Prediction models for serviceability deterioration of stormwater pipes

Abstract

Reducing the number of annual blockages and the consequential flooding events is one of the most important tasks for stormwater pipe infrastructure managers in Australia. Blockages are more likely to occur with pipes experiencing serviceability deterioration, resulting in a reduction of hydraulic capacity. When changing from a problem-based approach to a proactive maintenance and rehabilitation (M&R) approach, the asset managers need predictive information on the serviceability condition of pipes in order to firstly prepare the necessary resources from limited annual budgets and, secondly, to allocate these resources for the maintenance of the deteriorated pipes as precisely as possible. This paper investigates the application of a Markov model and an ordinal regression model for predictions of serviceability deterioration of stormwater pipes. The first model provides the prediction at a network level, which satisfies the first requirement, and the second model predicts serviceability condition for individual pipes, given the attributes of the pipes, in order to satisfy the second requirement. Both models are calibrated using Bayesian inference and Markov Chain Monte Carlo (MCMC) simulation techniques on a dataset supplied from the City of Greater Dandenong, Australia.