Seismic Maintenance of Water Pipe Networks Using Stochastic Combinatorial Optimization

Water pipe networks are one of the lifeline networks of modern cities. Despite the importance of these networks, empirical evidence shows a substantial chance of these networks being damaged by seismic events causing significant disruption to their serviceability. To avoid such disruptions, utility managers often need to identify effective maintenance schemes while minimizing use of a limited rehabilitation budget. This identification is often made challenging by the inherent uncertainties associated with seismic events, uncertainty associated with seismic fragility estimation of water pipes, exponentially increasing decision space, and limited maintenance budget available to the utilities. Hence, there is a critical need for an effective approach to identify optimal seismic maintenance plans for minimizing post-earthquake serviceability loss. In this study, a stochastic combinatorial optimization formulation was used to model the problem. Monte Carlo simulation integrated with quasi-pressure-driven hydraulic analysis was used to evaluate the serviceability of the network for each maintenance scheme. Two commonly used metaheuristics, namely genetic algorithm and simulated annealing, were used to identify the optimal maintenance scheme that minimizes the post-earthquake serviceability loss. The effectiveness of the methodology was demonstrated by using two city-scale water pipe networks as the testbeds. The performance of these metaheuristics was then compared in terms of their computational runtime and solution quality. The comparison showed that both genetic algorithm and simulated annealing based optimization can effectively identify economical seismic rehabilitation plans for city-scale water pipe networks. The findings of this study are expected to be highly valuable to utility managers managing water pipe networks exposed to significant seismic risks.