Comparison Of Statistical Failure Models To Support Sewer System Operation

Achieving appropriate operational performance of urban wastewater infrastructure has become a high priority for water utilities. Recent research has focused on developing models to support proactive maintenance and rehabilitation of sewerage systems. This study evaluates two predictive software tools that use different statistical models: (a) the FAIL software (Martins et al., 2013) and (b) the SIMA software (Rodriguez et al., 2012). Comparisons among a singlevariate Homogeneous Poisson Process (HPP) implemented in the FAIL software and two different Non-Homogeneous Poisson Processes (NHPP) implemented in the SIMA software are conducted in this study. Two contrasting urban wastewater systems are studied: Bogota (Colombia) and SIMAS Oeiras and Amadora (Portugal). Furthermore, three different types of sewer failures named blockage-related failures, sediment-related blockages and structural failures are analysed. In order to evaluate the prediction efficiency of each model, the number of predicted failures obtained using each model were compared with the observed number failures. The obtained results showed that both models were capable to point towards the same number of observed failures. On the other hand, the HPP model range of prediction was wider than the NHPP models, showing that the latter has a higher prediction precision. Three casestudies also evidenced that NHPP models are more accurate when compared with the HPP model: the number of observed failures are within the prediction range in a higher percentage of the fits.