Development of Bridge Management Tools for Predicting Bridge Replacement Projects
Abstract
One of the responsibilities of the NCDOT is the prioritization of highway bridge replacement projects throughout the North Carolina transportation system. The Priority Rating Index (PRI) is a multi-criteria formula that is currently used to provide a score based on condition and functional data in the Bridge Management System (BMS) for each of the highway bridges in North Carolina to aid in ranking the priority of potential replacement projects. The PRI comprehensively utilizes many of the performance measures considered to be important by the NCDOT Structures Management Unit, the group responsible for maintaining North Carolina’s highway bridge network. However, anecdotal evidence from Structures Management Unit personnel, supported by an analysis of PRI score distributions among bridges selected and not selected for replacement, suggests that the PRI is a poor indicator of whether a bridge will actually be scheduled for replacement. In addition, the PRI double counts some performance measures, uses nonlinear and case-based formulas that do not produce a transparent link between measures and priority, and neglects some important maintenance related considerations that influence priority for replacement. Therefore, the purpose of this study is to develop an objective decision-support tool for prioritizing bridge replacement candidates that accounts for the multiple goals and preferences of the Structures Management Unit. Critical criteria and performance measures are proposed through a review of BMS improvements from other states as well as discussions with the Structures Management Unit. Additionally, new performance measures are introduced to incorporate historical maintenance burden and current maintenance needs. The trade-off preferences of the Structure Management Unit for each of the performance measures are modeled with value functions through a developed Excel VBA macro. Data driven prioritization formulas are created through statistical regression of binary logistic and constrained linear least squares models using the statewide bridge inventory to result in utility functions that provide a priority score for each of the replacement projects. The statistical models provide insight on the performance measures that have been statistically linked to bridge replacement projects as well as their relative importance. Analysis of the predictive accuracy for binary classification of projects, distributions of prioritization scores, and odds ratios computed from the predicted prioritization scores are used to compare the performance of the models and arrive at a recommended best model. Risk attitudes are incorporated with logistic regression and constrained linear least squares, resulting in a utility function that provides a priority score for each of the replacement projects. The results of this study seek to provide transparent, defensible ratings for bridge replacement projects that can be used in future budget planning and can be adjusted if the goals of the NCDOT change.