Multiobjective Approach for Pipe Replacement Based on Bayesian Inference of Break Model Parameters

Abstract

A planning strategy is presented that aims at establishing the optimal replacement schedule for a water distribution network. Two performance indicators are defined. The first accounts for structural state (total cost defined as the sum of pipe replacement cost and the expected cost of pipe break repairs) and the second for hydraulic performance (minimization of the pressure deficit). A multiobjective objective function is defined based on these two indicators and a genetic algorithm optimization technique is used to identify optimal solutions (Pareto front). Three management strategies are considered to choose a replacement schedule among those making up the Pareto front: (1) a prostructural strategy that only considers the structural indicator; (2) a prohydraulic strategy that integrates both structural and hydraulic indicators; and (3) a budgetary constraint strategy which assumes a predefined budget for replacement expenditures. The proposed planning strategy was tested on two hypothetical networks. Synthetic pipe break records were generated using a statistical pipe break model. A Bayesian inference approach was then used to estimate parameters values from these pipe break series. A comparison of the different management strategies is provided as advantages of using Bayesian inference are discussed.