The North Carolina Department of Transportation (NCDOT) currently oversees the design, construction, operation, maintenance, repair, rehabilitation, and replacement of more than 17,000 bridges. As funding to match the growing need for new infrastructure and for maintenance, repair, and rehabilitation (MR&R) of existing infrastructure becomes more difficult to obtain, maximizing the service life of existing bridges becomes increasingly critical. In support of data-driven planning, NCDOT's bridge management system (BMS) stores inventory data, including bridge characteristics, inspection data, and rating information, and uses deterioration models and economic models to predict outcomes and to provide network-level and project-level decisions.
The objectives of this project were to provide NCDOT with revised, updated deterioration models and user cost tables for use in the BMS software. Existing data in NCDOT's BMS was reviewed and steps to address data anomalies were identified and implemented. Updated deterministic deterioration models were developed for the existing data on the family level, with components grouped into families using established a priori classifications.
Additionally, a unique statistical regression methodology applying survival analysis techniques to better address characteristics of the historical condition rating data was developed and resulted in probabilistic deterioration models for bridge components and culverts that provide significantly improved predictive accuracy and precision over prior deterministic models. These models include transition probability matrices that account for the effects of design, geographic, and functional characteristics on deterioration rates over different condition ratings. These models were found to provide significantly improved prediction accuracy and precision over typical planning horizons used in network analysis. However, while this advanced model was found to best fit the historical condition rating data and provide unique insight on factors influencing deterioration over the life-cycle of each bridge component, it was also discovered that a simplified implementation of the probabilistic deterioration model was able to achieve similar performance without rigorously incorporating the effects of external factors on deterioration rates.
To aid in implementation and technology transfer, a software application was developed to facilitate routine updating of both the deterministic and probabilistic deterioration models. Preliminary work to evaluate the relative impact of individual maintenance activities on element condition ratings was performed, including the development of histograms of condition rating changes from prior actions to aid in development of action effectiveness models. Inputs and methodologies utilized to compute user costs in NCDOT's BMS were updated and enhanced using relevant, current resources that were locally or regionally sourced when possible. Specifically, the updates and enhancements to the user cost models address ADT growth rates, vehicle operating cost, vehicle distribution, vehicle weight distribution, vehicle height distribution, accident injury severity, accident cost, and an equation useful in forecasting the number of annual bridge-related crashes.
Analysis performed to generate the bridge-related crash prediction equation resulted in the identification of seven bridge characteristics that are most associated with bridge-related crashes. A sensitivity analysis on user costs indicated that, in NCDOT's BMS, user costs are most sensitive to accident costs.
