efficient scheduling of sewer pipe maintenance work by assigning different priorities to sewer pipes.