Session Details: 2019 CNAM Conference

The City of Calgary has built an extensive database of their sewer system including CCTV data, physical characteristics and maintenance data. This study uses the available data to build advanced deterioration models that mimic the physical process through the inclusion of nonlinear smooth P-splines functions. Missing data is accounted through a surrogate covariate that is the geospatial location of pipes inside the district where the pipes are buried. The geospatial location of pipes contains information that is not systematically collected through inspection campaigns such as soil types, loading conditions, groundwater table fluctuations, and many others. This information is valuable since the pipes installed in the same area are affected by the same factors. Thus, these pipes should have the same deterioration pattern unless some heterogeneities exist. The developed deterioration model accurately predicts the pipes scores that can be translated into internal condition. In addition, a map that represents the impacts of different districts on the deterioration of sewer pipes is produced. This map gives a visual approximation of the impact of the districts on the deterioration of sewer pipes and allows for rapid prioritization of inspections. The extension of the model to the analysis of the response quantiles allows to evaluate the partial effects of covariates on the response distribution and quantify the uncertainty in the prediction output.