One
of the primary functions of a bridge management system (BMS) is to inform
data-driven, risk-based decision making by forecasting future network level
needs and anticipating the costs and benefits of bridge replacement,
rehabilitation, and preservation actions.
Of these actions, bridge replacement projects account for the majority
of the current funding needs and annual allocations. Consequently, shortcomings in conceptual cost
estimating models used within bridge management systems can impose serious and
potentially costly errors affecting financial needs projections and project
selection and prioritization. Conceptual
cost estimating strategies currently used in the NCDOT BMS are simplified, do
not consider factors affecting construction, preliminary engineering, and right
of way costs, and have not been recently updated to reflect changes in
construction cost trends and inflation.
In this study, cost data for recent bridge replacement projects
completed in North Carolina were sourced and assembled into a database with
information on the characteristics of the replaced and replacement
structures. This database was then used
to evaluate current conceptual cost estimating strategies used by NCDOT,
identify factors influencing construction, preliminary engineering, and right
of way costs, and formulate new conceptual cost estimation models for bridge
replacements. Generalized linear regression
models and decision trees were developed to estimate unit costs for each
component of the replacement cost and cross-validation was used to arrive at
appropriately sized models. The
developed cost estimation models were evaluated by comparing goodness of fit to
the underlying project data as well as assessing the projected unit replacement
costs obtained when applying the developed models to all bridges in the
state. The recommended conceptual cost
estimation strategy uses generalized linear models to forecast unit
construction and unit preliminary engineering costs and a decision tree to
forecast unit right of way costs. The
recommended conceptual cost estimation strategy can be readily implemented
within the existing BMS with few required changes and empirical evidence
suggests that these revised models will significantly improve the accuracy of
the conceptual replacement cost estimates.
