Methodology for classifying the structural state of unin-spected pipes in sewer networks based on support vector machines

Abstract

The nearly unmitigated growth of cities has placed ever-greater pressure on urban water systems regarding climate change, environmental pollution, resource limitations, and infrastructure aging. Therefore, the development of methods to classify and assess the structural state of urban drainage infrastructure becomes very important, given that they can be used as support tools for proactive management plans. This paper presents a method for predicting and classifying the structural state of uninspected sewer pipes using Support Vector Machines, based on the physical characteristics, age, and geographical location of the pipes. According to the results, the methodology: i) correctly classified more than 75% of uninspected pipes; (ii) identified pipes in critical structural states, with low importance prediction error for 69% of pipes; and (iii) provided a guide for establishing the number or percentage of pipes that require inspection or intervention.