No abstract

No abstract

Once installed, water pipes “condition degrade” until failure. Utility managers anticipate the failure of water mains through a proactive management of urban water systems. In proactive management, deterioration models are used to predict the actual condition of uninspected pipes and forecast the future condition of all the pipes in the network. For an efficient planning, these models' predictions should be as close as the observed conditions of pipes during inspections. High‐degree polynomials can capture the complexity of water pipes deterioration process. However, these polynomials wiggly estimate such relationships and are unsatisfactory in some regions where they fail to fit the observed data. Flexible regression techniques that enable automatic data‐driven estimation of nonlinear relations between covariates and response constitute an alternative approach that can represent the stochastic deterioration process. In this article, a semiparametric deterioration model based on geoadditive quantile regression with smooth nonlinear function estimation of the effects of continuous covariates is proposed. The results confirm the nonlinear assumption of the continuous covariates and highlight factors contributing to the appearance of extreme values in the response variables. Maps representing the effects of the unobserved covariates on the pipes breakage rate are also produced.

No abstract

No abstract