The development of performance prediction and optimization models for bridge management systems

A major objective of a bridge management system is to assist highway programmers in making consistent and cost-effective decisions related to bridge activities on a systemwide basis. As part of an effort to develop a bridge management system for the Indiana Department of Transportation, a bridge performance prediction model and an optimization model for project selection were developed. The purpose of the performance prediction model was to find the relationship between bridge condition rating and bridge age. The performance curves, obtained by applying polynomial regression techniques, were used to predict the average condition of bridges at a given age. The prediction of future conditions of individual bridges was achieved by utilizing the theory of Markov processes. The optimization model was developed by using dynamic programming in combination with integer linear programming and Markov chains. The function of the model was to select bridge projects so that the effectiveness or benefit of a bridge system is maximized subject to the constraints of available state and federal funds over a given program period. The Markov chain model was incorporated into the optimization model to predict or update bridge conditions at each stage of the dynamic programming. One version of the model maximizes the systemwide effectiveness in terms of bridge structural condition weighted by traffic volume, bridge traffic safety index and community impact index. A second version of the optimization model combines the ranking and optimization techniques so that the results of ranking and optimization models could be compatible. The proposed models are described in this thesis, and the results are presented along with examples and discussions. The performance prediction and optimization models can be used to provide optimal solutions to systems with hundreds of bridges.