Application and Comparison of Regression and Markov Chain Methods in Bridge Condition Prediction and System Benefit Optimization

Abstract

This study is designed to assist highway planners and engineers in identifying and choosing an appropriate method for bridge condition predictions. Mathematical optimization techniques, such as linear programming and dynamic programming, can be used for the evaluation of a bridge system. For each bridge, the input data of the bridge project selection model includes the predicted bridge condition in future years, the recommended bridge repair action, the estimated cost of recommended bridge repair action, and the expected improvement or benefit from the repair action. Through mathematical manipulation, bridge projects are selected to maximize the total expected benefit of the bridge system while a number of constraints are simultaneously satisfied. This optimization process is based on the predicted bridge conditions. Bridge condition predictions will affect bridge project selections and the corresponding system benefits. Using the analysis method presented in this paper and bridge condition data from Indiana, Markov chain methods were shown to yield more accurate condition predictions than the polynomial regression method.