Condition Prediction of Deteriorating Concrete Bridges Using Bayesian Updating

It is well known that the U.S. infrastructure is in need of extensive repair. To ensure that the scarce resources available for maintaining the U.S. bridge inventory are spent in an optimal manner, bridge management programs have been mandated by the Federal Highway Administration. However, these programs are mainly based on data from subjective condition assessments and do not use time-variant bridge reliability for decision making. Many nondestructive test methods exist for the detailed inspection of bridges. Predictions based solely on inspection data may be questionable, particularly if limitations and errors in the measurement methods that are used are not considered. Through the application of Bayesian techniques, information from both inspection data and engineering judgment can be combined and used in a rational manner to better predict future bridge conditions. In this study, the influence of inspection updating on time-variant bridge reliability is illustrated for an existing reinforced concrete bridge. Inspection results are combined with prior information in a Bayesian light. The approach is illustrated for a reinforced concrete bridge located near Pueblo, Colo. For this bridge the effects of corrosion initiation time and rate on time-variant strength are illustrated using simulation. Inspection results are combined with prior information using Bayesian updating. Time-variant bridge reliability computations are performed using a combined technique of adaptive importance sampling and numerical integration. The approach presented allows accounting for inspection results in the quantitative assessment of condition of bridges and shows how to incorporate quantitative information into bridge system and component condition prediction.