Efficient Optimization for Multi-Objective Decision-Making on Civil Systems Using Discrete Influence Diagram

Abstract

The breakdown of civil systems, e.g. bridge networks and water distribution networks, has a significant social and economic impact, highlighting the importance of optimal decision-making on such systems. Modeling and optimization of probabilistic decision-making problems for civil systems, can be facilitated by graphical methodologies such as influence diagram (ID). However, the converging structure in IDs representing civil systems, which relates the random variables standing for component events and that for system event, results in the exponential increase in the number of modeling parameters and variables to be optimized as that of component events increases. In order to address these challenges, in this paper, the recently proposed matrix-based Bayesian network (MBN) is employed to quantify the IDs. To facilitate the optimization process, a proxy objective function is also proposed. The proxy func-tion not only significantly reduces the number of variables to be optimized, but also allows an efficient framework for multi-objective optimization in which the weighted sum of the objectives is optimized to obtain a set of non-dominated solutions. Three numerical examples demonstrate the performance of the proposed methodology.