Risk-based bridge project selection using genetic algorithm optimization

In the past few years, state and federal government agencies began to develop and implement comprehensive systems to monitor and control the deteriorated state of the nation's infrastructure. First generation Bridge Management Systems and Pavement Management Systems are now present in many states. Most existing systems rely on the concepts of cost-benefit ratios, life-cycle costs and average condition ratings as the basis of comprehensive management programs.;This study proposes risk (reliability) based bridge or pavement management. The reliability approach allows the incorporation of the many sources of uncertainty involved in inspection, performance assessment, and deterioration modeling of bridges.;The proposed methodology consists of two phases. First, the risk to the public is evaluated for each bridge using reliability methods. Second, repair projects are selected optimally over a specific number of years so that risk is minimized.;The risk from a single bridge is based on current reliability methods and combines the elemental reliabilities found for the separate parts of a bridge. These elemental reliabilities are found by either first order methods or by subjective assignment. In either case, the method for "combining" element conditions takes into consideration the elements participation, thus it is more realistic than simple averages.;The deterioration of the elements is modeled with Markov Chains. This stochastic process blends naturally with the existing condition rating scales. The two levels of uncertainty allow the risk from a bridge with any combination of element repairs to be found.;In any bridge management system, the optimization task is very large. Traditional systems resort to simplifying heuristic criteria to obtain the solutions. The optimization approach proposed in this research makes no shortcuts of this nature; instead combinatorial explosion is overcome by introducing Genetic Algorithm as a means of optimization.;The most important findings from this study are: the lack of correlation of existing averaging formulas with system reliability methods, the practicality of condition based inspection intervals, the benefits of extending the optimization to include many years of repairs, and the usefulness of Genetic Algorithms for bridge project selection.