ABSTRACT
Highway agencies use deterioration models to monitor the performance of bridge components (deck, superstructure and substructure) and to predict their remaining service lives on the basis of attributes of the bridge and its operating environment. In this respect, there are three issues that need to be addressed. The first is that probabilistic models provide more robust predictions of future condition. Secondly, in a departure from transition models (that predict the probability that the bridge condition will transition from one state to another), there is a need to examine bridge deterioration from another perspective that considers how long bridges stay (or transit) in a certain condition state or the probability that a bridge will stay in that condition state given its age and other attributes. The condition state is a discrete variable ranging from 1 to 9: defined as 8–9 (Excellent), 7 (Good), 6 (Fair), 1–5 (Poor). Thirdly, there is a need to account for observation-specific attributes that influence bridge deterioration. In addressing these issues, this paper develops ordered probit (OP) models with random effects that predict the probability of bridge component staying in a certain condition state. The paper considered using random effects because not only can they account for the wide variability in the NBI data but also can address the unobserved variance among the individual observations.
An intended major contribution of this paper is the “Bridge Maintenance Matrix (BMM)” in which three maintenance interventions, i.e., repair, rehabilitation and replacement/reconstruction are defined in terms of condition jumps of bridge components (from a lower to a higher condition state), as shown in Figure 1. Using NBI database, through tracking the condition panel data spanning 23 years (1992 to 2014), from over 5,000 in-service bridges; the average length of time during which each intervention was effective, in other words, the number of years its takes to completely dissipate the maintenance effectiveness (the maintenance effectiveness is defined in terms of its ability to reduce the probability that the bridge component will stay in a given condition state) was determined. Such probability reduction was termed the “decreasing factor”. Bridge Maintenance Matrix. https://s3-euw1-ap-pe-df-pch-content-public-p.s3.eu-west-1.amazonaws.com/9781315207681/cd556cd4-4dcf-4efe-8e29-56fc67b8bfbd/content/fig182_1.tif"/>
In building upon past work in this area, this paper throws more light on the impact of newly-considered explanatory variables such as the maintenance history. The paper analyzed the post-application effectiveness of different types of maintenance work in terms of the probability that the recipient bridge component will stay in a given condition state. Using the developed OP model, any of several possible scenarios (combinations of the input variables) can be analyzed via simulation to visualize the predicted probability trend of bridge components being in different conditions over time (shown in Figure 2 for deck only). The input variables include internal factors such as age, design variables, maintenance history, and external factors such as administrative region, climate, and traffic loads. This knowledge is critical for bridge policy making, cost allocation, and other management and administration issues associated with bridges. Probability Prediction curve for bridge deck. https://s3-euw1-ap-pe-df-pch-content-public-p.s3.eu-west-1.amazonaws.com/9781315207681/cd556cd4-4dcf-4efe-8e29-56fc67b8bfbd/content/fig182_2.tif"/>
