ABSTRACT
Highway agencies use deterioration models to monitor the performance of bridge components (deck, super-structure and substructure) and to predict their remaining service lives on the basis of attributes of the bridges and their operating environment. Most bridge deterioration models are deterministic in nature. However, due to recent federal legislation, agencies are now paying increasing attention to risk-based performance evaluation and decision-making, and seek to use models that incorporate stochastic elements. Such probabilistic models provide more robust predictions of future condition. This paper describes the development of ordered binary probit (BP) models that duly account observation-specific correlation effects. The models describe the bridge component deterioration trends, specifically, the probability that the component condition will drop from one state to another. The paper acknowledges past similar or related efforts in this direction but presents new insights and also simplifies the complexity that is associated with the BP model. In demonstrating the model application, the paper uses data from over 5,000 in-service bridges, including component age, superstructure material type, type of service under bridge, highway functional class, truck traffic, climate severity, rehabilitation history, condition switching state in last inspection period, and current condition rating. Using the developed BP models, the study conducted a simulation to predict the probability of the component condition dropping from one state to another; therefore, the predicted the future condition is based on simulation involving the predicted probability and the current condition. The paper also presents visualizations of the deterioration trend simulation for each component.
All recent federal legislation has encouraged highway agencies to pay more attention to risk-based performance evaluation and decision-making. As such, highway agencies now seek to use models that incorporate stochastic elements. Such probabilistic models provide more robust predictions of future condition.
In the present paper, we described the development of binary probit (BP) models that incorporate random effects, to capture the deterioration process of bridge components, specifically, to describe the probability that the component condition will drop from one state to another.
In the paper, we acknowledged past similar or related efforts in this direction but simplified the complexity that is associated with the BP model. In demonstrating the model application, the paper used data from over 5,000 in-service bridges. The data included component age, superstructure material type, type of service under bridge, functional class, truck traffic, climate severity, rehabilitation history, condition switching state in last inspection period, and current condition rating.
From the model results, the most significant factors that influence the probability of a bridge component switching to a lower condition state, are: age, current condition, whether a transition occurred in last inspection period, and the number of years to last transition. Also, variables that were found to be generally influential for all the three components are: road functional class, region variable (a proxy for climate), number of freeze-thaw cycles, and whether the component received a recent rehabilitation. For the deck deterioration model, the additional significant factors are: level of truck traffic, type of wearing surface, and the number of cold days (below 32°F); for the superstructure deterioration model, the additional significant factor is the superstructure material type; for the substructure deterioration models, type of service under bridge (waterway or otherwise) was found to be influential, much more so compared to its effect on deck and superstructure deterioration. The developed probabilistic modeling methodology can be applied by highway agencies to develop more robust deterioration models to monitor the performance of their bridge components (deck, superstructure and substructure) and to predict their remaining lives on the basis of attributes of the bridges and their operating environment.
