STOCHASTIC OPTIMIZATION SUBSYSTEM OF A NETWORK-LEVEL BRIDGE MANAGEMENT SYSTEM

The prediction and stochastic optimization modules are two of seven modules that make up a stochastic network-level bridge management system which is under development. An overview of major portions of the bridge management system is provided. The prediction model of structural degradation generates initial estimates of transition probabilities (tp values). A tp value is defined as the probability that a bridge segment will move from one condition state to another within 1 year given the maintenance scope assigned to it. The tp values are updated with new survey data using a Bayesian updating procedure. Methods are developed to account for the fact that structural surveys may be performed on a multiyear basis, while yearly tp values are needed for the optimization models. The optimization module, which minimizes cost subject to top management's performance objectives, is a Markovian-based linear program that stratifies the bridge network to improve degradation predictions. Rather than using single ratings for a major bridge element (e.g., bridge deck), the program optimizes on a bridge segment level to maximize the use of structural condition information. The condition state of a segment can include selected functional deficiencies as well as structural condition ratings.