Determination of Bridge Elements’ Weights Using the Random Forest Algorithm

AbstractSignificant bridge inspection data has been collected over the years at the component and element level to improve management practices in the United States. A widely adopted systematic approach to correlate the weight or importance of the bridge elements to the overall bridge performance, which influences the maintenance, repair, and replacement (MRR) schedule and resource allocation for structures, does not exist given the existing data. Some transportation agencies use a cost-based approach to assign weights to bridge elements, which can be in terms of the loss accrued during downtime or the amount needed for the replacement of the element. However, this approach does not consider the bridge elements’ structural relevance to the overall performance of the bridge. This study proposes a novel framework to synthesize component and element-level bridge data to showcase their relationship using the random forest algorithm, which is essentially an ensemble of decision trees to evaluate the importance of different elements relative to the overall condition of the bridge. The analysis focused on eight bridge design types predominant in Delaware, Maryland, Pennsylvania, Virginia, and West Virginia, and analyzed 104,699 bridge records consisting of the condition rating and element-level data from the National Bridge Inventory (NBI). The random forest algorithm showed that bridge elements’ weight (or importance) is not constant as implied by the cost-based approach; rather, bridge elements’ weight varies based on their relevance to the bridge’s structural performance. The resultant bridge elements’ weight, which is the element weight multiplied by the component weight, can be used to improve the existing Bridge Health Index (BHI) equation found in the Manual for Bridge Evaluation (MBE) using this data-driven approach. Given more available component and element-level bridge data, this formulation provides a framework for transportation personnel to determine which set of bridge elements to prioritize in their maintenance actions and ascertain if the elements receiving the highest priority in the MRR schedule and budget allocation are also the same set of elements that bridge inspection reports regard as needing attention. Practical ApplicationsThe United States bridge inventory is made up of several bridge design types with distinct deterioration characteristics based on their structural configuration and needed to make decisions about maintenance and repai