Optimal bridge maintenance planning based on probabilistic performance prediction

Current automated maintenance planning procedures for deteriorating bridges are usually based on deterministic prediction of bridge performance and whole-life maintenance costing. In these procedures, uncertainties associated with the deterioration process under no maintenance and under maintenance are not taken into consideration. In this paper, such uncertainties are confined to the parameters that define the selected computational models and their effects are evaluated by means of Monte Carlo simulations. A multiobjective genetic algorithm based numerical procedure is used to locate, in the Pareto optimal sense, the best possible tradeoff maintenance planning solutions with respect to three objective functions, namely, condition index, safety index, and cumulative life-cycle maintenance cost. By computing these objectives in terms of either sample mean or sample percentile values, bridge managers’ specific confidence levels on the performance of maintenance solutions can therefore be conveniently incorporated into the optimization process.