Risk-Averse Proactive Seismic Rehabilitation Decision-Making for Water Distribution Systems

Abstract

Earthquakes could have enormous destructive impacts on water distribution networks. Utility managers are challenged to make proactive rehabilitation decisions under seismic and network uncertainties. These utility managers have different risk appetites. However, existing seismic rehabilitation decision-making models of water distribution networks do not consider decision-makers’ attitudes toward risk making existing models practically limited. The objective of this research is to formulate a risk-averse stochastic combinatorial optimization model to identify the critical pipes of a water distribution network for proactive seismic rehabilitation with controllable risk aversion levels. The functionality of the water distribution system is quantified by the post-earthquake serviceability index, the expected value of which is maximized by the objective function. A value-at-risk (VaR) constraint is used to control risk levels. This methodology includes four steps: seismic repair rate calculations, integrated multi-physics modeling, Monte Carlo simulation, and risk-averse stochastic combinatorial optimization. The repair rate of each pipe subjected to seismic loads was calculated using empirical fragility curves. These curves were generated based on the locations of the pipes, soil corrosivity in different locations, pipe diameters, pipe materials, and pipe joint properties. Network’s hydraulic behavior and seismic vulnerability assessment were simulated using an integrated multi-physics model. Monte Carlo simulations were performed to consider the probabilistic nature of damages to the water distribution systems. These damages were represented by leaks as well as breaks in the individual pipes. The model used to ascertain the susceptibility of the water distribution system to earthquake hazard was fused with a stochastic formulation of combinatorial optimization to maximize the serviceability index of the distribution system while minimizing risk. The solution to the optimization problem of detecting the critical pipes for a given resource constraint was obtained through a risk-averse simulated annealing approach. The approach was implemented on a widely used benchmark network to detect the critical pipelines of that water network. The introduction of risk-averse stochastic combinatorial optimization models equips decision makers with a proper model to make rehabilitation decisions at a controllable risk aversion level.

Topics & Concepts