Life-Cycle Management For Sustainable Infrastructure Systems Using Stochastic Dynamic Programming And Evolutionary Algorithms

Abstract

The objective of this paper is to develop a comprehensive framework for Life-Cycle optimization of large infrastructure systems, and identify the procedure of searching optimal multistage maintenance actions and schedules with multi-objective goals. The complex optimization problem is solved by a decomposed two-phase approach. In facility level, an integrated method that combines stochastic duration model with dynamic programming is developed to identify the optimal maintenance actions and schedules for individual infrastructures under various states. In system-level, the problem is formulated by combinatorial optimization with multiple objectives and multiple budget constraints. A heuristic algorithm based on NSGA II and simulation is proposed to identify the systems optimal maintenance decisions and estimate the future budget requirements. To demonstrate the advantage of the model, we present a numerical example of pavement preservation management system. The studies captured the stochastic property of infrastructure deterioration, provided the optimal strategies to allocate the limit budget, analyzed the trade-off between economic costs and emissions, and demonstrated the importance of embracing a life-cycle perspective.